Writers of Pro Football Prospectus 2008

Most Recent FO Features

SmithSte01.jpg

» The Week In Quotes: August 29, 2014

This week: Josh Shaw lies, Steve Smith intimidates, Le'Veon Bell relaxes, Matt Simms dances, and Clint Trickett kisses and tells.

07 Jul 2008

DPAR is Dead

by Aaron Schatz

It's a big day around Football Outsiders. We've now officially put the KUBIAK fantasy football projections on sale. You can purchase them either alone for $20 or in a package with Premium DVOA Database access for $50. (This year, there is no coupon code to get $10 off Premium later on -- you need to purchase them both together at the same time to get the discount.) We've also put the 2008 game charting data in the store; you can purchase that for $60.

We also found out that Pro Football Prospectus 2008 is now shipping from online booksellers. In fact, we even have a reader who found a copy in a Barnes and Noble last Thursday. As longtime FO readers get their hands on America's number one preseason football guide and smash-tastic weapon of mass mosquito destruction, they will page through to look at the comments on their favorite players, and immediately wonder one thing.

Where the hell is the DPAR?

The newfangled statistics we feature here at Football Outsiders are by no means a finished product, and we are always looking for ways to improve them. The best time to do that is usually the offseason, where we can introduce the changes when we publish the new version of PFP. That's what we did when we updated team DVOA to the current "version 5.0" in 2007. This year's project was to work on the individual stats, which needed to be improved in two ways: accuracy and accessibility.

The new version of individual DVOA is based on five years worth of data, 2003-2007, and has a number of changes that will bring them in line with the changes we made the year before in team DVOA.  Individual DVOA ratings now use more complicated (and accurate) opponent adjustments. Passing and receiving DVOA now includes defensive pass interference as a positive play, just like team DVOA does. We found that while there's no correlation from year to year in how often a player gets DPI, the year-to-year correlation of (catches + DPI) is slightly higher than the year-to-year correlation for just catches. In addition, the bonus for plays in the red zone for individual DVOA has been dropped to 10 percent, as opposed to 20 percent for team DVOA. This improved DVOA correlation for players from season to season.

Once we had improved the individual baselines and opponent adjustments, the next step was to improve the accuracy of "replacement level." Originally, we estimated replacement level by simply using a scale similar to the one our partners at Baseball Prospectus use for hitters and pitchers, putting replacement level about three-eighths of the way between average (0%) and the worst starters in the league. For example, since the worst quarterbacks are usually around -35% DVOA, the replacement level was -13.3% DVOA. It was time to come up with something we felt was more accurate than "this is sort of based on what BP does."

For quarterbacks, we analyzed situations where two or more quarterbacks had played meaningful snaps for a team in the same season, then compared the overall DVOA of the original starters to the overall DVOA of the replacements. We did not include situations where the backup was actually a top prospect waiting his turn on the bench, since a first-round pick is by no means a "replacement-level" player. By comparing the replacement-level quarterbacks to the quarterbacks they replaced, as well as the quarterbacks who played the entire season, we determined that the replacement level for quarterbacks is roughly -12.5% DVOA, fairly similar to what we had before.

The same is not true at other positions. There was no easy way to just separate running backs and receivers into "starters" and "replacements," since unlike at quarterback, being the starter doesn't make you the only guy who gets in the game. Instead, we decided to use a simpler method. First, we ranked players at each position in each season by attempts. The players who made up the final 10 percent of passes or runs were split out as "replacement players" and then compared to the players making up the other 90 percent of plays at that position. This took care of the fact that not every non-starter at running back or wide receiver is a freely available talent. (Think of Jerious Norwood or Anthony Gonzalez, for example.) Replacement level is now higher in most ways, but by different degrees, and replacement level actually went down for both running back receiving and wide receiver rushing


Measurement Old New
Passing -13.3% -12.5%
QB rushing -17.1% -12.5%
RB rushing -13.7% -8.0%
RB receiving -9.8% -12.6%
WR receiving -12.7% -7.6%
TE receiving -15.0% -12.3%
WR/TE rushing -26.0% -43.0%

That takes care of the improvements to accuracy, but what about accessibility? This has admittedly been a problem at Football Outsiders since the very beginning. The challenge of any new stat is to present it on a scale that's meaningful to those attempting to use it. DPAR (Defense-adjusted Points Above Replacement) came about because the DVOA system measures success in "success points" that represent yardage and progress towards a first down all wrapped together. Saying that Carson Palmer was worth 106.3 success value points over replacement in 2007 has very little value without a context to tell us if 106.3 is good total or a bad one.

The solution to this problem was to translate "success points" into "actual points." I put together a complicated spreadsheet that estimated how many "success points" ended up in X number of points scored, combining both offense and defense, and we used that estimate to translate success values into DPAR. For example, Carson Palmer in 2007 was worth 51.8 passing DPAR.

Unfortunately, DPAR isn't exactly the easiest thing to understand either. The reason is obvious: Do you know anyone who refers to a player's value by how many points he scored? I mean, we know touchdowns are six points, and we talk about kickers in terms of points when we talk about fantasy football, but people don't think of football players in terms of points. They think in terms of yards.

Therefore, along with the new set of replacement baselines comes a new way to think about Football Outsiders' advanced stats for individual players: DYAR, or Defense-adjusted Yards Above Replacement. Carson Palmer's season is now worth 1,215 DYAR, which ranked sixth in the NFL last year. By comparison, Jeff Garcia's season was worth 694 DYAR, Rex Grossman's season was worth -168 DYAR (yes, below replacement level), and Tom Brady's season was worth an all-time record 2,788 DYAR. Turning "success points" into yards was actually easier than turning "success" into points, since we happen to have tangible yardage totals for all the players we are measuring.

Here's a comparison of the top players in 2007 according to the old method (DPAR) and the new method (DYAR).


Old Method
Player DPAR
12-T.Brady 200.6
18-P.Manning 133.5
4-B.Favre 102.9
9-D.Brees 100.6
9-D.Garrard 91.2
9-T.Romo 89.1
9-C.Palmer 84.6
8-M.Hasselbeck 78.2
7-B.Roethlisberger 75.2
6-J.Cutler 72.1
New Method
Player DYAR
12-T.Brady 2,788
18-P.Manning 1,841
4-B.Favre 1,438
9-T.Romo 1,297
9-D.Brees 1,285
9-C.Palmer 1,215
9-D.Garrard 1,086
6-J.Cutler 972
8-M.Hasselbeck 937
3-D.Anderson 797

There are a number of reasons why players shift around a little with the new DYAR values. Drew Brees is hurt by the slight rise in replacement level, because he had so many pass attempts. The more complex, more accurate opponent adjustments make the difference between DVOA and VOA smaller for some players (Tony Romo) and larger for others (Ben Roethlisberger, who now ranks 11th). Changes in the baseline for various down-and-distance situations will have a small impact on every player, in ways that are hard to see without breaking it all down. Of course, small changes in our stats should not be treated as gospel truth. We believe the new individual baselines and formulas are more accurate, but it is possible that the next time we improve things, David Garrard's 2007 season will move back ahead of Tony Romo's 2007 season. So don't flip out when, for example, the 1995 quarterback stats finally go up and you see that Brett Favre's MVP season, which was number one when I first ran the numbers, now has a passing value 30 DYAR behind Scott Mitchell and 22 DYAR behind Erik Kramer.

The changes are larger when we look at old individual DVOA and new individual DVOA, primarily because improving opponent adjustments tends to move extremely high and low DVOA ratings towards the mean. Minimum 100 passes here, as usual.


Old Method
Player DVOA
12-T.Brady 62.1%
9-D.Garrard 49.5%
15-T.Collins 49.3%
18-P.Manning 43.9%
4-B.Favre 29.7%
7-B.Roethlisberger 28.0%
7-J.Garcia 25.9%
9-T.Romo 24.7%
18-S.Rosenfels 21.8%
6-J.Cutler 20.5%
New Method
Player DVOA
12-T.Brady 56.9%
18-P.Manning 40.6%
9-D.Garrard 37.4%
15-T.Collins 31.6%
4-B.Favre 28.0%
9-T.Romo 25.5%
18-S.Rosenfels 24.3%
5-Q.Gray 20.9%
9-C.Palmer 20.1%
7-J.Garcia 19.5%

Translating player performance into yardage rather than points allows us to introduce one more statistic, which we hope will make Quick Reads much more accessible to the general population of football fans. The new stat is called Equivalent Yards (EqYds). It is very simple: Equivalent Yards (EqYds) are DYAR with replacement-level performance added back into the total. This provides an easy comparison: in general, players with more Equivalent Yards (EqYds) than standard yards played better than standard stats would otherwise indicate, while players with fewer Equivalent Yards (EqYds) than standard yards played worse than standard stats would otherwise indicate.

(Note July 8: Unfortunately, as some readers have noticed, Equivalent Yards (EqYds) does not seem to be showing what we thought it was supposed to show. We're looking into how we can fix it.)

Here's the same list of top ten quarterbacks from 2007 in DYAR, along with the top ten quarterbacks in Equivalent Yards (EqYds) and in standard passing yards.


Top 10 Passing Yards
Player Pass Yds
12-T.Brady 4,806
9-D.Brees 4,423
9-T.Romo 4,211
4-B.Favre 4,155
9-C.Palmer 4,131
8-J.Kitna 4,068
18-P.Manning 4,040
8-M.Hasselbeck 3,966
3-D.Anderson 3,787
6-J.Cutler 3,497
Top 10 Equivalent Yards (EqYds)
Player True Yds
12-T.Brady 6,428
9-D.Brees 5,355
18-P.Manning 5,008
4-B.Favre 4,703
9-C.Palmer 4,668
9-T.Romo 4,439
8-M.Hasselbeck 4,425
8-J.Kitna 3,894
3-D.Anderson 3,824
6-J.Cutler 3,808
Top 10 DYAR
Player DYAR
12-T.Brady 2,788
18-P.Manning 1,841
4-B.Favre 1,438
9-T.Romo 1,297
9-D.Brees 1,285
9-C.Palmer 1,215
9-D.Garrard 1,086
6-J.Cutler 972
8-M.Hasselbeck 937
3-D.Anderson 797

Obviously, Equivalent Yards (EqYds) are not the best way to measure value because they are heavily dependent on usage. That's why Jon Kitna and his 616 pass attempts (including DPI and sacks) rank eighth in Equivalent Yards (EqYds). The fact that Kitna has more actual yards than Equivalent Yards (EqYds), however, gives a clear indication that Kitna's performance was generally below average (-1.2% DVOA, to be exact).

For those curious, the replacement level is 4.93 yards per pass attempt. However, it is 5.63 yards per equivalent pass attempt, "equivalent pass attempts" being the translation of the "league-average" baseline on all that quarterback's passes into a pass attempt number. The reason for the difference is that there end up being more "success points worth of pass attempts" than there are actual total pass attempts. Yes, this is complicated, but the good news is that nobody needs to see the gears behind the curtain on a week-to-week basis.

The colors in that table reflect that numbers on the player stats pages will now be color-coded. DYAR and DVOA, the main opponent-adjusted stats, will be blue. Equivalent Yards (EqYds) will be red. Everything else will be the usual black.

How will this affect Quick Reads? The plan is to still list players by DYAR, but to give an example, here are the top quarterbacks from Week 14 of 2007 in standard passing yards.

That was the week Peyton Manning eviscerated Baltimore on Monday Night Football, with 249 yards on just 17 passes, while Tom Brady threw for nearly 400 yards against Pittsburgh. You can see how things change when we look at the same quarterbacks through the lens of Equivalent Yards (EqYds) and DYAR.


Player Opp PaYd Comp Att TD INT Sack True
Yards
DYAR
12-T.Brady PIT 399 32 46 4 0 0 569 285
13-K.Warner SEA 337 28 46 4 5 5 337 20
9-D.Brees ATL 328 28 41 3 0 0 429 169
9-T.Romo DET 302 35 44 2 0 3 404 107
8-C.Redman NO 298 23 39 3 1 3 220 -15
14-B.Griese WAS 295 27 43 0 2 2 296 42
11-K.Clemens CLE 286 24 39 0 2 4 199 -48
8-M.Hasselbeck ARI 272 22 32 4 0 2 342 142
12-L.McCown HOU 268 25 38 0 0 4 252 -3
4-B.Favre OAK 266 15 23 2 1 0 248 119
18-P.Manning BAL 249 13 17 4 0 0 322 228
8-J.Kitna DAL 248 22 36 0 0 1 299 87

It might sound like things will get confusing with all these stats, but it won't, because not every one of these stats will exist from here on out. As of August, Football Outsiders will never be using DPAR again. Pro Football Prospectus 2008 lists all players in DYAR, and we'll list all players with DYAR and Equivalent Yards (EqYds), not DPAR. Right now, the 2007 stats pages have been changed to represent DYAR instead of DPAR. You can take a look at quarterbacks, running backs, tight ends, or wide receivers. Over the next couple weeks, I'll be writing a series of articles looking at the best and worst seasons and games of the DVOA Era (1995-2007) according to the new stats. As we do each position, we'll put up all the DYAR and Equivalent Yards (EqYds) stats going back to 1995. Yes, finally.

Posted by: Aaron Schatz on 07 Jul 2008

230 comments, Last at 17 Aug 2008, 8:32am by Ronjan

Comments

1
by Mikey (not verified) :: Mon, 07/07/2008 - 4:17pm

So in short, true yards measures how many yards a QB would have passed for against a replacement level defense that day?

2
by Tom D (not verified) :: Mon, 07/07/2008 - 4:19pm

But how do I pronounce it?

3
by Tony (not verified) :: Mon, 07/07/2008 - 4:30pm

2: You pronounce it "jar." Sorta like how "D'Yer" isn't pronounced duh-yer.

4
by KR (not verified) :: Mon, 07/07/2008 - 4:36pm

2: I prefer a pirate-y approach to pronunciation, with Dee-Yarrr.

5
by admin :: Mon, 07/07/2008 - 4:38pm

It was a common joke when I started doing this that Tampa Bay always leads the league in YAR.

By the way, I'll try to answer the inevitable DYAR questions that will be coming in this discussion thread, sometime in a mailbag later this week.

6
by inkakola (not verified) :: Mon, 07/07/2008 - 4:45pm

no, its dee-yar and say the 'yar' like the sea captain from the simpsons.

"deyaaaaaaaaaar"

7
by Theo, Netherlands (not verified) :: Mon, 07/07/2008 - 4:45pm

D'Yarrrr is great!
Because that's Pirate and Pirates are the chosen people.
There is a strong correlation between the decline of Pirates since the 1800's and global warming since then.
RAmen

8
by Yaguar (not verified) :: Mon, 07/07/2008 - 4:52pm

I second those who prefer the pirate pronunciation.

9
by PackMan (not verified) :: Mon, 07/07/2008 - 5:14pm

+1 for the pirate pronunciation.

10
by LeverageLover (not verified) :: Mon, 07/07/2008 - 5:15pm

Still no WinEx or leverage-adjusted stats.

As they say on the north side, "there's always next year."

11
by jimm (not verified) :: Mon, 07/07/2008 - 5:15pm

I like the new stat and adjustments.

It seems to fit more closely with my perception of how individual players performed on the Vikings.

12
by Independent George (not verified) :: Mon, 07/07/2008 - 5:22pm

Because a good 'yar!' should come from the back of the throat, shouldn't there be an umlaut? And if there's an umlaut, shouldn't it be printed in a gothic font?

13
by NewsToTom (not verified) :: Mon, 07/07/2008 - 5:23pm

If we're going with the pirate pronunciation for DYAR, can we rename KUBIAK to SYMONS? KUBIAK has always bothered me, because my mother, who'd be hard-pressed to name more than 10 NFL players, recognized immediately it referred to Gary Kubiak, who was John Elway's backup. With Kubiak now being an NFL head coach and thus not at all obscure, renaming it seems appropriate.

And who better to rename it for than B.J. Symons?
(1) He's obscure, and will likely remain so.
(2) He was a draft pick of the Houston Texans, the team Kubiak now coaches.
(3) He played his college football at Texas Tech for noted pirate aficionado Mike Leach.
(4) TTU QBs put up ridiculous fantasy stat seasons, without necessarily being particularly good players.
C'mon, who's with me on this?

14
by JasonK (not verified) :: Mon, 07/07/2008 - 5:36pm

#13: Clever, but SYMONS fails for one important reason: unlike KUBIAK (and PECOTA, for that matter), SYMONS doesn't sound like a name of a computer from the '40s.

15
by Yaguar (not verified) :: Mon, 07/07/2008 - 5:42pm

KUBIAK projections:

Tom Brady is listed as Red (High) risk.

I think you probably rushed the spreadsheet out the door, which is fine, but I hope a fix is in the making.

16
by Ryan (not verified) :: Mon, 07/07/2008 - 6:12pm

15: Why is that bad? Red means he is likely to fall short of the projection for whatever reason (not just injury).

17
by Yaguar (not verified) :: Mon, 07/07/2008 - 6:17pm

16: His projection is for a reasonable 37 touchdowns, and his overall point total is only a couple higher than that of Manning, who is Green. Brady is obviously not meant to be Red, which is used for players likely to miss playing time due to injury or benching. Neither one has happened to Brady in his career. Also, Brady was ranked above Manning in the KUBIAK preview posted a while ago. It's an oversight.

I have more KUBIAK questions and complaints and stuff, but I'm not sure if this is the right venue.

18
by David (not verified) :: Mon, 07/07/2008 - 6:23pm

Rex Grossman played below replacement level? C'mon, you didn't need all that newfangled mathematics to figure that one out, did ya?

19
by Crabbie (not verified) :: Mon, 07/07/2008 - 6:34pm

I'll admit to being diappointed with the shift to DYAR - I always thought that DPAR was pretty intuitive, and hasn't one of the site's frequently made points been that yardage is a very flawed statistic for making comparative judgments about players? Sure, this is "DYAR" or whatever, but that's pretty nonsensical to me. What are these yards that are better or worse than actual yards? "True" yards seems silly - a true yard is 36 inches. When Tom Brady completes a pass, do his superhuman powers over space and time shorten that yard to 30 inches, making it easier for his team to march down the field?

I mean its just a collision of the abstract and the concrete that strikes me as being incredibly silly, and a needless attempt to dumb something down that wasn't particularly esoteric in the first place. It actually makes a hell of a lot more sense to me to think of Carson Palmer as, over the course of a season, being about ten touchdowns better than a replacement quarterback, rather than 1,215 imaginary yards better. Maybe this is because this site has done such a good job of training me not to look at yardage, but if you wanted to see how many [b]yards[/b] better one player is than another, wouldn't you just look at, y'know, their actual yardage and yards per attempt numbers?

The more I think about it, the dumber it seems. There's really no arguing about what constitutes a yard, whereas there's plenty to be parsed out when it comes to trying to figure out how much some players contribute (or fail to) to the success of their offenses. Which is measured in points.

Anyway, that's just the 2 cents of a longtime reader and rare commenter. It's nice to see the DVOA tweak, which with the given examples corresponds more closely to my own intuitive judgment of qbs last year than the previous numbers (and I guess corrolates more nicely).

20
by Brian (not verified) :: Mon, 07/07/2008 - 6:36pm

How does anyone know if any of the stats make sense? Do they double count things? What aspects of performance are used, and how are they weighted? Etc?

21
by Monkey Business (not verified) :: Mon, 07/07/2008 - 6:41pm

Brady would be listed as a high risk for missing his projected TD total. There's no doubt that he's an excellent QB, but he's never been a stat machine. For a guy coming off a 50 TD season and a devastating loss, I wouldn't be surprised if he breaks 30 TDs.

The reason Manning is listed in green is because he routinely generates big numbers, and is realistically capable of breaking the single season TD record every season.

22
by Becephalus (not verified) :: Mon, 07/07/2008 - 6:53pm

I would second the confusion a FO thinking "yards" was somehow easier to understand than points. Sounds like a statistic fox would cook up. PAR is pretty damn straightforward if you ask me.

Alas it doesn't really matter to me I guess as I will understand what you are getting at either way.

23
by Goo (not verified) :: Mon, 07/07/2008 - 6:54pm

I second #19. I picked up PFP08 yesterday and haven't put it down since, but I'm not sure I like DYAR -- isn't the whole point to move beyond yards and think about what actually helps teams win games? You don't win by getting more yards than the opponent, you win by scoring more points, and that was one of the big appeals of DPAR. I like the adjustment to DVOA and the replacement level, but I think you've unintentionally muddied the water by switching from DPAR to DYAR.

24
by mikeabbott (not verified) :: Mon, 07/07/2008 - 6:58pm

It'll be fun to see what the mainstream audience thinks of this.

25
by Yaguar (not verified) :: Mon, 07/07/2008 - 6:59pm

19: I agree to some extent. One question is what sort of yards these are. A yard that counts towards a new first down is far more valuable than a yard that doesn't. The difference between three and four yards on first and ten is huge, whereas the difference between a 23 yard pass or a 24 yard pass on first and ten is negligible. Points are points, but yards have very situational value.

21: Again, it's true that Brady's projection is pretty unstable; he could go anywhere from 28 to 48 TDs and not surprise anyone. But that's not what the Red risk level has meant in the past few KUBIAK spreadsheets.

26
by MCS (not verified) :: Mon, 07/07/2008 - 7:11pm

It's disappointing that I order my copy of PFP months ago and here I sit waiting while all I had to do was wait for it to come out and go to the bookstore.

I pre-ordered and pre-paid.

27
by Crabbie (not verified) :: Mon, 07/07/2008 - 7:12pm

That's exactly the thing, thanks for putting it that neatly - yards are situational in value, whereas points are static. Which is what makes DPAR a far more comprehensible way of presenting the statistic to me.

28
by Becephalus (not verified) :: Mon, 07/07/2008 - 7:15pm

Here is another dirty little secret MSC. Casual customers are a lot more important and treated better than loyal ones (Obviously this isn't 100% true, but it generally is).

Companies are out to get the most money, not give out the most satisfaction or fairness. As long as you are just satisfied enough to be a customer, thats good enough.

29
by MCS (not verified) :: Mon, 07/07/2008 - 7:25pm

Wow, does 26 sound crappy. It was intended to be somewhat backhanded but not as strong as that. Sorry.

BTW. I wait for my copy of the book so I can deep dive these new stats a little more before rendering my opinion.

30
by jsa (not verified) :: Mon, 07/07/2008 - 7:30pm

Count me among those not happy with the change.

You ask: "Do you know anyone who refers to a player’s value by how many points he scored?"

One of the things I love about this site and DPAR is that it gave a way to compare how many points a player contributed compared to another player. So that one could say "Tom Brady's performance last year contributed about 120 more points than Carson Palmer's." That makes sense to me, it tells me in terms I can easily understand how much a player helped his team compared to another.

This DYAR stat confuses me. Of course it could be because it's new. But one of the basic lessons I heard early on from FO is that yards is not a good basis of measuring players; instead, by considering performance in the context of the game situation and opponent, we can measure how much a player helps his team score points. If this were baseball, it would be like telling me how many more adjusted hits a player got rather than his VORP.

I assume the FO staff deliberated on this decision and isn't about to reconsider, but I wish that they would.

31
by Stevie (not verified) :: Mon, 07/07/2008 - 7:34pm

Stats are stupid

32
by MCS (not verified) :: Mon, 07/07/2008 - 7:37pm

Loyal is right Becephalus. I'm such a flunkie that, when my wife threw out PFP2005 during our last move, I made her buy me another copy.

33
by anotherpatsfan (not verified) :: Mon, 07/07/2008 - 7:51pm

It will be interesting to watch the debate of DYAR, and I hope and expect Aaron will answer some of those questions being asked.

However, there is no need to debate which was the best way to buy the book, or imply that FO is somehow responsible for the fact that someone preordered and is still waiting their book (or imply that FO is somehow mistreating loyal customers versus casual customers). You either waited to get the book at the bookstore at $21.95 plus tax and gas, or you preordered it online from Amazon, etc. for about 14 bucks (and likely didn't pay much if anything for shipping). However you decided to get PFP 2008, you didn't pre- or otherwise order it from FO.

Whatever truisms there may be about casual versus loyal customers, and while the casual customer may help the number of books sold by online and B&M booksellers and thus help indirectly help FO, FO is still at the mercy of the publisher and the retailers as to when the book gets in your hands. It seems this might be the rare case where the loyal customer is the key to the success of the business, as ultimately FO probably makes more money from the non-casual visitors here who buy Kubiak, Premium, etc., which are likely beyond the interests of the "casual" FO visitor from Fox message boards or elsewhere.

34
by MCS (not verified) :: Mon, 07/07/2008 - 8:19pm

Relax Yetanotherpatsfanonapatsfanladenintrrnet, it's not nearly as serious as you make it out to be. I've been floating around this site since 2003 and I preorder every year. Ain't no thing.

35
by Mike Tanier :: Mon, 07/07/2008 - 8:21pm

If it makes anyone feel better, I haven't gotten my copy of the book yet.

36
by LearnFromTheMasters (not verified) :: Mon, 07/07/2008 - 8:22pm

Funny that Baseball Prospectus is mentioned in this article. While they provide stats like EQA and QERA (which are on the same scale as batting average and ERA, respectively), the stats they use for overall player comparisons are from the VORP/WARP category -- expressed in runs ("points") and wins ("tens of points").

This may be because, as noted above, scoring/winning is more intuitive/important than "stats that look like stats everyone already knows, except not."

37
by Luz (not verified) :: Mon, 07/07/2008 - 8:28pm

...long live DPAR!

38
by sam! (not verified) :: Mon, 07/07/2008 - 8:37pm

Count me as a longtime follower who prefers points to yards.

39
by Lou (not verified) :: Mon, 07/07/2008 - 8:52pm

i feel the need to echo the sentiments of Crabbie and others. To me DPAR was simple to understand, where the value of a yard changes in context, the value of a point rarely does.

as others have pointed out BP has VORP and WARP. So why scrap DPAR? why not just print it next to the new stats?

40
by Thok (not verified) :: Mon, 07/07/2008 - 8:58pm

Shouldn't it be Yards Adjusted for Replacement and Defense, a.k.a. YARD?

41
by Raiderjoe (not verified) :: Mon, 07/07/2008 - 9:18pm

Book says Raiders have 0% chance to contend for Super Bowl and they have 4% chance to be playoff contender.
This going to be very embarasing when Raiders do make playoffs.
Book also says Russell going to throw 20 ints. No way does that happen. Jamarcus ussell is very accurate passer. More like 12 ints.

42
by Andrew (A.B.) (not verified) :: Mon, 07/07/2008 - 9:26pm

I'll second (third) the disappointment with the death of DPAR.

Frankly, I wish more of the advanced stats had migrated toward a "points" framework, instead of moving in the opposite direction. As someone pointed out, it's parallel to the BP use of "runs" in many of their metrics.

43
by Tom D (not verified) :: Mon, 07/07/2008 - 9:27pm

I like DYAR, it does a much better job of showing the differences between players. DPAR was too compressed.

44
by Devin McCullen (not verified) :: Mon, 07/07/2008 - 9:28pm

Here's a quickie for the mailbag, although it's not YAR-related. You mention that Eli Manning had the second-worst completion percentage over a 5-game stretch since 1995 (among QBs with at least 15 attempts). So who's #1?

45
by Yaguar (not verified) :: Mon, 07/07/2008 - 9:55pm

44: I'd bet it's Ryan Leaf, starting with that legendary awful game.

40.3% over five games.

46
by Bill Barnwell :: Mon, 07/07/2008 - 10:14pm

That's actually a research error on my part that didn't get found until after the book. Manning's completion percentage for the five games was actually eighth-worst of all-time, not second.

Leaf was indeed first at 40.3%. Eli in 2004 is seventh.

47
by Gerry (not verified) :: Mon, 07/07/2008 - 11:04pm

Bill, any chance at the rest of that list?

Eli Manning. Worst 'good' quarterback ever, or best 'bad' quarterback ever? Discuss.

48
by The Ninjalectual (not verified) :: Mon, 07/07/2008 - 11:04pm

Is it April Fools Day again? Or did FOX get their sinister, dirty noses into FO business? I wouldn't blame anyone for selling out though, if the price is right.

It may be just a fear of anything new and unfamiliar, but I don't trust this DYAR. I liked and understood DPAR; it made intuitive sense to me even when I first encountered it. It took some time to get a feel for what the scale of what numbers meant, bit I knew immediately that it was measuring something important. This DYAR doesn't have the same appeal; like others have said, this stat doesn't really add anything new that traditional stats--interpreted through DVOA--didn't already.

Like it was said above, "scoring/winning is more intuitive/important than “stats that look like stats everyone already knows, except not.” Now I will spend all season wondering what Clinton Portis' DPAR would have been, and I feel like I've lost an important tool in assessing how much Jason Campbell is improving. Yards don't do it for me anymore!

I also second that YARD (Yards Adjusted for Replacement and Defense)is a better acronym.

49
by Vick (not verified) :: Mon, 07/07/2008 - 11:20pm

"Unfortunately, DPAR isn’t exactly the easiest thing to understand either. The reason is obvious: Do you know anyone who refers to a player’s value by how many points he scored? I mean, we know touchdowns are six points, and we talk about kickers in terms of points when we talk about fantasy football, but people don’t think of football players in terms of points. They think in terms of yards."

The whole point is about the points and not the yards. Why do we need FO if we have to still go by people thinking ? Maybe, DPAR wasn't accurate enough. Couldn't you guys could do salvage point-based measurement while improving its accuracy ?

50
by shake n bake (not verified) :: Mon, 07/07/2008 - 11:20pm

obviously points are more important to outcomes, but how many people intuitively know how many points an offense usually scores over a season. I don't and I consider myself a pretty hardcore football stat guy. Football fans know that 4,000 yards (or DYAR) is a solid QB season, what does 75 DPAR tell you without the context of the a DPAR ranked list? It's above replacement level. Is it mediocre or very good? I honestly don't know because I pulled 75 out of thin air. I'm going to check as soon as I post this because I really don't know.

51
by Raiderjoe (not verified) :: Mon, 07/07/2008 - 11:22pm

I dont know what DPAR and DVOA and DYAR is, I just read the comments in the book.

52
by Goo (not verified) :: Mon, 07/07/2008 - 11:34pm

Yeah, but nobody has 4000 DYAR either -- Brady's record-breaking year was apparently worth 2788, and only a handful of guys were over 1000... So DYAR isn't really on a familiar scale, either. Now, for a familiar scale you supposedly look at true yards, but like Aaron says, those are too dependent on attempts to be used to measure value.

You know, at least with DPAR I could subtract it from a team's point total or swap players and use the old pythagorean formula to run "what ifs" with different players; how do you do that with DYAR? The more I think about it, the more I see this as a move backwards, taking FO's signature individual stat and needlessly making it worse.

53
by The McNabb Bowl Game Anomaly (aka SJM) (not verified) :: Mon, 07/07/2008 - 11:47pm

Re: 50

So the average fan knows how many yards is "good" for a QB, RB, or WR, and therefore "True Yards" is more intuitive than DVOA. Hmm, I'll have to think about that. It's hard for me to think about rating players based on yards because I've been conditioned not to. Seems like this is a concession to people who insist on counting yards in their assessments- "You want to count yards? OK, here are some more accurate yards."

I definitely liked that DPAR represented points. That enabled us to see concretely how much a given player added to the offense. I realize you can do that with yards, but yards don't win games, points do. With DPAR, we used to be able to say, "That RB was critical to his team's win. A replacement back wouldn't have generated enough points for them to win." Or, "That team could be winning every game if only they had a replacement-level QB, but their crappy starter is sucking the points away." Now, of course this wasn't perfectly accurate since the teams don't generate points precisely according to DVOA on a game-to-game basis, but it was useful.

How can we do that with yards? A RB generates a certain number of DYAR. Great, he got more yards than a replacement would have. But what does that mean for the outcome? Points impact the outcome, yards do not.

Bottom line, don't scrap DPAR yet. How about a trial run for Quick Reads using both stats, to see which one people like more?

54
by pressrow (not verified) :: Mon, 07/07/2008 - 11:56pm

I love how we have an ultra-researched post about mathematical equations and accuracy, and the first chunk of comments are all about pirates. Love it.

On a more serious note, does this mean that "dpar" has been moved from the list of security words to leave comments?

55
by Tom D (not verified) :: Tue, 07/08/2008 - 12:21am

Re 53:

I actually think yards is much better way to represent how much a player helps his team. It goes back to the 99 yard drive capped with a 1 yard TD. People intuitively realize the TD wasn't as important as the yards in this situation.

Assigning how many points a play is worth was always kind of arbitrary because a great player can be a on a bad team, and generate a lot of yards, but not reach the end zone.

56
by D (not verified) :: Tue, 07/08/2008 - 12:25am

I'll reserve final judgment until I get the book, but based on this article I have to put myself in the pro-DPAR faction. Points simply makes more intuitive sense to me in determining a players value than yards. #19's comment about Carson Palmer being worth 10 TDs more than a replacement QB being easier to envision than 1200 yards better. It probably doesn't help that I'm used looking at teams point differential to determine its real value.

57
by Bill Barnwell :: Tue, 07/08/2008 - 12:29am

Sure thing, although I'm kinda swamped, so you just get the final week of the streak listed. Note that I don't include extended streaks, so although Ryan Leaf would technically be #1 and #2 since he has wks 1-5 as well as weeks 2-6 that qualify, I'm only including each player's worst non-overlapping five-game stretch:

- Ryan Leaf wk6 1998 40.3%
- Stoney Case wk5 1999 44.0%
- Aaron Brooks wk16 2002 44.1%
- Akili Smith wk6 2000 44.2%
- Kerry Collins wk15 1995 44.3%
- Eli Manning wk15 2004 44.4%
- Mike McMahon wk15 2005 44.4%
- Eli Manning wk16 2008 45.1%

58
by Tom D (not verified) :: Tue, 07/08/2008 - 1:00am

Re 57:

The lack of Rex Grossman on that list surprises me.

59
by PaulH (not verified) :: Tue, 07/08/2008 - 1:29am

Put me on the anti-DYAR list.

A bit of background first... I discovered FO in 2006 after I heard it mentioned while some friends were talking about BP. I checked it out, and absolutely loved it immediately. I was quickly hooked, and I check this place almost every single day, not to mention reading all of the PFPs, and the Year of the Scab.

I was born and raised in the heart of Alabama, and after being named after Paul "Bear" Bryant, it's fair to say that football is something of a religion with me. I've always loathed the talking head "experts" that fill the football landscape, and FO is one of those few places that you can truly get great insight and research from.

However, as many have mentioned earlier, I think the move to DYAR is a backwards step for FO. This place has gotten bigger by the day, and it has done so with DPAR and DVOA at the core. Changing DYAR just isn't a positive step, in my opinion. One of the cornerstones of this site has been that yards constitute a relatively meaningless number that was a poor indicator of value and efficiency. And now we're going to do a 180 by emphasizing yards... I just don't get it.

Beyond that, who was complaining about DPAR in the first place? I know I certainly wasn't, and I don't know of many people that were. I can say firsthand that it was / is much more intuitive to me than this whole DYAR stuff. And I know no one discusses player value in terms of points, they all talk about yards. That's exactly why I read FO! As mentioned earlier, it's a respite from the talking head "experts" and ignorant fans with no knowledge whatsoever of the game.

I'm still going to keep reading FO regardless, and I'll be glad to get my hands on the new book, but in my humble opinion this is a step backwards. Ditch DYAR and bring back DPAR.

60
by disco dack (not verified) :: Tue, 07/08/2008 - 1:36am

quinn gray is now top 10 in DVOA?
I may have a foggy memory of jacksonville football last year but quinn gray was bad. He had two great games where he threw for 300 yards and had 6 td's to just 3 int's. However, every other game, against good secondaries, he struggled. Someone needs to help me understand how Gray's value per play was so good.

61
by JPS (not verified) :: Tue, 07/08/2008 - 1:41am

I have to vote in favor of True Yards. We all know from experience that 450 yards passing is a monster game, and that 250 is only decent. So if a QB passes for 450 yards in a game but only earns 250 True Yards, we know what's being said. In contrast, if a QB throws three TDs but has only a DPAR of 12.7, we understand that the actual performance was worse than the points scored, but for me there's a much less intuitive feel about the statistic. For me, 12.7 points only has meaning on the average, on a per-game basis averaged over several games, whereas saying "yards" appeals to my intuition directly; I don't have to go through an "average" filter.

62
by Alex (not verified) :: Tue, 07/08/2008 - 2:08am

How does anyone know if any of the stats make sense? Do they double count things? What aspects of performance are used, and how are they weighted? Etc?

I don't know how long you've been here, but if you're new, maybe you missed this, where you can scroll down to "Does DVOA really work?" Or, maybe your questions would be better answered by going here, which goes into some detail about how the stats made.

as others have pointed out BP has VORP and WARP. So why scrap DPAR? why not just print it next to the new stats?

I agree, why can't we keep DPAR and bring in DYAR? What, is DPAR going to get jealous? Did DYAR make you put a clause in its contract so that it would be the only advanced statistic measuring value above a replacement player, or its entire contract would be guaranteed? D*mn poison pills clauses.

Seriously, though, having DPAR is nice, not just because it's intuitive/more directly tied to winning, but as mentioned before, you can use it in conjunction with Pythagorean wins to make rough estimates of the change in winning percentage of switching certain players. That's a lot of fun, and you can't do that with DYAR (or at least, it's not as easy). So why can't we at least see the DPAR alongside the DYAR? Why take away a good measurement?

Still, I'll say this for DYAR, it does give us the pretext for talking like pirates and making ridiculous jokes about this fact, so it can't be all bad.

It may be just a fear of anything new and unfamiliar, but I don’t trust this DYAR.

Oh, don't worry, that's just perfectly normal paranoia.

I also second that YARD (Yards Adjusted for Replacement and Defense)is a better acronym.

Does YARD give us a ready-made excuse to talk like pirates? Didn't think so.

63
by qsi (not verified) :: Tue, 07/08/2008 - 2:35am

Hm. I don't comment much here, but I've been reading FO for so many years now that I can't even remember when I first found this site.

My first reaction to reading the "DPAR Is Dead" title was one of excitement: FO has invented something even better! My second reaction to DYAR, after reading about it, is that I don't like it. I've become very comfortable with DPAR, so there may be some status quo bias in my reaction. The rationale for moving to DYAR (making it more intuitive) is getting it exactly wrong: DPAR works as an better stat because it is not tied to old notions of yards, and you can evaluate it without the mental baggage that yards bring. This is an advantage.

Not only that, I also ended up confused by the nomenclature of DYAR: "True Yards" is a horrible name because they're not true yards. The true yards are the traditional stats you see plastered on TV. At least, that is how my DPAR/DVOA-centric mind works now.

I am not opposed to introducing DYAR and DAFY (Defense-Adjusted Fake Yards?) in addition to DPAR, but getting rid of DPAR strikes me as very wrong.

64
by Victor (not verified) :: Tue, 07/08/2008 - 2:48am

Points are also relative. When you are down by 5 with one minute let, you'll be given 3 for free. Similarly, when you are down by 2 with almost no time left you'll take the 3 even if you were 51% as likely to 6 had you gone for that. So the "relative"-ness of yards is a problem that also applies to points.

If you wanted a real comparison you might have to have some notion of an individual contribution to wins and then you could say, "If you replaced Carson Palmer with an average replacement qb, the bengals would have won x fewer games". That also means a team with an average replacement qb, (the Bears if we are being charitable) know how many wins it would be worth to them to trade for Carson Palmer. They would also know if it was worth trading another player at another position for Carson, because wins are wins. Wins are also relative, though. The difference between 1 and 2 wins is a lot less important than the difference between 8 and 9 wins, for example.

The only real measure would be something like Defense adjusted Super Bowl Wins above replacement. Because so few teams win the super bowl each year, though, the confidence intervals would be huge (that means you can't trust the stat at all) and the percentages would be infinitesimal (how much more likely does Anthony Gonzalez make a team to win the SB? )

But DYAR has virtues that DWAR and DPAR don't. Yards happen all the time, and its a lot easier to assign them to individuals, because they happen on every play. If you were Willie Parker at a time when Jerome Bettis was the RB on all redzone plays, you could conceivably be a starter who was never on the field for points, and that would skew the stats. Yards provide a better measure. I don't think they are perfect, but if you want to compare Joe Jurivicious to Eric Parker, its a much better because the confidence on the stat is much higher.

Count me as willing to give it a season to see how it goes, and appreciating that you guys are still thinking about this stuff.

Also, here's a suggestion for looking at Replacement Level. By revealed preference, all teams prefer their staff at their salaries to the best unsigned free-agent at a position. It would be interesting to look at the DYARs of unsigned free agents against those of players with jobs and figure out how much a team is wiling to pay for DPAR, or conversely, how much a FA has to lower his expectations before he gets picked up to replace some cheap guy who is worse than he is (Kevin Jones is better than a lot of back ups in this league, even with his age, how much can he ask for and still get a job?)

65
by Darth Goofy (not verified) :: Tue, 07/08/2008 - 2:57am

I fall into the group liking the DYAR stat. I think a lot of confusion is coming from FO initially indicating that you can't necessarily judge a player by the yards they earn, because of team and opponent factors. For example, Tom Brady could pass for 400 yds and 6 TDs against the Raiders and Peyton Manning could pass for 200 yds and 1 TD against the Jaguars. Everybody would say that TB had a better game, but look at his competition. (Intentional barb for RaiderJoe.) I think what FO is trying to do is use a stat that appeals to the average joe (another barb), uses the basis of DPAR and tries to amalgamate the two into a usable statistic as a foundation for further research.

I like the stat and look forward to this year with D`YAR (best pirate impression) and it's accompanying analysis.

66
by phill (not verified) :: Tue, 07/08/2008 - 3:47am

re 57 - the list of 8 worst QB 5-game stretches in terms of completion percentage (during a new moon while geese are migrating). Is it just random that 5 out of 8 on the list end on week 15 or 16 - is this a case of QBs (or teams generally) being shell-shocked by the end of a crappy season / giving up on a crappy season, or just one of those random things.

Re DYAR vs DPAR - DYAR I can live with. I was fine with DPAR. I tended to think in terms of DPAR as being essentially DVOA multiplied by no. of plays anyway, on a rather arbitrary scale rather than in real 'point' terms. So DYAR is basically just DPAR mutliplied by an arbitrary constant - simply changing the scale.

True Yards seems conceptually confused to me though. You measure how many extra yards a player has gained vs what a replacement level player would do in that position. Okay. Replacement level player would gain RLP yards. Real player gains RLP + DYAR yards (well, RLP + YAR strictly). Then you produce a stat of his actual yards plus DYAR, which is RLP + 2 DYAR. Which seems pretty meaningless. If a player is X above replacement level, then why do we want to know what X above his actual level is - 2X above replacement level?

True yards is the answer to the question "How many yards would player X throw for in a game in which a replacement level player threw would as many yards as player X did in a given game?" Or to put it more simply... Huh?

What would make more intuitive sense is "how many yards would player X have thrown for against an average (or replacement level) defence rather than the one he did play against?". Or "What would a replacement level players stats have been in this game? (Real world yards MINUS YAR or DYAR, not plus).

Please, please either explain what I have misunderstood or go back to the true yards concept and think about what it is meant to represent again.

67
by Kulko (not verified) :: Tue, 07/08/2008 - 4:09am

Re: 64
If you were Willie Parker at a time when Jerome Bettis was the RB on all redzone plays, you could conceivably be a starter who was never on the field for points, and that would skew the stats. "

That misses the point of dpar completely. In a scenario like this Parker gets 95% of the dpar, because that is exacly what FO metrics are about. You are right, that traditional yards are better than traditional points, but that why Fo introduced a nmetric which spread the points of a team over all offensive skill players.

In generell I believe Aaron is trying to solve a problem that is not solvable.
Last year the fox reader wrote "What are your stupid dpar anyway, can't you see that Kitna passed for trillions of yards he is a hero"
So this year they will write, "What are you stupid true yards anyway, the true yards are the one on his box score. Alex Smith will always be our hero."

Also I must say, the more I think about it, the less I understand what these "yards" mean.
Bradys true yards come down at 6428. For this number to mean anything, there should be a sentence like "If Brady played sixteen times against an average defense then he would have thrown 6500 Yards."
Firstly I do not believe such a sentence to be "true" for any if, but I guess I just can't fathom how superhuman the pats where last year.
But secondly the way FO stats works, the real sentence is probably like: "If Brady plays 16 games against average defense and plays all snaps from either 1 and 10 from his own 47 yard line or 2nd and 7 from the oppenets 35 yard line (or whatever an average difficulty situations for a QB are), the he would have thrown for 6500 yards. If you consider that the real sentence is probably even more complicated, I do not see where this creates a stat that is more accessible than DPAR.

68
by KG (not verified) :: Tue, 07/08/2008 - 4:20am

I do like the DYAR addition, but I'd much rather be able to work with DPAR if we can only have one.

69
by ammek (not verified) :: Tue, 07/08/2008 - 4:44am

Another vote for offering both DYAR and DPAR stats, at least on the FO stats pages. I've just about gotten used to DPAR!

Not many commenters have mentioned it, but there are some excellent-sounding changes in your upgrade: red zone and replacement level, in particular, make a lot of sense.

Now that you are working from five years of data (2003-07), I wonder: has the value of Average in DVOA changed at all? If, to take an extreme scenario, passing success on first down increased by, say, 15 per cent in 2008, would DVOA adapt to that? Or does Average mean 'average for the era'?

70
by phill (not verified) :: Tue, 07/08/2008 - 6:11am

re myself in 66 - obviously I am reading impaired this morning :(. What I wrote about True Yards was complete rubbish - it's not player's actual yards plus DYAR, it's 'replacement level yards' plus DYAR - something a bit like how many yards they would have passed for against that defence if each play was 1st and 20 from their own 20, or some such metric.

71
by Admore (not verified) :: Tue, 07/08/2008 - 6:47am

Well, one nice thing about the new stat is that you not only can make Pirate Sounds (always good) but Philadelphia Story reference too.

KH: "My, that 2005 Peyton was yar!"
CG: "ITruly he was, Red. Yar, indeed."

But that still doesn't solve a baseline problem I have with football analysis. Unlike baseball stats, hardly any football stats really resonate. Baseball Prospectus has a couple of revolutionary stats, but one of them, Eqa, looks like something you grew up with, batting average. It takes offensive performance in totality and renders into a batting average like number.

What's the football equivalent of batting average, or era? It's hard to say, since the roles of players, and their relative success in those roles, is so different.

It's possible to get to important and revealing stats, and I believe FO has made great progress with this. DYAR (yes, why not YARD, or YARDA?) I think is an attempt to get to a more familiar context than points. I just don't think either is so familiar that it really matters much which you pick for translation.

Personally, I'd rather see wins (warp), or fractions thereof (which, hey, probably would look like batting averages!). To say a QB is worth 5000 D'YARS! versus an average of 3000 D'YARS! isn't terribly meaningful to me on first glance. Whereas if you said "This QB is worth 3 wins above replacement, given a average team around him, and an average opponent." I'd think, hey, that's huge! That guy is great.

72
by steelberger (not verified) :: Tue, 07/08/2008 - 7:19am

19....I agree. Points win games, not yards.

And personally, as a Steelers fan, I find any system that ranks Sage Rosenfels ahead of Ben Roethlisberger (new DVOA) sort of ridiculous. Maybe I just read it wrong.

73
by Kulko (not verified) :: Tue, 07/08/2008 - 7:56am

Bug Report:
The new table for Quarterback Rushing is sorted by true yards instead of DYAR like every other table.

74
by Harris (not verified) :: Tue, 07/08/2008 - 8:12am

I'll put in another vote for DPAR. Yards, as most people here agree, are a poor measure of success. We've often ridiculed the NFL for designating the best offenses and defenses by yards instead of points and it seems like DYAR is a step in that direction.

I'm sure I'll get used to the new formula, but I'm not sure that I'm going to like it.

75
by ammek (not verified) :: Tue, 07/08/2008 - 9:34am

Hey, that 1995 NFC Central was positively stacked at the QB position! Mitchell, Kramer, Dilfer...

76
by Joe T. (not verified) :: Tue, 07/08/2008 - 9:37am

But isn't DYAR a measure of individual success, not team success? So wouldn't yards be a more accurate measure of an individual player's contributions than points scored? I'm just thinking of a running back or wide receiver who routinely moves the chains for a team, but once his team gets in the red zone the fullbacks and tight ends get a disproportionate number of TDs (ie Warrick Dunn & Mike Alstott).

So a red zone fullback would have his DPAR skewed in one direction, and his DYAR skewed in the other?

77
by NewsToTom (not verified) :: Tue, 07/08/2008 - 9:48am

Re #54
I'm waiting for my copy of PFP to give a fuller analysis. In the meantime, I concur generally with qsi in #63 and with #19 in wondering what non-FO partisans will think of DYAR. I'll see what results experiments on my co-workers, who generally have not cottoned on well to DPAR, render.

And to answer your second question, judging by my current security word, no.

78
by mawbrew (not verified) :: Tue, 07/08/2008 - 9:48am

I kinda like the idea of DYAR on an individual game basis (i.e. Quick Reads). In evaluating/measuring individual game performance it does seem more intuitive than DPAR. As a cumulative measurement I think I still conceptually prefer DPAR though there's probably not a big difference (in the table above no QBs ranking changed more than two spots).

79
by A. Diggity (not verified) :: Tue, 07/08/2008 - 10:00am

I think that quote cuts to the quick of the matter at hand better than anything that's been said. The inherent problem isn't really replacing points for yards in a qualitative metaphor; rather, it's the nature of a qualitative metaphor itself. The point of metaphors like this is to translate a long set of performance metrics into one number that states whether something (in this case, a football player) is good or bad in a way that a layperson could relate to his own experience. (e.g., "Well, 100 horses can pull more than 75, so this engine is stronger than that one.") The key word, here, is "layperson", which will come into play more and more as Football Outsiders gets more popular and becomes more revenue-driven. Now, that's not to say that the FO guys should boil their stats down to "on a scale of 'teh suck' to 'ZOMG!', Vince Young is 'meh' with a slight chance of becoming 'feh' in the future", but I appreciate that they have received some criticism of casual users of their stats (like the readers of Quick Reads), and have at least made an attempt to meet them halfway.

80
by A. Diggity (not verified) :: Tue, 07/08/2008 - 10:02am

Whoops, #79 was supposed to open with:

Translating player performance into yardage rather than points allows us to introduce one more statistic, which we hope will make Quick Reads much more accessible to the general population of football fans.

Sorry 'bout that.

81
by jsa (not verified) :: Tue, 07/08/2008 - 10:23am

I hope DPAR is dead like Favre is retired.

82
by The McNabb Bowl Game Anomaly (aka SJM) (not verified) :: Tue, 07/08/2008 - 10:29am

Re: 76

If I understand correctly, DPAR and DYAR are based on the same performance factors. A red zone fullback would not have a good DPAR and a bad DYAR, they would both be either good or bad because they are saying the same thing, just expressing it in different terms. DPAR was never based on distributing the actual points scored, it was always based on drive success points (on a per-play basis), and DYAR is based on that, too.

DYAR is not qualitatively different from DPAR. The debate is whether we want value above replacement to be scored on a points scale (DPAR) or a yards scale (DYAR). I personally like points, and I don't see how yards are more meaningful to anyone. In fact, It seems confusing to me that a player could earn more "True Yards" than he actually got. What does that mean, really, in terms of value to his team? We all know Brady got a lot of yards. But is his value determined purely by getting yards? Only in the sense that yards lead to points. So why not skip that step and go straight to points?

(However, I do like that "True Yards" and DYAR will be able to put to rest the myth that throwing 4,000 yards is a good thing. We'll be able to look at a 4,000 yard guy and say his true yardage was far less than that, while a guy who had 3,800 normal yards actually has a higher number of "True Yards." So the new stats are useful for comparison of players at the same position to each other, but not as useful as DPAR for measuring a player's contribution to his team's wins.)

Again, I end with a plea to keep DPAR around for comparison, to see if people like it more.

83
by The McNabb Bowl Game Anomaly (aka SJM) (not verified) :: Tue, 07/08/2008 - 10:33am

P.S. I realize that the charts make it seem like DPAR and DYAR are different, but that's because the DPAR numbers are from before the baseline changes and the DYAR numbers are from after those changes. If you recalculated DPAR with the new baseline numbers, it would rank the players in the same order and with the same proportions as DYAR.

84
by gat (not verified) :: Tue, 07/08/2008 - 10:49am

Are kickers going to be ranked using DYAR instead of DPAR? That seems... stupid.

85
by throughthelookingglass (not verified) :: Tue, 07/08/2008 - 11:10am

I think the improvements to the formula, especially the new replacement levels, reflect excellent work. Good job, guys!

However, when I saw the headline "DPAR is Dead", I hoped FO was moving to a wins-based metric. After all, the point of yards is to score, and the point of scoring is to win. It seems like this stat is moving backwards. It's not a bad thing on its own-more stats are better, as long as they are clearly labeled, explained, and understood. However, if forced to choose between DYAR and DWAR, I'd prefer to have an individual metric that corresponds more directly with the whole point of the freakin' game (corresponds in my mind, I realize the actual difference would be negligible).

86
by admin :: Tue, 07/08/2008 - 11:20am

82 and 83 are correct.

87
by Reinhard (not verified) :: Tue, 07/08/2008 - 11:21am

I love DYAR!!!

Kickers + yards: those are yards that you don't have to get to get into good % field goal range, since your kicker 'already' got them.

88
by The McNabb Bowl Game Anomaly (aka SJM) (not verified) :: Tue, 07/08/2008 - 11:30am

Re: 82 (myself)

I made a mistake in saying that "True Yards" are good for comparing players at the same position. They are not. "True Yards" are only useful in determining whether a player was better, even with, or worse than his actual yardage number (which is of dubious utility, IMHO).

89
by NewsToTom (not verified) :: Tue, 07/08/2008 - 11:43am

Re #88
I generally share your skepticism w/r/t the probable value of True Yards, but one thing I noticed and sent Aaron an email about is, looking at RB True Yards, RBs like Fred Taylor and Purple Jesus have significantly lower True Yards, while Joseph Addai has more True Yards. Coincidentally or not, MIN and JAX rank 1st and 2nd in 10+ yards and IND ranks toward the bottom. The OL rankings discount 10+ yards, which makes strong intuitive sense. But, from those examples, it seems like True Yards do, too, and I'm not sure THAT makes sense. It's particularly confusing w/r/t somebody like Purple Jesus-both DYAR and DVOA say he's very good, and then there's True Yards saying "whoa, wait a minute, he's not nearly as good as he looks."

90
by Devin McCullen (not verified) :: Tue, 07/08/2008 - 11:57am

re 57 - the list of 8 worst QB 5-game stretches in terms of completion percentage (during a new moon while geese are migrating). Is it just random that 5 out of 8 on the list end on week 15 or 16 - is this a case of QBs (or teams generally) being shell-shocked by the end of a crappy season / giving up on a crappy season, or just one of those random things.

I suspect it has something to do with lousy weather.

91
by Joey Jo-Jo Junior Shabbadu (not verified) :: Tue, 07/08/2008 - 12:04pm

#87: I'm not sure that would be correct; I don't think 1998 Gary Anderson (regular season version) could have kicked a 60 yard FG even though he undoubtedly would have had a really high DYARrrrr.

92
by gerry (not verified) :: Tue, 07/08/2008 - 12:22pm

This was my first thought: "with DPAR I could subtract it from a team’s point total or swap players and use the old pythagorean formula to run “what ifs” with different players; how do you do that with DYAR?"

And this was my second: "How does anyone know if any of the stats make sense? Do they double count things? What aspects of performance are used, and how are they weighted? Etc?"

And this was my third: Who wants to create an open-source version of FO? We could create customizable stats and unlift the veil of the black box. Imagine understanding why the 2007 projections had Dallas with 6.4 wins and Philly with 11? More accurate and sensible FF projections. Who's with me?

This comment will be purged in 3-2-1...

93
by Wes (not verified) :: Tue, 07/08/2008 - 12:32pm

As a loyal outsider promoter since nearly day 1, I've got to throw my hat into the pro-DPAR camp. I see this as epic failure and huge step backwards. I've always been able to say to non-believers that yards don't win, points do. So listen to these guys (FO) who are talking about how many points a player is worth... Now what do I do? Yards don't win, so look at these yards! They're different numbers! I've not been this disappointed with FO ever before.

What type of bike did you use to jump the shark? Scooter? Chopper? Vespa? Schwinn?

Fail.

94
by Jimmy (not verified) :: Tue, 07/08/2008 - 12:32pm

If the point of the revamped statistics is to make complex statistical data more easily understandable for a lay person (ie the statistically ignorant) then I think DYAR may fail (for all its piratey goodness).

You should have gone for something everyone can relate to and measured production in PEPSIs or DONUTs. Anyway I am sure you can think of something.

I may be missing something obvious here but where you went for 'Defense-adjusted Yards Above Replacement' or DYAR, couldn't you have used 'Yards Above Replacement adjusted for Defense' which would have been YARDs. Not quite as piratey, but if you try you can still make it quite menacing.

95
by Bad Doctor (not verified) :: Tue, 07/08/2008 - 12:35pm

I'm not liking the new stats, but maybe it's because I'm missing something. True Yards is really throwing me.

So Tony Romo threw for 302 yards on 44 passes against the Lions, a craptacular passing defense. Given the context of his pass attempts, we'd expect a replacement level QB to throw for 195 yards in such a situation; as such, Romo was 107 yards above replacement. But when we add back in what the replacement level QB would do, we say Romo contributed the equivalent of his 107 yards above replacement plus 297 yards, which came from something to do with equivalent pass attempts, so his true value was 404 yards. I don't get it. How can 300 actual yards against a crappy pass defense really mean you were as valuable as if you threw for 400 yards against an average pass defense? The latter is a much more impressive performance.

To cite phill in #66, what are we proving here? What is the end to this sentence?: "By passing for 302 yards against the Lions, Tony Romo was as valuable, by our calcuations, as if he had thrown for 404 yards . . ."

96
by The McNabb Bowl Game Anomaly (aka SJM) (not verified) :: Tue, 07/08/2008 - 12:48pm

Re: 95

There's something wrong with your numbers. Clearly, if Romo passed for 302 yards against a bad defense, then his "True Yards" would be lower than 302 because of opponent adjustments. You must have calculated wrong.

97
by admin :: Tue, 07/08/2008 - 12:49pm

Sigh. Unfortunately, it looks like Tom Gower is correct about "True Yards." I'm not sure why I didn't notice this before, but it is not working right. Actually, that's not true -- it is working right, but it isn't saying what I thought it was saying, because the standard deviation is smaller than the standard deviation for regular yards. I'll be looking into this and trying to improve it over the next couple days.

As for DYAR, I'll address this further when I do a "DYAR Mailbag" but it is the exact same stat as DPAR, just on a different scale. Basically, we've gone from Fahrenheit to Celsius. Freezing is still freezing.

To give an example, here's the top 10 QB from last year in terms of DPAR, but with the new baselines and formulas. It is the same as the ranking in DYAR.

12-T.Brady 188.3
18-P.Manning 125.3
4-B.Favre 98.5
9-T.Romo 88.9
9-D.Brees 89.3
9-C.Palmer 84.0
9-D.Garrard 74.0
6-J.Cutler 67.3
8-M.Hasselbeck 65.6
3-D.Anderson 55.9

98
by Chris (not verified) :: Tue, 07/08/2008 - 12:54pm

Dyar seems more like a fantasy football concept where you are seeking intrinsic value in terms of yards/stats. Dpar seemed like a better stat for true WIN the game value.

The problem I have with Dyar is that not all yards are created equal. It is so much easier to rack up passing yards on artificial turf in a dome, than it is to pick up yards in frozen temps and swirling winds of Giants stadium or Buffalo NY in December just like it is easier to hit a Home Run in Colorodo as opposed to a pitchers ballpark. I hate to bring up baseball, but even in the same stadium, a lefty can have a huge advantage over a righty in Yankee Stadium.

If anything, the very familiar Colts are an example of why Dyar might not be as good as Dpar. I believe it was Gunter Cunningham's/Herm Edwards Chiefs/Colts playoff game in 06' game in Indy. There were reports that the Chiefs were going to play ultra conservative until the Colts got into the field goal range and then start to play "real" defense and hope to limit them to just 3 points on lots of execution instead of the possible quick 7 point score. This strategy would give up yards, but it would make them " earn them" and slow down the game and keep it close etc. They knew the probability of a mistake ( fumble/tipped ball) was about as likely as a punt so they just tried to limit the big play down the field and hope for false starts, tipped balls etc.

99
by Temo (not verified) :: Tue, 07/08/2008 - 12:58pm

95. To my understanding, true yards is what a replacement level player would have done PLUS the added production of the player. So actually a Replacement level player would have thrown for 297 yards. (404-107=297).

Which makes sense-- that's 6.75 yards per pass attempt against a crappy Detoit Pass Defense- much more than the 4.4 yards/pass number you get using your numbers. That's why DYAR is the better comparison stat b/w QBs than TYAR.

Or so I think.

100
by Chris (not verified) :: Tue, 07/08/2008 - 12:59pm

97- If it is a question of easier divisibility, you can convert POINTS to the tenth, hundreds, or Thousands column. DPAR can read Carson Palmer was worth 26.789 points more than a replacement level quarterback.

101
by Temo (not verified) :: Tue, 07/08/2008 - 1:06pm

97. Brees has a slightly higher number but is ranked lower.

102
by FarenheitCelciusKelvin (not verified) :: Tue, 07/08/2008 - 1:09pm

Any chance of conversions between yards, points, and wins? These seem like very, very simple regressions which would let readers use a simple multiplication to convert between DYAR, DPAR, and a theoretical DWAR.

103
by Chris (not verified) :: Tue, 07/08/2008 - 1:14pm

102- I don't see an even exchange between Yards, Points, and wins. They would be correlated but it isn't like 100 yards is worth 6 points in a 1 for 1 trade etc.

If my team scored 42 points in a given game, you would predict that we would beat the unknown team B. Now if I told you we played in St. Louis and that team B scored 45 points then it wouldn't matter that we scored 42 points as far as wins go.

If I told you the Giants were going to Score 23 points on the Pats in the super bowl...

104
by Temo (not verified) :: Tue, 07/08/2008 - 1:25pm

95. I'll take a shot at this, anyone feel free to correct me:

I'm thinking, and correct me if I'm wrong, that since DYAR like DPAR comes from "success points", there is more importance given to a 2 yard pass on 4th-and-2 than a 9 yard pass on 3rd-and-10-- perhaps even if the 9 yards came against a crappy defense. Therefore, the QB that has gotten 9 yards against a staunch defense is probably getting less true yards than the QB who got 2 yards on a crappy defense.

So even though Romo threw for 302 yards against the lions- his value to the team was more in line with a QB who threw for 404 yards against the lions.

105
by Temo (not verified) :: Tue, 07/08/2008 - 1:27pm

104. Slight correction- since it's defense adjusted, that would be 404 yards against ANY team, not the Lions. I think.

106
by Patrick (not verified) :: Tue, 07/08/2008 - 1:39pm

@74: "Yards, as most people here agree, are a poor measure of success. We’ve often ridiculed the NFL for designating the best offenses and defenses by yards instead of points and it seems like DYAR is a step in that direction."

what? i think most people here agree the opposite: total yards are a better measure of offense than points. points happen thanks to field position, and often happen without the offense seeing the ball at all. the 2006 bears were no where near the 2nd-best offense in the league. neither yards nor points are actually good though, which is the point of FO.

People talk about VORP for baseball, but baseball fans are used to crappy numbers like runs and RBIs. VORP is really just an advancement on those numbers.

there is no traditional football player stat that DPAR compared to.

My question(s) about DYAR:

with DPAR (or VORP), you could take a replacement offense (maybe 150 points? '06 raiders?), then add up all the DPAR from individual players and it would give you a 'predicted points scored'.

with DYAR, can I take a replacement team yardage (4000-yard season maybe?) and add all the individual DYAR totals and get a predicted yardage? will their be a table for how many points a team with a given teamwide True Yardage would be predicted to score?

Are we now doublecounting? In the olden days, the DPAR for a completed pass were divided between the QB and receiver, right? Now, it's fairly clear there is overlap for True Yards, Brady had more True Yards alone in 2007 than his team had in real yards.

But are we dividing DYAR between QB and WR? On the one hand, this would tend to make QB/WR #s lower than you would expect (because traditional numbers double count for passing yards),

but if you don't, then an RB DYARd is worth as much as 2 QB/WR DYARds, which would be awful.

107
by mrh (not verified) :: Tue, 07/08/2008 - 1:40pm

We’ve also put the 2008 game charting data in the store; you can purchase that for $60.

If we could get the 2008 game charting data now, it would be worth a lot more than $60.

108
by Tally (not verified) :: Tue, 07/08/2008 - 1:41pm

I think what's throwing people is that they think just because the new stat includes the term yards that suddenly FO is putting more value on yards.
That's simply not true, any moreso than they placed emphasis on points scored when using DPAR. There has been no sudden paradigm shift from points to yards. DYAR and DPAR are both derived from DVOA, which has been updated for greater predictive value.

109
by Bobman (not verified) :: Tue, 07/08/2008 - 1:47pm

I'll reserve judgement for a while, but my gut says "points is right." Crabbie and others make a bunch of sense to me.

Shifting into 2nd gear, one of my favorite features on FO is the drive success rate table, and a fair amount of that is yards-based, no? That's more of an "event-based" metric, including points, yards, fumbles, TOs, etc.

As a consumer and a businessman, I wonder how this is related to the USNWR annual college/grad school rankings: they change every year to keep it fresh. Nobody (except my wife) wants to see Wharton as #1 every year. Will an updated DPAR return in 5 years and then a refined DYAR(rrrr) after that?

I appreciate that the FOers have continuously refined their system--sometimes even mid-season if it seemed necesasary. In the past it has steadily improved, so I will assume this is an improvement I don't quite "get" yet. And if it's generally an "expand the business" move, I can understand that as well. I may not love it, but I'm not going anywhere else.

(confession: I kind of like the Aikman Ratings--sort of a halfway step between FO stats and traditional NFL stats--talk about abstract, like a QB rating, there's no tangible measureable they correlate to, but since they are just one O and One D number, you can take in a broad league-wide snapshot in one glance. They-re crude--the QB/RB/OL performance is not broken out--it's all based on 11 guys at a time.)

110
by Bobman (not verified) :: Tue, 07/08/2008 - 1:48pm

107 SWEET! You're right. I'll charitably bid $100 for that data.

111
by Temo (not verified) :: Tue, 07/08/2008 - 1:53pm

109. I don't think there's been any monumental shifts in how they calculate their ratings though. It's mostly a result of having more data to analyze as the seasons pass. As opposed to the school ratings you describe, which continually analyze the same data (well, in a manner of speaking... obviously schools do change from year to year) but arrive at a different conclusion.

112
by Bad Doctor (not verified) :: Tue, 07/08/2008 - 1:53pm

99:Right, Temo, but there's my question ... if a replacement level QB would have thrown for 297 yards, and Tony Romo threw for 302, how in hell does he have a 107 DYAR? And if a replacement level QB would have thrown for 297 yards, and Tony Romo threw for only 5 more yards than that, how does he have the 6th best quarterback performance of the week?

And if there are two different levels of replacement value, how does FO know they're using the right one for DYAR and the right one for Total Yards? Couldn't they be the same? Couldn't they be reversed?

113
by gerry (not verified) :: Tue, 07/08/2008 - 2:00pm

Here's a trick question. How are Rex Grossman's actual yards and True Yards identical (1221) when he also posts negative DYAR (-121)?

114
by Temo (not verified) :: Tue, 07/08/2008 - 2:08pm

112. Well, that means that his DVOA for the game was fairly high. Just like Jon Kitna had more Actual Yards than True Yards because he had a negative DVOA... so does Romo have more True Yards than Actual Yards because his DVOA was positive.

Looking at the boxscore (linked) it looks like Romo passed for 20(!) first downs on the day, meaning that almost half of his attempts resulted in first downs for the Cowboys.

115
by Chris (not verified) :: Tue, 07/08/2008 - 2:12pm

Maybe this Dyar, Dpar, True yards, fake yards, actual yards stuff is all just a scam to get the general public to quit this site, so that the TRUE, die hard football fans can build it back from the ashes and "keep it real".

In the NWFOO ( New World Football Outsiders Order) only 5 star commenters exist and all fans are Patriots fans.

116
by Jimmy (not verified) :: Tue, 07/08/2008 - 2:15pm

#109, Bobman

Do they still publish the Aikman ratings? I haven't seen them in a while, but I haven't been looking all that hard.

I think it is probably true that any purely statistical analysis of football is likely to be more accurate when looking at units as a whole (ie offense/ defense/ special teams). Any play that happens in a game will have an impact on any others that follow, and any player on the field on a given play will have a unique impact on said play. FO does a great job in providing great new information, and its efforts to incorporate analytical game data are highly commendable.

There are always going to be some people who will never be happy with any rankings short of their own created by watching NFL films game film of every game played and creating their own scouting system. You could also substitute your own scouting getting hold of the pro-scouting reports of a trusted NFL front office (best of luck to anybody who tries this option - but if you do manage it, can I get a pdf?).

117
by Red Dwarf (not verified) :: Tue, 07/08/2008 - 2:46pm

I think math geeks (and I mean that affectionately, I love this site) sometimes forget how much words matter. To judge by how much confusion still exists despite attempts to explain that's it's just the scale changing (great fahrenheit to celsius example, for all the good it seemed to do), I suggest avoiding the words 'yard' or 'points' in any overall rating. Heck, apparently a lot of people still think the FB poaching a 1 yard TD after someone else did all the work to get that close to the end zone means the FB gets more DPAR (which is either not true, or I've misunderstood everything here for years).

My suggestion: call the final ranking number Win Shares or Percentage Above or Below Replacement (PABR) or something. If the goal is to make it more accessible and less confusing, then avoid real world football related terms like yards or points as those have specific meanings in most fans minds that they find difficult to ignore when looking at your ranking, which obviously leads to mass confusion even among your core fans who follow this site in the offseason.

118
by FarenheitCelciusKelvin (not verified) :: Tue, 07/08/2008 - 2:58pm

Chris(103):

All of these stats are based on averages and expectations. For example, the commonly used Pythagorean formulae take points scored and points allowed and calculate an expected number of wins, with constants adjusted per-sport. That doesn't mean the team WILL win that many games, it just means that's the number they're MOST LIKELY to win.

There's no reason a yardpointwin conversion shouldn't be possible.

119
by Biff Tannen (not verified) :: Tue, 07/08/2008 - 3:13pm

re:107

I'm not impressed.

120
by krugerindustrialsmoothing (not verified) :: Tue, 07/08/2008 - 3:16pm

if that's not a headline that threatens to rip your stomach out, I don't know what is. Can't wait to see the Pirate stat in action.

121
by clonmullin (not verified) :: Tue, 07/08/2008 - 3:42pm

DPAR, DYAR, True Yards ... I don't care. I have gotten word from Amazon that the book is on the way and the first KUBIAK predictions are out. However quiet the off season is for ye guys just imagine what it's like here in Ireland. Football season is finally here :-)

122
by NewsToTom (not verified) :: Tue, 07/08/2008 - 3:45pm

Oh, just because DPAR/DYAR has dominated the discussion...

1. Thank you for more sophisticated opponent adjustments.
2. Thank you for finding the new replacement values.
3. Thank you in advance for the forthcoming series on the best single seasons in FO history, including putting the pre-2000 data on the site.

All of these are genuinely Good Things, and we're thankful for your hard work in doing them, even if we don't necessarily act like it.

With that, we now return to Scale Fight 2008: Points Versus Yards!

123
by Devin McCullen (not verified) :: Tue, 07/08/2008 - 3:47pm

I think the one thing we can all agree on is that the next time Aaron decides to rename a major stat, he needs to run a focus group first.

(And DPAR is still coming up as a security word.)

124
by sam! (not verified) :: Tue, 07/08/2008 - 3:59pm

Aaron,

My apologies. After re-reading the article (plus your post), DYAR is just fine with me.

125
by Quentin (not verified) :: Tue, 07/08/2008 - 4:02pm

Of course, small changes in our stats should not be treated as gospel truth.

I don't think that's gonna be a problem.

126
by Dan (not verified) :: Tue, 07/08/2008 - 4:17pm

The switch from DPAR to DYAR is not a big deal - based on Aaron's lists it looks like the conversion is pretty close to DYAR=14.5xDPAR. So, for those (like #102), who wanted a conversion calculator: just divide yards by 14.5 to get points. My initial reaction the DPAR/DYAR controversy is that I liked the old way and there was no need to change it, but it's not a big deal and I might come to prefer the new scale once I get used to it.

What I don't understand is why FO plans to give "True Yards" so much prominence. I may revise my views once I see it in action, but right now it just doesn't seem like a number that's very meaningful, interesting, or useful to look at. Why should I care what a player's True Yards were? Comparing True Yards to actual yards in order to see whether someone is better or worse than their traditional stats is complicated and it only gives a very rough sense of how good someone has played. Unfortunately, the name "True Yards" makes it sound like this is a really important stat (especially to people who aren't familiar with the acronym stats), which could confuse people and make it harder to get people to buy into DVOA & D[P/Y]AR.

Perhaps True Yards per attempt would be a more meaningful stat? That would basically be yards per attempt, with bonuses or penalties based on opponent adjustments, types of usage, and how good a player has done at getting progress towards first downs, getting touchdowns, and avoiding turnovers. Basically, it's DVOA translated to a more intuitive scale for each position. And the folks at Pro Football Reference have already found this sort of yards per attempt based scale pretty convenient to work with - they love to use adjusted yards per attempt. True Yards per attempt would sort of be like a souped up version of that, which uses play-by-play data instead of just box score data.

127
by D (not verified) :: Tue, 07/08/2008 - 4:42pm

Aaron would you mind posting the conversion from DYAR to DPAR? When I took a quick look at the numbers you posted in #97 I thought it was about 14.5 x DPAR = DYAR but I ran the numbers and that's not quite right.

128
by D (not verified) :: Tue, 07/08/2008 - 4:48pm

To expand on my previous point the multiplier increases as DYAR increases. I'm sure if my math chops were a little sharper I could calculate it myself (at least for QBs) but its summer and I haven't taken a math class in about 5 semesters.

129
by Gerry (not verified) :: Tue, 07/08/2008 - 4:54pm

"This comment will be purged in 3-2-1…"

I have two things to say.

1) gerry is not Gerry. Maybe I need to go back to Dales...
2) FOMB don't play that way. Now, if you had said that Brady is obviously better than Manning, then just maybe you would have something to worry about.

130
by Chris (not verified) :: Tue, 07/08/2008 - 5:06pm

118- A Yard to Point scale would be much more possible ( Only flawed by an Alstott/Dunn, Bettis/Parker combo).

Are you talking about a PF/PA in a game and predicting the outcome, or a PF/PA in a season and predicting season win totals?

Since pretty much the entire conversation was player specific, I would assume you were talking about individual players. I don't think you can say that say 75 rushing yards = 1 TD = .333 wins etc or the comp for Quarterbacks.

A lot of individual stats are dependant on the venue as a huge variable which doesn't effect either say the QB or the defense. Playing in a snowy windy game in Buffalo with 5 degree temps will hurt both QB's stats and help both Secondaries.

131
by Yaguar (not verified) :: Tue, 07/08/2008 - 5:14pm

92: I'd love to see more of the inner workings of FO calculations revealed. (After all, I did reverse-engineer the Lewin projection system so I could use it on the 2007 draft class before the book came out. Maybe this is a sign that I need to get out more.)

Unfortunately, I don't think it's really in their business interests.

132
by Tally (not verified) :: Tue, 07/08/2008 - 5:25pm

Re 127, 128:

It does look linear (r^2=0.9998) but you're right in that the multiplier isn't constant.

133
by Reinhard (not verified) :: Tue, 07/08/2008 - 5:26pm

I think most of the opposition is sentimental, and I know it isn't logical.
DPAR NEVER CALCULATED POINTS. It takes basic yardage plays and estimates how many points that was worth. So we aren't going from points to yards. But from yards to True Yards. And some yards are obviously more valuable than others. 3rd and 1 getting 1 yard is a valuable yard. 4th and inches that's a very valuable inch! That's the whole concept that FO and it's stats are built around. I am very surprised by the reaction I am seeing. It will go away soon.

134
by Temo (not verified) :: Tue, 07/08/2008 - 5:28pm

130. I would certainly like to "reverse-engineer" (if you want to put it that way) some of FO's work. If only because I too should probably get out more.

135
by Reinhard (not verified) :: Tue, 07/08/2008 - 5:29pm

96: Maybe what happened is that he made very valuable plays during the game, such as converting long 3rd downs, didn't turn the ball over, threw td's and such. So although those plays would be even MORE valuable against a good pass D, he could still have produced more than you would guess just by looking at his raw yardage totals.

136
by NewsToTom (not verified) :: Tue, 07/08/2008 - 6:00pm

Re #127
Keep in mind that the changeover also included some calculation adjustments, such as strength of schedule, so the numbers won't perfectly correlate.

137
by Dan (not verified) :: Tue, 07/08/2008 - 6:16pm

131: Looking for the linear relationship in a statistical package, I get DYAR = -42.49 + (15.03)DPAR [which is roughly DYAR=15(DPAR-3)]. The intercept is significantly different from zero (p=.0004), and the 95% confidence intervals for the numbers are (-56.82, -28.15) for the intercept and (14.88, 15.17) for the slope. This equation has an r^2 of 0.99981, and there's no visible pattern to the residuals. I don't see why there should be an intercept, so my guess is that Aaron made some last-minute tweaks to the formula between when he calculated the DPAR numbers in #97 and when he calculated the DYAR chart in the post.

138
by Goo (not verified) :: Tue, 07/08/2008 - 6:32pm

132. And why do you think those certain yards are more valuable than others?

Because gaining those yards increases the probability of scoring ...drumroll please... points.

139
by Goo (not verified) :: Tue, 07/08/2008 - 6:33pm

Whoops, that was in response to 133.

140
by mikeabbott (not verified) :: Tue, 07/08/2008 - 6:46pm

I prefer points but realize that, while people who prefer DPAR as a measurement to DYAR are sophisticated enough to use either, the FO staff may suspect that DYAR will be easier to explain to Fox Sport yahoos (in the Swiftean sense).

Has anyone ever used FO numbers for fantasy football?

141
by Stuart Fraser :: Tue, 07/08/2008 - 7:46pm

People wondering about the acronym:

The reason we're using DYAR and not YARD is because occasionally FO people appear on talk radio shows, podcasts etc., and having an FO stat which sounds identical to a widely used conventional stat would not be helpful in this medium.

We did consider DRAY (Defense and Replacement Adjusted Yardage - another stat which Tampa Bay ought to be good at) but DYAR had the votes. Me hearties.

142
by Conor (not verified) :: Tue, 07/08/2008 - 7:55pm

Going back to something arlier in the thread, why would Brady be a red light to meet his 37 TD projection? Wouldn't that just mean it is a bad projection? Or maybe closer to his 75th percentile projection? The idea that you can be a bad risk to meet a projection kind of seems off to me. If that is the case then why is it the projection?

143
by Stuart Fraser :: Tue, 07/08/2008 - 8:09pm

Re: Brady, and noting that this is not in any way an "official" answer.

Roughly the projection system goes:

"Hey, Brady had a really great season in 2007. All the variables relating to the New England offense say it should be fine, he's not old enough to be declining, so let's project more of the same".

Then other parts of the projection system say:

"Hang on. Brady's 2006 wasn't that good. In fact, Brady has never thrown anywhere near that many touchdowns and he's only gone over 4,000 yards once before. And after that one he dropped back down to around 3,500. Who wrote this projection anyway?" and gives it a "RED" reliability rating.

Most players are not particularly consistent and therefore a Green rating is something of a rarity.

144
by shamne (not verified) :: Tue, 07/08/2008 - 8:46pm

Just got my hands on PFP '08 and I can barely contain myself.

145
by Alex (not verified) :: Tue, 07/08/2008 - 9:23pm

How about we settle things this way:
Aaron tells us DPAR junkies how to convert DYAR to DPAR, and we give this new DYAR thing a shot.

146
by Yaguar (not verified) :: Tue, 07/08/2008 - 10:07pm

143: The projection system applies its own risk ratings? I always thought it was done manually. Chad Johnson has a Red rating, for example, because of the trade rumors and behavior issues and stuff. The projection system can't possibly know about that.

147
by jstreet (not verified) :: Tue, 07/08/2008 - 10:14pm

I'll be happy either way, but if we still can see player DVOA, the DYAR change makes a lot of sense to me.

148
by the silent speaker (not verified) :: Tue, 07/08/2008 - 10:16pm

Since the problem here is some yards being more important than others, does it make any sense to measure yards in some fusion of fraction of the distance to the end zone, fraction of distance to the first-down line and fraction of 100 yards? Something simple, like cube the first, square the second, multiply the cube and the square and the last one, then take the cube root. A two-yard touchdown is cbrt(1*1*.02)=.2714=27.1 Fused Yards by this literally made-up-on-the-spot metric, whereas a nine-yard non-first-down from midfield is cbrt(.18^3*.9^2*.09)=.0752=7.5 Fused Yards. I'm sure the good people at FO can improve on the off-the-cuff version, maybe adjusting for down situation (twiddle the middle term somehow?) but I think the concept is fundamentally okay.

Also, re #94: The reason we don't measure production in PEPSIs and DONUTs is that the Player-Equivalent Point-Scoring Index is rubbish. Amd don't even get me started on Defense-Optimized Non-Ugly Throws.

149
by Goo (not verified) :: Tue, 07/08/2008 - 10:40pm

Former QB-turned-stathead Virgil Carter's research said the value of field position in points (starting a drive, 1st & 10) V = 5.91 - 0.077 Y, where Y = the # of yards from the opponent's goal line. I'd assume that someone has expanded this to include other downs and distances, but I can't find anything on Google right now...

150
by Pat (not verified) :: Tue, 07/08/2008 - 11:00pm

#145: It's roughly ~14.5 DYAR/DPAR, plus or minus a few percent. (Just divide the two lists Aaron gave).

Or you could just take 100 yards divided by 7 points, and get roughly the same answer.

151
by Tally (not verified) :: Tue, 07/08/2008 - 11:46pm

Re 149:

Yes, someone has expanded on the values of yardage on various downs and distances. You're on their website right now.

DVOA values yards relative to first downs and touchdowns (e.g., 8 yards on 2nd and 5 is valuable; 8 yards on 3rd and 15, not so much). Position relative to the goalline is considered, as well.

Guys, DYAR doesn't place any more emphasis on yards or less emphasis on points than DPAR. They're both based on DVOA.

152
by Goo (not verified) :: Tue, 07/08/2008 - 11:53pm

I should have said, "someone expanded on it and made the formula open-source". Obviously DVOA is the ultimate expansion of that work. DVOA is also a black box.

153
by Alex (not verified) :: Wed, 07/09/2008 - 12:18am

Or you could just take 100 yards divided by 7 points, and get roughly the same answer.

Wow. That seems so obvious now, but I don't know if I'd have ever realized it. Thanks, that helps.

154
by asg (not verified) :: Wed, 07/09/2008 - 12:22am

So, speaking of Lewin, he's not in the new book. Did he not do a projection article this year, or will it appear in another forum?

155
by cd6! (not verified) :: Wed, 07/09/2008 - 1:20am

Maybe it's the engineer in me, but I'm really unhappy with the name of the stat "True Yards."

A true yard, like an "actual yard" or a "yard" is 36 inches, period. An FO "True Yard" is a calculated stat.

Moreover, this name may lead to conversations like this:
"LT had a DYAR of 200 Sunday"
"Well he had 195 true yards"
"Ahhh but he had 250 True Yards"
say what?

As Stuart pointed out earlier, this kind of thing is going to cause headaches when casual listeners tune in to radios & podcasts.

Beyond that, I'll pick up a copy of PFP in the store before I weigh in on the DYAR vs DPAR debate. I'm looking forward to reading it.

156
by Bobman (not verified) :: Wed, 07/09/2008 - 2:02am

cd6!, think about "true yards" this way (10 years ago):

Guy 1: Barry Sanders racked up 130 yards today.

Guy 2: Yeah, but FO says it was only 90 True Yards because so many were backwards.

Guy 1: Sorry, my brain just exploded. What the hell did you say?

Maybe they should keep with the pirate theme and call that stat Yard Arms. (or should that be a passing stat?) Hell, there's a whole litany of untapped pirate vocabulary they can mine... "walking the plank" is any 3rd and 10 situation, "booty" is points, "keel haul..." no, I'll stop lest the guys at ESPN do an entire Sports Center broadcast in pirate lingo. Actually, it would be fine the first time, but by the 75th time, it would get a little old.

157
by Crabbie (not verified) :: Wed, 07/09/2008 - 2:29am

Yeah, it's important to remember that DPAR (my anti-spam word) is the exact same measurement as DYARD, only on a different scale. It's mostly a semantic difference.

But it's a huge one, as I and #155 and others are saying, and a huge step in a pretty rediculous direction. I mean, really, how anyone could've thought it was a good idea is pretty beyond me.

But that being said, I don't think it warrants any kind of OMG TEH SELLOUTS!!!1!!1! reaction. I'll still keep reading what Schatz and the stats have to say (since DVOA is still around and improved), and have always been much more interested in the team stats as the individual ones will almost certainly remain pretty flawed due to the structure of the game. There's just never going to be any kind of completely satisfactory way to definitively quantify individual contributions in football, given that there are 21 other players on the field and they're usually all doing something germaine to the outcome of the play. And then there's the incredibly important role of coaching, given the enormous range of ways those players can be arranged and instructed to interact on the field, especially given that play occurs as a series of discrete events (downs). I mean it really is just about as extraordinarily complicated a game as currently exists.

Anyway, just so long as unit DVOAs are around, I'll continue to have much love for Schatz, because they really are pretty far ahead of other publicly available models. I continue to suspect that the Vegas casinos have something similar or perhaps even better (and have likely had it for years), but they aren't sharing.

158
by Scott (not verified) :: Wed, 07/09/2008 - 6:55am

True Yards? Looks like the most useless statistic ever. Since when do we care about yards here anyway?

I'm waiting for some people with the will and means to really advance the statistical breakdowns of players in football. I want to go to a QB's splits and see what he does when he has 2 seconds to throw, 3 seconds to throw, 4 seconds, etc. What's his average protection time for a game or season? I want to see his splits for passes thrown into coverage versus passes thrown to wide open receivers (and I want the degree of coverage or openness). I want stats for individual offensive linemen that keeps a % of how many times they were beat or pancaked in a season.

It can be done.

159
by steelberger1 (not verified) :: Wed, 07/09/2008 - 7:05am

Sometimes new stats dont really improve on what they replace. Sometimes, you just have to look at your stats and apply some common sense.

Derek Anderson, for example, threw for a lot of yards (9th in the league) and TD's (5th).

He also only completed 56.5% of his passes (28th), threw 19 INT's (4th most), and averaged 7.2 ypa (14th).

I like this site, but I gotta say in this case Passer rating passes the sniff test on Anderson (82.5, 17th) and DYAR doesnt (10th).

160
by phill (not verified) :: Wed, 07/09/2008 - 8:39am

159 wrote: [quote]I like this site, but I gotta say in this case Passer rating passes the sniff test on Anderson (82.5, 17th) and DYAR doesnt (10th).[/quote]

Maybe you are misunderstanding DP/YAR then (he's 10th in DPAR also, obviously). He's 15th on DVOA, which (to my mind) is a meaure of how good he is. But DVOA is a normalised stat, where DYAR isn't. DYAR is cumulative. Which means that it increases proportinally to how many passes he throws. He threw more times than average, so his DYAR is higher as a consequence. Think of DYAR as DVOA x no. of passes thrown (x arbitrary constant).

(That's not quote true, since DVOA goes negative at average player level, while DYAR goes negative at replacement level, so they are zeroed at different points, but conceptually it gets the idea over).

DYAR is a measure of how valuable a player has been to the team, which depends on how the team uses that player quite considerably. DVOA attempts to measure how [i]good[/i] a player is.

(Which is why I never cared much for DPAR personally - DVOA all the way).

161
by Chris (not verified) :: Wed, 07/09/2008 - 9:20am

157- Vegas is way ahead of the curve.
158- That would be awsome and it will happen one day. Baseball already has all kinds of graphic tools that show that when a ball is hit into that exact spot in the outfield, the play is made 79% of the time etc.

If the whole question of tweaking DPAR was a measurement thing, then why not just convert 56 Dpar into 56.875 Dpar instead of converting to yards to be more precise?

The whole thing has me wondering about different style players.

Player A-Rod is a quarterback who feasts on weak defenses. He will rack up yard after yard on the garbage defenses of the NFL to the point where he is a stat king.

Player Horry is a quarterback who does ok against bad competiton, ok against average competition, but raises his game against the best on the biggest stages.

Traditional Stats love the A-Rod, but DYar loves Horry and gives him credit for beating big time opponents.

So who is right?

It is situational specific. There are trends to the whole mess of stats out there. Some pretty good teams feast on the weak but lose to the best, while other pretty good teams beat the best but don't show up against crap teams some nights.

162
by Suupa Jim (not verified) :: Wed, 07/09/2008 - 9:21am

Am I the only person who didn't even realize DPAR was calculating points? I always just kind of figured it was DVOA x the number of attempts or something.

163
by Robert Plant (not verified) :: Wed, 07/09/2008 - 9:53am

oh oh oh oh oh ohhh you don't have to go

164
by Temo (not verified) :: Wed, 07/09/2008 - 9:55am

161. Same here

158. Game charting information includes whose fault a sack was- particular O-Lineman, came free (on a blitz/blown pick -up), or coverage sack. It also includes whether or not the reciever was uncovered on the play (i.e. the infamous "hole in zone" and most RBs catches are uncovered receivers). There are also (I think- don't have a game chart in front of me) boxes for showing where a QB scrambled on a play, whether he passed the line of scrimmage or not. However, recording seconds of protection would not only cause a problem of subjectivity (there's a big statistical difference between 2.5 and 3.0 seconds, but to a human it might not be that great) as well as exponentially increasing the amount of time it'd take to do a whole game. I'm sure pro scouts do stuff like this however. But that's their job, this we do for free :)

165
by sam! (not verified) :: Wed, 07/09/2008 - 10:43am

161:

DPAR wasn't counting points. It was counting "Success Points" - awarded for the quantitatively-defined success of a given play.

166
by LooseOnTheLead (not verified) :: Wed, 07/09/2008 - 11:42am

DPAR wasn't analogous to yards, so why would one choose to call a re-scaling of DPAR "yards"? Inthe future, when I try to describe DYAR or True Yards to someone, I'm going to have to say something like, "Yes, 'yards' is in the name and it's sort of on the same scale as yards, but don't think of it as yards, exactly." How will that be easier to understand?

Maybe I'll change my mind as I become more accustomed to the new nomenclature, but for now, I expect that if I'm asked what DYAR stands for, I will answer, "Don't worry about that. It's just a number."

167
by LooseOnTheLead (not verified) :: Wed, 07/09/2008 - 11:48am

165:
DPAR wasn’t counting points. It was counting “Success Points” - awarded for the quantitatively-defined success of a given play.

DPAR was based on success points, but it was expressed in "actual" points--conceptually, at least. Here's a quote from the article above:

The solution to this problem was to translate "success points" into "actual points." I put together a complicated spreadsheet that estimated how many "success points" ended up in X number of points scored, combining both offense and defense, and we used that estimate to translate success values into DPAR.

168
by Chris (not verified) :: Wed, 07/09/2008 - 12:09pm

164- They do that in practice at the college level... of course they do that in NFL games.

169
by cjfarls (not verified) :: Wed, 07/09/2008 - 1:10pm

I'm really torn on the DPAR/DYAR question.

However, as I look at it, I think it is this:
DYAR in itself is confusing, and provides zero benefit (except for pirate talk) over DPAR. The scale still doesn't match with laypersons mental maps of traditional stats (because the baseline isn't included), so all it does is confuse the situation and create confusion by linking success points to a highly variable traditional stat.

However, I kind of like the concept of True Yards (but really hate the name, which I also think will be very confusable). I think there is value in creating a statistic on a similar relative scale to how we already think about player performance (aka 300+ yard passing games, 100+ yard running/receiving games, etc).

Remember, that "True Yards" aren't measuring yards... its simply a different way of expressing success points, but in a scale that is similar to what traditional stats use. But that is why the name should be totally different... you don't want people thinking they are in anyway related to traditional yards.

So i guess my ideal would be to report DPAR as the variable +/- replacement, and a renamed "True Yards" to translate that total into "layperson" terms.

Oh, and the new DVOA tweaks sound great... heading to the bookstore tonight...

170
by The McNabb Bowl Game Anomaly (aka SJM) (not verified) :: Wed, 07/09/2008 - 2:15pm

Here's why I don't like DYAR:

It claims to measure yards, but it really doesn't.

Imagine the following scenario: QB Manning throws for 315 yards against Pitt, a good pass defense. A hypothetical replacement QB (let's call him "Sorgi") would only have thrown 165 yards on the same number of attempts against such a good defense. So is Manning's DYAR 315-165=150? NO! It's not! Because DYAR is actually based on success points, not yards. So Manning's DYAR in that game is some other number, based on getting first downs and avoiding sacks and interceptions, etc.

So DYAR is not really "Yards above replacement." It's really "Success points above replacement converted into yards." But SPARCIY is not as nifty an acronym (though SPAR fits with the pirate theme). DPAR made sense as a measure of offensive contribution in points. But what does a yardage scale teach us? Really, nothing.

171
by Chris (not verified) :: Wed, 07/09/2008 - 2:36pm

In High school every offensive lineman got a grade. You either got a + or a - on any given play and every player received a grade. You were 73% efficient or whatever. The problem with that is it doesn't factor in quality enough. Did you cheesegrade your guy into the ground, or did you seal him off so he couldn't make the play?

Every evaluation method has its flaws, and DYar at worse looks to be named incorrectly.

172
by Temo (not verified) :: Wed, 07/09/2008 - 2:55pm

170. But you're assuming that DPAR meant something too. It wasn't DPAR=Actual Points-Replacement Points. So why should DYAR=Actual Yards- Replacement Yards.

173
by Temo (not verified) :: Wed, 07/09/2008 - 3:04pm

Although in a sense, don't we measure success points as the relative amount that a play helps you score points? So then saying "Looking at our history of data, "A" success points for a team correlates to "B" many actual points, defense adjusted... and since Manning had X success points, he contributed Y many actual points."

Turn that around into a statement relfect DYAR, and it doesn't make as much sense: "Looking at our history of data, "A" success points for a QB correlates to "B" many actual yards, defense adjusted... and since Manning had X success points, he contributed Y many actual yards." Assuming, of course, that success points is indeed predicated on measuring how many points an action is expected to contribute.

174
by GlennW (not verified) :: Wed, 07/09/2008 - 3:34pm

I have to concur with the majority opinion here on the preference for DPAR. With DPAR, I could make a simple league-wide calculation followed by a comparative player value assessment such as: "In a league where the average team scored 335 points, Tom Brady's (or more precisely, the Patriots passing offense's under Brady) performance was estimated at a whopping 188 points above QB replacement level".

I don't know how to make such a complete team-based performance evaluation with DYAR, rather only how to compare a player against his peers at the position. Intuitively I don't know what to do with it.

175
by Chris (not verified) :: Wed, 07/09/2008 - 3:59pm

In baseball Billy Beane would look at the effects that Player A in the lineup measured by how many more runs they score than player average. ( what people want Dpar to be).

He also studied how say... a lineup for of Kevin Youkalis's would score if Youk was cloned 9 times and took ever at bat for the saux to a different kind of value.

In football the work is a heck of a lot harder and there are at least 22 variables including the field, temp., wind, etc. etc. etc.

176
by Nathan (not verified) :: Wed, 07/09/2008 - 4:53pm

I have read many complicated articles on here, and this is the first one I haven't understood in the slightest.

After reading all the comments, I understand it and dislike it.

177
by qsi (not verified) :: Wed, 07/09/2008 - 5:22pm

I'd like more clarity on the relationship between DPAR and DYAR. From the DPAR ratings Aaron posted, as others have remarked already, it looks like there's 14.5 DYARs per DPAR. This works out at 43.6 DYAR for 3 DPAR ("FG") and 101.6 DYAR for 7 DPAR ("TD").

So if it's just a simple scale factor that converts DYAR to DPAR I'm left scratching my head over both the furore and the rationale for change in the first place. But whether this is true or not I'm still not clear on as it's not stated in the article.

What makes me suspect there is more to it than that is the fact that the DYAR/DPAR scale factor decreases steadily, though not monotonously, in the ten data points we have. This goes from 14.81 for Tom Brady down to 14.26 for Derek Anderson. If there is a non-linearity in the relationship we're not well placed to estimate it with this data because we know we don't have a population sample; we only have the extreme upper end of the distribution. On the other hand, it could simply be a random artefact of the data. Sample size is too small to tell.

Could one of the Outsiders shed light on this? I think this is a fairly central point as it would enable DPAR stalwarts (amongst whom I would include myself) to convert DYAR to DPAR. But then the rationale for the change boils down to marketing.

178
by Temo (not verified) :: Wed, 07/09/2008 - 5:52pm

177. The stated purpose all along was for "marketing" as you say. The only reason for this change is not that DPAR is broken in some fundamental way, but that they feel more casual football fans would understand DYAR more than DPAR.

By Aaron's rationale, we are more likely to say that Tom Brady threw for 400 yards rather than Tom Brady scored 28 points. Therefore, it would be better to say that Tom Brady contributed 200 more yards of offense (notice I didn't say he threw for 200 more yards) than a replacement QB would have (200 DYAR) rather than saying that Tom Brady contributed 14 more points of offense than a replacement QB would have (14 DPAR).

The whole purpose is to make the FO stats more "accessible" to the average fan. Of course, these numbers are rough and maybe I'm getting some stuff wrong. But that's my understanding.

179
by Tom D (not verified) :: Wed, 07/09/2008 - 6:03pm

Re 174:

That seems like a pretty easy change to me, just compare Brady to how many yard the average offense gained (which is about 5 300).

180
by DPARDIEU (not verified) :: Wed, 07/09/2008 - 9:08pm

DPAR better. DPAR will make pro bowl this year.

DYar got no rigns. Worst stat.

181
by mrh (not verified) :: Wed, 07/09/2008 - 10:02pm

Re 119 - I was thinking of you when I posted 107.

182
by Pat (not verified) :: Wed, 07/09/2008 - 11:50pm

#170: Imagine the following scenario: QB Manning throws for 315 yards against Pitt, a good pass defense. A hypothetical replacement QB (let’s call him “Sorgi”) would only have thrown 165 yards on the same number of attempts against such a good defense. So is Manning’s DYAR 315-165=150? NO! It’s not! Because DYAR is actually based on success points, not yards. So Manning’s DYAR in that game is some other number, based on getting first downs and avoiding sacks and interceptions, etc.

Over a large number of games, it would be ~150, if you normalized it right. That's the entire point of figuring out a conversion between success points and yards or success points and (game) points.

I really don't see the reason for caring.

183
by dmb (not verified) :: Thu, 07/10/2008 - 12:54am

So ... any chance of DPAR pulling a Lazarus?

I agree that even if the new stats are kept, "True Yards" really needs a new name.

And I think that The McNabb Bowl Game Anomaly had the best three posts in this very long thread.

184
by PantsB (not verified) :: Thu, 07/10/2008 - 1:09am

Since the feeling seems clearly in the pro-DPAR camp, is there any reason not to maintain both? If it runs off essentially the same data (maybe that's an incorrect assumption on my part) the actual processing should be trivial/automated.

I've always felt one of FO's strongest foundation points is that yards can be deceptive, so using adjusted yards as the base stat seems wrongheaded to me.

185
by D (not verified) :: Thu, 07/10/2008 - 2:10am

Thanks #137 that seems like a pretty damn good approximation.

186
by Scott C. (not verified) :: Thu, 07/10/2008 - 6:49am

MOST people here are missing the meaning of True Yards, and the "problem" with it.

Its pretty simple folks!

Why does Brady have 6500 True Yards? It is NOT saying that if he played an average defense for 16 games he'd throw for 6500 yards.

It is saying:

If he played a league average defense for 16 games, it would be as valuable as 6500 yards by an average QB -- but the average QB's yards are not as valuable as his!

Why? Well TOUCHDOWNS and FIRST DOWNS are worth success points. Thus, they are worth DPAR, and since DYAR is just:

DYAR = DPAR * ( some constant )
We're basically just converting points into "point-equivalent yards".

SO, the whole ugly mess is that although DYAR works well in many contexts, True Yards is HORRIBLY misleading.

Its useful, but in Brady's case its converting all those touchdowns into yards, and not subtracting a lot of INT's or fumbles.

Think of it as:
Defense adjusted yards, with bonus yards for touchdowns, negative points for bad plays, and scaled for efficiency and down/distance.
In other words, True Yards has NOTHING to do with comparing to real yards.

AT ALL.

187
by DYARjoe (not verified) :: Thu, 07/10/2008 - 9:05am

Dyar rocks with a patch over its eye, Dyar to champion for next 8 outsider books. Dpar is dead for good. Jemarcus Schatz to win league MVP not even close.

188
by Patrick (not verified) :: Thu, 07/10/2008 - 9:35am

wow, tight ends look horrible in True Yards.

looking at the list, the best TEs in the league still had fewer True Yards than real yards,

indicating that the best TEs in the league are below average. Apparently.

189
by LooseOnTheLead (not verified) :: Thu, 07/10/2008 - 10:33am

178:
"By Aaron’s rationale, we are more likely to say that Tom Brady threw for 400 yards rather than Tom Brady scored 28 points."

And therein lies the other half of my objection to DYAR. (The first is that it isn't yards.) Most fans I know of do NOT compare QBs by yards. They compare them by passer rating, of which a yardage state is only one component. Passer rating is actually a nice case-in-point for this discussion. It's widely used and the average fan is comfortable with it, yet it is on a scale that doesn't relate or translate directly to yards, touchdowns, or anything else that would otherwise have been familiar. With respect to fans' comfort zones, about the only thing the scale for passer rating has going for it is that it looks very roughly like a grading scale from school. (You could say that a rating in the 80s gets a 'B', for example.) But even that relationship is weak...and, come to think of it, it doesn't exist for passer ratings for college QBs, yet fans still have little apparent trouble with it.

There is just no good case for a metric called something yardage-related or scaled roughly like yards. It only increases confusion.

I realize that whenever something new is introduced, there will be those who resist the change. That's normal and doesn't suggest, obviously, that changes should never be made. I'm trying to keep that in mind, but...I still strongly dislike this "improvement". I simply cannot see how it improves anything.

190
by Temo (not verified) :: Thu, 07/10/2008 - 10:44am

187. DYAR measures how much better than "average" (not quite average, but close enough) a player is, not True Yards. True Yards=DYAR+Replacement Player's Yards. It doesn't *really* tell you anything about actual yards.

188. I agree with a lot of your points, I was just trying to explain what the rationale was. Hopefully Aaron can clear up a lot of stuff in that mailbag he's supposed to write.

191
by The McNabb Bowl Game Anomaly (aka SJM) (not verified) :: Thu, 07/10/2008 - 10:46am

Re: 183

Thanks.

Re: 186

Maybe "True Yards" should be renamed "True Yardage Value" or something like that, because what is true about them is the value, not the yards.

If you're right about what "True Yards" actually means, then it's even less useful than I had thought, since it's really hard to get a feel for what that number means. I mean, Brady threw for as much value as if an average QB had thrown for 6500 yards, so that's good, but how good? On how many attempts? Do we need another stat called "True Yards per pass attempt"? Or would that be exactly the same as DVOA? And if so, then what do true yards actually add to the conversation? (Saying "as good as an average QB throwing for 6500 yards" is not really useful without considering the number of attempts. I mean, it's obviously very good, but what happens when we start looking at guys who are not huge outliers?) So basically the question is this: Are True Yards per pass attempt identical to DVOA? If not, then TYPPA is probably way more useful than simple True Yards. If TYPPA are the same as DVOA, then I'm still not sure why True Yards are useful.

Re: 187

There's something screwy about those numbers. Players with high DYAR and high DVOA should not have fewer True Yards than actual yards, I'm pretty sure. I think Aaron said he's still tinkering with the formula so maybe it will be resolved by an update.

192
by GlennW (not verified) :: Thu, 07/10/2008 - 11:12am

> That seems like a pretty easy change to me, just compare Brady to how many yard the average offense gained (which is about 5300).

Yeah, I guess that's true. As with others, I just feel that a points-based system has a more intuitive feel to it, as a derivative value statistic. With DYAR, in my mind I somehow want to compare the stat with the raw yardage stats we know and hate, when that's not at all what the stat represents.

FO's rationale that "(football fans) think in terms of yards" runs counter to their original objective, I think-- we want to get away from that kind of thinking. As someone else suggested, that's tantamount to baseball sabermatricians moving from a runs-based system to adjusted batting average just because old-school fans love BA, even though the adjusted BA statistic will have nothing to do with actual batting average. Sure, I can still use the stat, but I already intuitively know what a run represents in the game, and don't want to have to make another mental calculation to understand what a point of adjusted BA represents.

193
by bravehoptoad (not verified) :: Thu, 07/10/2008 - 11:23am

re: 97

As for DYAR, I’ll address this further when I do a “DYAR Mailbag” but it is the exact same stat as DPAR, just on a different scale. Basically, we’ve gone from Fahrenheit to Celsius. Freezing is still freezing.

If this is true, then it should be easy enough to supply a simple algorithm for converting from DYAR to DPAR for all those people who are missing it.

194
by Bad Doctor (not verified) :: Thu, 07/10/2008 - 11:27am

Thanks to everyone who posted, this is a very good discussion, and I'm learning a lot.

I think my initial aversion to the DYAR switch was because I thought DPAR was based on true points. As many of you have pointed out, it's really success points. So I'm learning that DYAR isn't weak, DPAR just wasn't as strong as I thought it was.

Clearly, the best answer for any advanced football stat is to move from yards to points to wins, just as advanced baseball stats have moved from bases to runs to wins. But I think Aaron's dad had a saying about that sort of thing. Since we're not moving backwards from points to yards, the move seems fine ... well, scary now, but I'm sure we'll all adjust to the new scale. I do wonder if just using "success points" as the measure wouldn't work better than confusing things with mention of points and yards. QB rating got mainstream acceptance, even though it works on its own scale. Game Score sticks out as an advanced baseball stat that did the same.

And even if I didn't mostly "approve," it bears mentioning that the FO crew is doing a fantastic job. Can't wait for my PFP08 delivery and the season to come.

195
by Temo (not verified) :: Thu, 07/10/2008 - 12:03pm

194. I'm sorry, but I don't think you're exactly right on DPAR (although I could be wrong). To my best understanding, it's success points as they correlate to actual points. In other words, since Tom Brady has 50 success points (made up scale), and 50 success points usually leads to 28 actual points, we say that Brady has DPAR=28- 14(arbitrary replacement level performance).

196
by Lou (not verified) :: Thu, 07/10/2008 - 12:27pm

As for DYAR, I’ll address this further when I do a “DYAR Mailbag” but it is the exact same stat as DPAR, just on a different scale. Basically, we’ve gone from Fahrenheit to Celsius. Freezing is still freezing.

you know theres a reason most thermometers have both scales on them- to be useful to the widest possible customer base. if you think DYAR will appeal to the more conventional fan thats fine, but by eliminating DPAR i think you're eliminating part of the potential customer base that would find that stat more useful. in short, i welcome these new stats but please bring DPAR back.

197
by GlennW (not verified) :: Thu, 07/10/2008 - 12:29pm

195: That's correct, DPAR did represent (or attempted to) the equivalent of honest-to-god points put up on the scoreboard above what a replacement player would have produced (which is why I liked it). From Aaron, above:

'The solution to this problem was to translate "success points" into "actual points." I put together a complicated spreadsheet that estimated how many "success points" ended up in X number of points scored, combining both offense and defense, and we used that estimate to translate success values into DPAR. For example, Carson Palmer in 2007 was worth 51.8 passing DPAR.'

198
by DGL (not verified) :: Thu, 07/10/2008 - 1:37pm

So since people think True Yards is misleading, why don't we go with the pirate theme and call them Adjusted Yard Equivalents. AYEs, me hearties.

199
by Pat (not verified) :: Thu, 07/10/2008 - 1:55pm

If this is true, then it should be easy enough to supply a simple algorithm for converting from DYAR to DPAR for all those people who are missing it.

It is. Divide DYAR by about 14.5. Poof, DPAR, within a percent or so.

I really, really don't get people's objections. Get rid of the whole "success points" thing in your head. They don't matter. They're a method of converting "real football" into "ideal football." That's all. Then when you want to move back from "ideal football" stats to "real football" stats, you can present those stats as yards, points, or whatever else you want. You could present them as turnovers, for that matter, since in "ideal football"-land, yards result in points, turnovers result in points, kickoff yardage results in points, etc.

One of the previous suggestions was that "fans compare quarterbacks by passer rating." You can't do that, since passer rating is a rate stat and yards/points/TDs/INTs/etc. are counting stats.

You can see this obviously because a player who attempts one pass can have a perfect passer rating.

You could convert DVOA into a passer rating, presumedly, but in my mind that seems a bit silly (but who knows).

The whole mess with True Yards I don't get either. (The people who are complaining about the name I really don't understand: a "True Football Yard" is worth 7/100 points. The fact that a yard is 36 inches is ridiculously immaterial to the discussion.) If you take DYAR and add replacement level yards back in, what you should get is "ideal football yards," which is "have Brady play the same games he did, several hundred times, and his total yardage should average out to something like this."

If that isn't what it does, then the conversion from success points to yards is wrong.

200
by NewsToTom (not verified) :: Thu, 07/10/2008 - 2:22pm

Re #199
If that isn’t what it does, then the conversion from success points to yards is wrong.
Ding, ding, ding! I think that's what it's supposed to be. I'm not sure that's what we're actually seeing, though, thus the ongoing work on True Yards.

201
by GlennW (not verified) :: Thu, 07/10/2008 - 2:36pm

> Divide DYAR by about 14.5. Poof, DPAR, within a percent or so.

Why should we have to do that though? I need a calculator to do so, and fundamentally it's a step backwards regardless.

I'll admit, as a SABR-type I'm a relative simpleton. I understand the concepts, but don't want to have to make any of my own calculations or adjustments-- I just want to crack the book and take it all in. There's an issue of standardization here. Go back into the archives and there are hundreds of discussions based on DPAR, which used to be a fundamental statistical building block. I do want to go back to my last three copies of Football Prospectus and have the data mean something (even if minor adjustments have since been made, which is fully understood). But poof, DPAR is now gone. To be blunt, yes, it's aggravating. Even though it's an evolving discipline, if these kinds of wholesale basis shifts continue to be made, I don't know that I'd want to continue to pay money to keep up with them.

Sincere question-- was any kind of market research done here? Even an informal customer survey or the like?

202
by DGL (not verified) :: Thu, 07/10/2008 - 4:13pm

Pat @#199:

"...a player who attempts one pass can have a perfect passer rating."

Only if it is complete for at least 13 yards and a touchdown.

203
by Peter Libero (not verified) :: Thu, 07/10/2008 - 4:39pm

202: ...meaning it can.

I honestly am a little shocked at how few people understood DPAR previous to this change, and even more shocked at how few people understand the "change" (which it's really not) to DYAR. Kudos to Pat for clearing things as much as possible.

Seriously though, I mostly agree with Phill in #160... DVOA is what this site is actually based on, DPAR is just a way to account for our hypothetical 1-attempt guy.

204
by bubqr (not verified) :: Thu, 07/10/2008 - 5:05pm

So everyone agree that the main debate is more around : "Why replacing DPAR by Dee Yarr when you can have both ?"

And btw, DYAR sounds awful in french. I much prefered DPAR(Not that I know anyone in France who knows anything bout DPAR but me..)

205
by kidneypuncher (not verified) :: Thu, 07/10/2008 - 6:47pm

Hey anybody,
Is there any discount on kubiak if the book is bought?

206
by MJK (not verified) :: Fri, 07/11/2008 - 11:28am

Totally immaterial to the argument, but I kind of like the way DYAR sounds. It goes well with Mike Tanier's rather piratical appearance. (Apologies if that's already been said--haven't had time yet to read the whole thread).

207
by anotherpatsfan (not verified) :: Fri, 07/11/2008 - 12:29pm

205 -- no discount for book/Kubiak that I know of. I do get both the book and the website Kubiak, but I barely even look at the Kubiak in the book (and focus on the other stuff in the book). As far as the projections go, the downloadable spreadsheet is more fun to mess with, is more informative (and the ability to see and mess with colors makes it easier to read), can be customized for scoring systems, and is updated numerous times between now and the beginning of the season, and at least once or twice early in the season.

IMO the download spreadsheet is a potentially valuable tool in the fantasy geek arsenal (as is the book, with its team and individual comments), and if you can swing it, do both.

208
by cjfarls (not verified) :: Fri, 07/11/2008 - 12:29pm

My understanding is the biggest benefit of DYAR is that it can then be adjusted to get to "True Yards" (horrible, confusing name, but I do like AYE)... which I think is a useful stat for communicating with lay-people (DYAR is not because it doesn't have the replacement baseline built in).

The disadvantage is you lose the history of DPAR.

Which is why I suggest keeping DPAR, but adding DYAR. In theory, if DPAR and DYAR are just F vs. C measurement scales, it shouldn't be difficult to keep DPAR and still add "True Yards" (shudder) for the mainstream articles.

209
by cjfarls (not verified) :: Fri, 07/11/2008 - 12:30pm

Dyaaarrr-gh.... I meant "keep DPAR and add true yards."

210
by Scott C. (not verified) :: Fri, 07/11/2008 - 12:32pm

#199:
If you take DYAR and add replacement level yards back in, what you should get is “ideal football yards,” which is “have Brady play the same games he did, several hundred times, and his total yardage should average out to something like this.”

If that isn’t what it does, then the conversion from success points to yards is wrong.

That ABSOLUTELY is not what it does. See several posts in the last 25 or so, including mine at 186.

True Yards = DYAR + Replacement Level.

This is really screwed up. VERY screwed up. It seems as you say, at first pass to indicate what someone would have done against a replacement level defense, but that is far from the truth.

DYAR is the "yardage worth", in point terms converted to a yardage scale, of a player over replacement level.
For a QB, this would mean various things that we are all familiar with:

Lets use the Peyton Manning example from the article:

13 for 17 for 249 yards, 4 TD's and 0 INTs against Baltimore.

228 DYAR!
322 True Yards...

Why? Well a replacement level QB, say Rex, would be expected to put up 94 yard-equivalent-success-points in 17n attempts against Baltimore.

No, its NOT saying that the 322 - 228 means that the replacement guy would only throw for 94 yards. Its saying the yard equivalent of success points. That probably translates to something more like 140 yards, 1 TD, and 2 INTs with a couple sacks.

Now, if Peyton had such a good day against a different defense, True Yards would have NOTHING to say about his yardage output there either! Remember its just DPAR * some number! It would have something to say about his VALUE against that defense, and first downs, TD's and efficient play raise his True Yards but have zero affect on his Real Yards.

So, True Yards is extremely inflated for a QB with good DVOA, if you are trying to take it for what it sounds like -- and deflated for one with low DVOA.

Again, look at the Brady example from this last year. It is NOT saying he would throw for 6428 yards against average defenses ... he played average or below defenses last year.
Its really saying, this:

An average QB against an average set of defenses would throw for X yards in a season with (tom brady # of attempts).
But Tom Brady had a DVOA of 56.9%.

So Tom Brady's True Yards is X * (100 + 56.9)% = 6428.

Thus, a average QB (DVOA = 0%) against an average schedule with the same # of attempts as Tom Brady would throw for 4097 yards.

But any player with DVOA greater or less than 0, you can't line True Yards up with expected yards.

As DVOA goes up, the value of each yard goes up (more points per yard, due to better play and efficiency).
As DVOA goes down, the value of each 'real' yard declines.

True Yards unfortunately adjusts for the yardage value by inflating the number, which makes it no longer correlate with what a fan thinks of as yards...
After all, its JUST DPAR * (~14.5).

But if you know what it is really doing, its a good stat. But the name and concept IMO is horribly flawed and confusing to most people.

I liked it when I first saw it. But thats because I'm not an average fan, and with a heavy science background I looked past the name to the numbers and their derivation.

But I think this re-normalization to "Yard equivalent succcess points" instead of "Point equivalent success points" is a really really really REALLY bad idea.

Sure, the average fan will "get" yards a bit more. But they STILL won't get DYAR any more than DPAR, with its negative numbers and all.
DYAR is misleading in itself though. When LT throws for 1 pass for a TD in a game, his DYAR for that game may be more than his actual yards.
From a layman's perspective --
Actual Yards In Game - DYAR = expected yards for a replacement. But this again is NOT what the numbers mean.

If LT does his thing and thows the option pass for a TD again his passing actual yards - DYAR will be negative.

In fact, on a play by play basis, things are all odd -- an incomplete pass is negative DYAR.

Sorry for the long post but lets do one more thought experiment:

An incomplete pass on third and 3.
An average QB would be expected to make this play. Lets say that in the old world scheme, this was a first down 2/3 of the time leading to say, 0.5 success points (normalized to points) for one play. Now, the replacement QB will be worth something like 6 "yard - equivalent success points".
If a QB makes an incomplete pass on such a play, the play itself will be slightly negative, perhaps -4 "yard equivalent success points". So, on this play the DYAR would be -10.

True Yards would be -4.
The actual yards are 0, and actual yards - DYAR is -10.

None of these numbers have anything at all to do with the expected actual yards of any players in ANY circumstance.

True Yards should be re-named to:

Yardage Value or
Total Value Yards or something else that makes it clear that the measuring stick here is
NOT YARDS. AT ALL.

Its really points, normalized to the average # of yards per point. And those points aren't real points, they are "value points" -- assigning point values to plays based on down/distance and importance.

IMO, this does not make the numbers more accessible, it adds one extra layer of complexity. Its a good stat, but the names are misleading, and throwing out and not printing DPAR is a bad idea.

Its going to be hard to describe to someone how a 9 yard reception on third and 11 is worth negative 'yards' in some cases and small positive 'yards' in another. Its FAR easier IMO to use points as the base scale. Anyone confused by points, will be confused moreso by yards.

211
by Patrick (not verified) :: Fri, 07/11/2008 - 12:45pm

@190: "DYAR measures how much better than “average” (not quite average, but close enough) a player is, not True Yards. True Yards=DYAR+Replacement Player’s Yards. It doesn’t *really* tell you anything about actual yards."

from the article: "players with more True Yards than standard yards played better than standard stats would otherwise indicate, while players with fewer True Yards than standard yards played worse than standard stats would otherwise indicate"

Which is to say, apparently: "The best Tight Ends in the league, as defined by our own metrics, played worse than standard stats indicate."

that statement is brain melting.

212
by Temo (not verified) :: Fri, 07/11/2008 - 12:59pm

So I had to look up piratical just to see if it really was a word-- and much to my amazement, it indeed is an adjective form of the word "pirate".

213
by Patrick (not verified) :: Fri, 07/11/2008 - 1:00pm

@164: "However, recording seconds of protection would not only cause a problem of subjectivity (there’s a big statistical difference between 2.5 and 3.0 seconds, but to a human it might not be that great) as well as exponentially increasing the amount of time it’d take to do a whole game. I’m sure pro scouts do stuff like this however. But that’s their job, this we do for free :)"

There are only about 3.5 sacks in a given game, and the timing is not nearly subjective as many of the things that are charted; such as whether the QB is hurried, and whether a given RB is covered by an LB in the vicinity or by hole-in-the-zone.

For the "exponential increase in game charting time", again with only 3 or 4 sacks in a game, it simply wouldn't take very much time, especially not compared to things that are currently charted, such as which corner is in coverage etc.

214
by Scott C. (not verified) :: Fri, 07/11/2008 - 1:19pm

Here is my main issue with the yard scale:

Its misleading. Why? Because the ENTIRE FOUNDATIN of Football Outsiders stats is POINTS.

Here is how I see it:
Football Outsiders statistics starts out by finding the value of plays, in points. It does this by looking at the down/distance/situation before and after a play:

How many points does a drive attain on average at first down and 10 from mid field?
How many points does a drive attain on average on second down and 5 from the opponent's 45?

These two point values differ slightly.

That first down play for 5 yards in between those two points is worth a value in POINTS that is the difference between the values above.

Thus, any play can be tied to expected value point differential. This needs further adjustments, but as far as I can tell, is the foundation of the statistics at F.O.

Once you do all that work, if you convert it to yards with a singular SCALAR, you mess it all up. Now, the system is predicting point values, using a scale normalized to yards. But normalized to one person's yards --- the
"average team" or "average player" 's yards. These yards are of course different in value to other team or players yards because not all teams, players, or plays are as efficient at at turning yards into points.

If you want to have a stat that predicts how many yards a player would have against an average defense, or any other question that has "how many yards" in it, you would need to build the FO stats up from the ground up again on a different fundamental question:

How many remaining yards on average does a drive go that is first and 10 at mid field?
How many remaining yards on average does a drive go that is second and 5 from the opponent's 45?

The difference in yards of those two questions leads to a system that could generate "ACTUAL" DYAR -- that is a DYAR that is making predicitions about yardage. The current scheme predicts points converted to yardage scale.

Now, of course predicting yardage is useless, since the entire concept of using points in the first place is that points win games, not yards. And an "ACTUAL" DYAR stat would be a great predictor of how many yards a player / team would get against average teams, or other teams, but yards again, don't translate to wins. You'd get lots of funny effects -- bad special teams plays often lead to more offensive yards. Turnovers in the first half lead to more passing yards in the second (on average).

Anyhow, thats my fundamental disagreement here. The whole system is based on points -- not actual ones, but very very close to them -- since points win games.

Since points correlate with winning much more highly than yards, this is a step the wrong way. Furthermore, since its points based underneath, the translation to a yardage scale confuses the issue.

NONE Of these -- DYAR, True Yards, or the replacement or average levels of them, are meant to predict a single thing about yards. They predict success, and points.

215
by nat (not verified) :: Fri, 07/11/2008 - 1:31pm

"True Yards" is a horribly misleading name. They are neither true nor yards. Nor are the yards in DYAR the same scale as real yards.

Let's just call them "doubloons". They are an arbitrary measure of value that includes yardage gained, first downs, TDs, and turnovers, and adjusts for down and distance.

QBs and WRs get high doubloons per yard. RBs get a mix. TEs get low doubloons per yard.

That makes a kind of sense, when you consider play designs. It doesn't mean TEs are all below average. It means that passes to the tight end are usually the result of a play that wasn't as valuable as hoped for.

216
by Temo (not verified) :: Fri, 07/11/2008 - 1:35pm

213. But why are we only measuring seconds until a sack? I had thought that the purpose was to measure pocket protection in terms of seconds afforded for a QB to throw the ball, so you measure seconds on all pass plays.

Otherwise, you're really only measuring if a sack is due to poor pass protection (say if the QB gets sacked 2 seconds after dropping back) or due to the QB holding the ball too long (say if the QB gets sacked 5 seconds after). In which case, you're not giving credit to the O-Line for offering a QB many seconds of protection in which to throw the ball, after which he does not get sacked but completes the pass for a big gain. So what happens when an O-Line gives its QB loads of time for all pass plays, but the one play he gets sacked, the RB misses his blitz pickup and the defender comes through untouched in 2 seconds?

217
by nat (not verified) :: Fri, 07/11/2008 - 2:15pm

Calling your yards "True" is like calling your news "Fair and Balanced".

It's a dead giveaway that something has gone terribly, tragically wrong.

218
by Nathan (not verified) :: Fri, 07/11/2008 - 2:52pm

I really want Aaron to address Scott C.'s excellent posts.

219
by Patrick (not verified) :: Fri, 07/11/2008 - 3:16pm

@216. I now see that 158 or whoever was suggesting charting every pass play for protection time, which I wouldn't argue is necessary.

I guess there are two very worthwhile questions when it comes to pass rushing that can be assisted by game charting:

(1) Who is to blame when sacks happen?

(2) Who is credited when sacks don't happen? (every play where a pass is attempted)

You could argue about which of these answers is more valuable, or you could say they are about equal. I don't know, but I think (1) is a little more important. Just me.

What you can't argue about is that (1) would take an order of magnitude less work to 'answer' than (2).

I realize that (1) only wouldn't give a complete picture, but I would argue it would be very worthwhile. And not just comparing O-lines and QBs,

but consider, e.g., two defenses, each pick up 40 sacks in a season. However, with this you'd be able to say "well, defense A got their sacks on average at the 2.9s mark, where defense B got theirs closer to 4s. It's likely that A has much better pass rushers and B a much better coverage secondary."

Again, (1) only gives a partial picture (maybe defense A shoots right through the O-line on 5% of plays, but allows all day to pass on the other 95%, where defense B gets to the QB around the 4s tick every play),

but it would give you a lot of help evaluating QBs, O-lines, DEs (DEs A and B both had 8 sacks, but all of Bs were after the 5 second mark)/individual pass rushers, defensive lines, secondaries.

That's a fat wad of untapped information. 3.5 plays per game.

220
by resident jenius (not verified) :: Fri, 07/11/2008 - 3:17pm

DYAR = BAD
DPAR = GOOD

Bad Aaron. BAD! No Cookie!

221
by Matt (not verified) :: Fri, 07/11/2008 - 3:33pm

For those curious, the replacement level is 4.93 yards per pass attempt. However, it is 5.63 yards per equivalent pass attempt, "equivalent pass attempts" being the translation of the "league-average" baseline on all that quarterback’s passes into a pass attempt number. The reason for the difference is that there end up being more "success points worth of pass attempts" than there are actual total pass attempts. Yes, this is complicated, but the good news is that nobody needs to see the gears behind the curtain on a week-to-week basis.

YIKES!

222
by Pat (not verified) :: Fri, 07/11/2008 - 4:15pm

Just as a hint to everyone, if you read right above the post entry point, it notes that HTML italics tags are allowed. Please, please, please use them instead of caps. Caps just makes you sound like an idiot.

This is really screwed up. VERY screwed up. It seems as you say, at first pass to indicate what someone would have done against a replacement level defense, but that is far from the truth.

Why would it indicate what someone would've done against a replacement level defense? (what good would that be? 'If Brady had played the 49ers last year 16 times, he would've put up X yards!'). It represents what someone would do against an average defense. Not a replacement level defense. Where did your idea of a "replacement level defense" even come from?

It's an average defense. That's what the defense-adjusted part is for. You add in 'replacement yards' because those are the yards that a replacement player would have gained against that average defense.

It won't "inflate" players with high DVOA. It will "inflate" players with high usage and low DVOA, who would typically get pushed down because they're contributing less than a replacement player would (but still contributing).

But that's what yards do, anyway, and people intrinsically do compare starters-to-starters anyway.

But if you know what it is really doing, its a good stat. But the name and concept IMO is horribly flawed and confusing to most people.

I don't get it. What I would interpret "True Yards" to be is "These are the yards that player X would've gotten had he faced an average defense and had the season been played a few hundred times."

Maybe I would interpret it as "these are the yards that player X would have gotten had the season been played a few hundred times," but certainly not the way you were saying it.

It would have something to say about his VALUE against that defense, and first downs, TD’s and efficient play raise his True Yards but have zero affect on his Real Yards.

I think you're not understanding how you would convert from success (or value) to yards.

All you do is say "what's the average yardage gained by players who gain X value in a season"? Poof. There's your conversion.

You're not "converting" first downs or interceptions or TDs into yardage. You're averaging them out.

Actual Yards In Game - DYAR = expected yards for a replacement. But this again is NOT what the numbers mean.

Uh. No? Actual yards in game - DYAR = something bizarre. Actual yards need to be adjusted for the defense, so you'd get DY - DYAR = expected yards for a replacement.

Actual Yards in Game - YAR would be expected yards for a replacement. There you'd have Y - (YAR) = R (since YAR = Y+R).

"True Yards" might be missing a "D", but "true" is kindof a catchall for "corrected for just about everything" anyway.

Football Outsiders statistics starts out by finding the value of plays, in points. It does this by looking at the down/distance/situation before and after a play:

No.

Read "Our New Stats Explained" again. Carefully. It starts out by calculating what Aaron used to call "V+". Value. Not points. Aaron tends to confuse the issue a little by calling the units that V+ is measured in "points." But it's not points. It's value. He never should have called it "success points." He should've called it "success," period, or just "V" (or "V+", or whatever).

Value is measured on an arbitrary scale. It's intended to normalize the output of a football play to a common scale, because football has no common scale inside the game itself.

But fundamentally, "success" can be translated back into yards, interceptions, field goals, first downs, safeties, touchdowns, or whatever the heck else you want.

Its connection to points as in points-on-a-scoreboard is just in the sense that when you normalize the scale to league average, you get a high correlation with actual points scored. But you also get a correlation with yards, so there's no big deal.

223
by The McNabb Bowl Game Anomaly (aka SJM) (not verified) :: Fri, 07/11/2008 - 4:19pm

I've said my piece on here, so I'll just say that Scott C. seems to understand what these stats mean better than most of us (including me) and we should all give his arguments serious consideration (including the Outsiders).

224
by Tally (not verified) :: Fri, 07/11/2008 - 4:24pm

Re 210:

Ditto 222. The biggest misconception here regarding DPAR/DYAR is that people think the stats here were always based on points.

DPAR is as arbitrary as DYAR; it's simply a unit designation, like inches or centimeters.

225
by The McNabb Bowl Game Anomaly (aka SJM) (not verified) :: Fri, 07/11/2008 - 4:38pm

Re: 224

True, but DYAR and (especially) True Yards are more misleading than DPAR, because they sound like something they are not.

DPAR also sounds like something it is not, but if you make that mistake, it's still less confusing and you're less likely to draw an incorrect conclusion from it that with DYAR and True Yards.

226
by Temo (not verified) :: Fri, 07/11/2008 - 4:46pm

222. You've (maybe) answered the big question that I myself had when I made an arguement somewhat like what Scott C. made. I stated that all of what I was talking about (about the relative value of DYAR v. DPAR) is contingent upon "Assuming, of course, that success points is indeed predicated on measuring how many points an action is expected to contribute."

Now you say that it is NOT. But that leaves the issue of what exactly determines how much Value a play has in terms of V+ (or whatever). If the success value of a play is not the point expectation added by the result of the play, then what exactly is it? How do we know that 4 yards on 2nd and 5 at midfield has a value of X while 6 yards on 2nd and 5 at midfield has a value that is, say, X*1.5 (or whatever)? The most logical thing I could come up with is that this value comes from the added expectation of points.

IE, 2nd and 5 at midfield has a point expectation, on average, of 2 points. 4 yards will set up 3rd and 1 at the 46, which has a point expectation, on average, of 3 points... so the success value of the play is 1. On the other hand, getting 6 yards will set up a 1st and 10 at the 44, which lets say has a point expectation of 3.5... so the success value of a 6 yard gain is 1.5.

Has the actual calculation (or even just the components) ever been revealed? Is this proprietary information?

227
by Temo (not verified) :: Fri, 07/11/2008 - 4:49pm

By the way, what I stated above is how baseball prospectus measures the value of stuff (baserunning, for example). Which is why I always assumed football outsiders functioned likewise.

228
by Cheech (not verified) :: Tue, 07/15/2008 - 10:08am

So Aaron is the DYAR Mak'er that Led Zeppelin has been singing about all these years! He might even know whether or not the mudshark incident is truth or rumor.

229
by bob40 (not verified) :: Tue, 07/15/2008 - 11:31am

Whew! Read through all the posts and got DYAR clear in my mind. Then got confused, but then got clear again but then got confused again. I can't tell if I'm clear or confused at this point. I'll read the posts and the original explanation all over again. There seem to be some excellent clarifications in the posts, some of which are then re-labeled as incorrect.
Cover me, I'm going back in.

230
by Ronjan (not verified) :: Sun, 08/17/2008 - 8:32am

I'm a bit surprised. If you're going to talk accessibility, DPAR was never the problem -- DVOA was.

DVOA represents itself as a ratio (percent improvement in performance) without being clear, at least to me, how much better is good and how much worse is bad. If my DVOA is +13%, is that guy 13% better than average? What does that really mean?

If anything DVOA on a "points per play" type scale would have been better.

Additionally, while there are certainly practical reasons for "more points = positive defensive DVOA", that's very counter to accessibility. If offensive DVOA and total DVOA both have "better performance = positive DVOA", defense should follow suit. I know this has been a point of contention for years, but taking a step back, if the goal is accessibility, all DVOAs need to have better performance as positive.