TCU has played much better in the second half of games this year. What other schools have seen dramatic shifts of play after halftime?
03 Aug 2005
by Aaron Schatz
We've received a number of questions and comments since the release of Pro Football Prospectus 2005. (On sale now!) We've put the best ones together in this mailbag -- which can also be used as an open thread to discuss the book.
Don't forget that we have a new contact form which you can use to e-mail any of the writers. We'll be doing a regular weekly mailbag during the season, and to be honest asking a question through the contact form is probably a better way to get your question answered than asking it in a discussion thread.
Be aware that I reference plenty of our innovative FO stats here, not to mention their unfamiliar terminology, so if you are a recent addition to the readership you might want to read this first.
I should also mention the Pro Football Prospectus 2005 book tour, coming soon to a town near you. All of the cities, times, and locations are listed here but here's the city-by-city rundown:
August 8: Boston
August 9: New York
August 10: Philadelphia
August 11: Pittsburgh
August 16: Indianapolis
August 17: Detroit
August 18: Minneapolis
August 20: Chicago
August 22: Denver
August 23: Dallas
September 6: Washington DC
September 7: Philadelphia, again
There is also a big radio day on Thursday August 4 where I'll be appearing in most NFL markets that aren't on the actual tour. You can find all those times and stations here.
Before we start in with the questions, I should apologize for the fact that there are a few errors and misspellings in the book. I know that some people feel that a book with misspellings automatically must be dismissed as the work of idiots who don't know anything about football. For the rest of you, the mistakes are just annoying. Trust me, I feel the same way.
The worst of the errors is the misspelling of Matt Hasselbeck's name right on the cover, but that one isn't actually our fault -- that comes from the folks who designed the cover. The other errors are all on us, though, and include:
Now that we have made a mea culpa for our human weaknesses and enjoyed a round of self-flagellation, let's get on to some real questions.
Michael Thompson: Will you be publishing the projected DVOA by team for 2005 in your normal format (offense and defense, rushing and passing)? I didn't see it in the book, maybe I missed it because I haven't read the entire thing yet.
Yes, the week the season starts we'll run:
One of the biggest complaints we've gotten is that the book doesn't give a specific prediction of each team's won-loss record. Instead, we give each team's percentage chance of finishing in five different categories, which looks like this:
Baseball Prospectus, as many of you know, doesn't include any team predictions in their book. Football magazines always predict exact won-loss records for every team, which is a complete fool's errand given the way NFL teams unexpectedly improve and decline from year to year. We thought this was a good compromise that gave an indication of how likely it was that each team would improve or decline. Well, apparently, the readers disagree, so we'll probably stick specific win predictions in next year's book, with the proper language about how each predicted record is really the mean of a range of projections, not a specific exact projection.
Mark Padden: Any chance of posting your fantasy projections online for book buyers? Just providing one list as an appendix was frankly a bit less than I expected from the book. Thanks.
You aren't the first person to suggest that we offer fantasy projections in a few different formats in the appendix next year. It is definitely on the list of things to consider changing to improve the 2006 book. On the other hand, we don't want to run table after table after table that just represent different league formats and increase the cost of the book to the consumer. (Some people have complained that we don't have tables of the best fantasy players by position, but I figured one big table took care of that -- if you need a running back, just look down until you hit the next listed running back -- while conserving space.)
Baseball Prospectus actually doesn't include a big list of all fantasy players in their book. What they do is offer that as a downloadable spreadsheet as part of a premium subscription package. So we were stuck trying to figure out what to do -- we don't have a premium subscription package. We knew people would want the projections in a list, preferably a spreadsheet they could play with. We can't really post the projections for book buyers because people who bought the book online would have to forward us the receipts, and that would still leave people who buy the book in actual stores out in the cold. What we decided on, as a one-year solution, was to offer the projections with a small $10 donation -- pretty cheap compared to most downloadable projection sheets at other fantasy sites, I think -- and let people know that we were trusting them to buy the book as well. We hope to have something better next year, and we hope people understand that this is our first time doing a book of this type, size, and magnitude so not quite everything was figured out perfectly.
Dan Babbitt: Will you all be issuing revised projections for Travis Henry and Chris Brown? Also, are you aware of anyone doing the type of work that you or Baseball Prospectus does, but with the NBA?
Yes, the spreadsheet we're sending out with a donation does include revised projections for Henry and Brown and a few other players. Lest anyone think we want them to send us $10 just for a handful of new projections, we present the most important changes here. They include the Henry and Gardner trades, Johnnie Morton signing in San Francisco, and changes in the expected roles in Seattle and Atlanta.
|Dez White||Just say no.|
As for your second question, you betcha! Start with Roland Beech, the man behind the NFL site Two Minute Warning, who also has an NBA site called 82games.com. You also have John Hollinger -- unfortuntely, he is no longer associated with the Prospectus family, but he's still writing a great annual book (now called Pro Basketball Forecast) and he writes for ESPN Insider as well. And of course you have to check out the "APBRmetrics" forum and this central links site for basketball analysis, both run by the talented and very friendly Kevin Pelton.
Paul Barrow: Please could you advise the degree to which you cover individual defensive players in the Pro Football Prospectus.
Unfortunately, there are no projections for individual defensive players for fantasy football purposes, if that's what you are looking for. Individual players are discussed in the unit comments that are at the end of each team chapter. We provide individual defensive player stats that are a little different from the standard ones -- plays rather than just tackles, as well as stops (plays stopped short of success), defeats (turnovers, plays stopped on third down before converting, plays stopped for a loss of yards), stop rate (stops per play made), plus good ol' interceptions, fumble recoveries, and sacks. We also give the average yards gained on each play where a defensive player made a play. We're hoping to have these stats on the website for 2005 and then really work on improving our understanding of them for next year's book.
Craig Matthews: Just got the Prospectus in the mail last night and it's become one of those books that you read, put down... and then see how long you can go before that nagging curiosity gets you to pick it up again. I mean, honey, the kids' dinner can wait-- I've got to find out if Clinton Portis's DVOA improved when the Redskins changed their blocking scheme last year. And I mean now!
In looking over the projections, however, I did see a few head scratchers I was hoping you can comment on.
1) Trent Green and Brett Favre ranked 15th and 20th at QB? I understand Green has a tough schedule and Favre is getting older but are you telling me if your life depended on it you would want David Carr (steady but unspectacular, as you note on your site) or Chad Pennington (recovering from a torn rotator cuff) airing it out for your fantasy football team? Ben Roethlisberger and Rex Grossman ranked ahead of Favre? Let's hope you don't run into Peter King in a dark alley (behind a Starbucks of course).
2) Where's the love for Steve Smith (ranked 36), Lee Evans (39) and Donald Driver (51)? Smith is just a year removed from an 88 catch, 1,000 yard season, Evans out produced Moulds in each of the final 7 games of 2004 and Driver had 1,200 yards and 9 TDs last year.
Both Green and Favre may look a little low because what we have here is a mean projection. Both Green and Favre have established a high standard of performance. But the older the quarterback, the higher the chance of injury or rapid decline; therefore, older quarterbacks come out with lower mean projections.
Trent Green's projection is in large part due to projected decline by the Kansas City offense as a whole, due to age and increased chance of injury, combined with Green's age and Kansas City's difficult schedule. On the other hand, I do think that the projection system underestimated how many attempts he'll have this year -- only 486 -- which would mean underestimating all his numbers. (Green has thrown at least 520 passes in three of his four KC seasons.) I wouldn't be surprised if he beats the projection, but I would be surprised if he had a year as good as last year.
Favre's projection had similar issues, with two differences: the projection for the Green Bay offense as a whole is even lower, but the Green Bay schedule is not projected to be difficult. Favre actually projected with the lowest number of games of any starting quarterback other than McNair, but I couldn't bring myself to actually predict Brett Favre would miss a game so I increased his numbers to reflect 16 games of the same quality.
Also remember that there is a difference between the question "which quarterback would you rather have" and "which quarterback would rather have in fantasy football." Go check out Houston's schedule.
On to the receivers in question. Steve Smith's projection simply reflects that he's coming off an injury. Driver is a 30-year-old receiver coming off a career year in an offense projected to decline. (Driver's projection doesn't take Walker's holdout into account, of course.)
I like Evans a lot more than Moulds, too. There is one major subjective component of these projections and that is "expected role," and Moulds ends up with a higher projection than Evans because we projected him to be "1" and Evans to be "2". Evans has fewer projected yards because he has fewer projected catches. That is how Buffalo seems to view them. However, change the expected roles, and you change the projections as so:
|Moulds as 1, Evans as 2:
Moulds: 853 yards, 5.0 TD
Evans: 796 yards, 4.2 TD
| Both as 1.5:
Moulds: 769 yards, 4.4 TD
Evans: 880 yards, 4.8 TD
| Moulds as 2, Evans as 1:
Moulds: 684 yards, 3.8 TD
Evans: 965 yards, 5.4 TD
Yes, projected touchdowns are not whole numbers, reflecting the fact that they are mean projections and not absolute facts.
(Craig then responded with another interesting observation.)
Craig Matthews: It seems there is this rookie running back who struggled early on but then really hit his stride around Week 10 and ended up averaging over 100 yards per game over the last seven games, scoring touchdowns along the way as his team looked like it had finally found the franchise back to fit their system and lift their team out of mediocrity. Of course, most people will recognize this back as PFP-predicted 2005 rushing champ Kevin Jones. Fits the bill, but he's not who I am referring to. Julius Jones? Another good match, but no. To the horror of all of the fantasy football players planning on building their teams around one of the Joneses in the upcoming draft, the guy I am talking about is William Green.
In 2002, Cleveland drafted Green and critics hailed him as the back Butch Davis had been waiting for. After some early struggles he ran for 726 yards and 5 touchdowns in the last seven weeks and helped get the Browns to the playoffs. Of course, ever since life has been a horror show for Green and the Brownies.
Now, as someone who drafted Mr. Green in the second round in 2003 it is painful for me to even type his name. In 2005, I have been once again looking at the previous year's rookies to build my team around and the two Jones's seem to be the consensus picks to have breakout seasons. What I am looking for is for the Football Outsiders to convince me (please!) that Kevin Jones and Julius Jones are not William Green.
Best arguments I can make are:
1) The character issue. Unless Ricky Williams gets ahold of the Joneses, I think they have a better chance of staying on the straight and narrow.
2) The coaching staff. I'll take a couple of Super Bowl winners (even if they have horrible nicknames like Mooch and the Tuna) over Butcher Davis any day.
What do you say?
There is one number missing from your analysis, and that is carries. Here is an analysis of all three backs split into Weeks 1-10, and Weeks 11-17:
|William Green 2002||Runs||Yards||Avg||DVOA||DPAR|
|Kevin Jones 2004||Runs||Yards||Avg||DVOA||DPAR|
|Julius Jones 2004||Runs||Yards||Avg||DVOA||DPAR|
Kevin Jones's second half explosion went far beyond those of Julius Jones and William Green, so I don't think Kevin Jones and Green are comparable. Julius, however, is comparable. The biggest differences there are:
1) Coaching Staff, as you noted
2) Ganja, as you noted
3) Julius Jones's record only includes the good games at the end of the year. Green had both the bad first half and good second half that could be used to judge him. Green went from bad to good in 2002. Both Joneses went from INJURED to good in 2004.
Also remember, Green wasn't great in yards per carry the following year, but his biggest problem was that he played only seven games, not that he didn't produce when he was playing.
Paul Tarr: I purchased Pro Football Prospectus 2005. My specific reason was to read your updated analysis of the greatest quarterback season ever.
I am a member of SABR and an avid football fan. In general, I was particularly impressed with your analysis at the play-by-play level. You are a step ahead of baseball in this regard. Need to study your techniques more.
I was generally very impressed with your updated analysis on the greatest football season ever:
That being said I don't agree with your rationale for excluding Otto Graham from your ratings. 1953 was the 8th year in pro football for Graham under his coach, Paul Brown. It was also his 4th in the NFL. Brown's procedures and innovations were no longer new. The original article "It's Otto-Matic: Manning's No. 1" showed great analytical insight on your part in comparing Graham to Ruth. In 1920, Ruth hit 54 home runs. The second highest total in the major leagues was 19. Ruth's slugging percentage was .847 compared to .632 for the second highest in the major leagues. Talk about domination. Ruth, like Graham, played a different game, based on a livelier baseball. Yet Ruth was never excluded from comparisons over time. His dominance was recognized as a level to be achieved and superceded.
Under these circumstances I would love to find out what Graham's PAA was for 1953 based on your procedure. It would seem that you have done enough on Graham's 2004 projections to compute an estimate, even with some limitations. It hasn't been published in your book but I think the data is relevant.
I have to admit that I struggled mightily with trying to figure out where to make the cutoff for the Best Quarterback Seasons article. I always talk about Otto Graham when I am asked about great quarterbacks, and I did want to make sure that I mentioned his great 1953 season. Your reasons for including Graham are certainly valid. The main reason why I felt uncomfortable including him was the low passing touchdown total, so dramatically different from quarterbacks of the past 45 years. Even after adjusting for a 16-game season and the modern offensive environment, Graham would come out with just 16 touchdowns passing, and no matter how much I try to adjust one NFL era to fit another, that doesn't seem like the top quarterback based on what the term "quarterback" means today.
To answer your question, however, Graham works out to 186.3 Points Above Average. (The top quarterback of the past 45 seasons, according to the article in the book, is Bert Jones in 1976, with 163.8 PAA).
As long as we're mentioning Graham, however, we should also mention two other historical quarterbacks with astonishing performances in the early days of professional football. Sammy Baugh's 1945 season, projected to the 2004 offensive environment, is worth 228.6 PAA (passing data only). That includes a translated 92.5% completion percentage and a 27-to-4 TD/INT ratio. Sid Luckman's 1943 season, projected to the 2004 offensive environment, is worth an absurd 268.1 PAA (passing data only) including 6103 yards, or 11.7 yards per pass attempt. Plus, ya know, Sid's one of my peeps.
So, I guess the issue is where to draw the line. Depending on where you draw the line, the best quarterback season of all time belongs to Bert Jones, or Otto Graham, or Sid Luckman. All great quarterbacks.
Reinhard: The top running backs in DVOA last year were LaMont Jordan, Tatum Bell, Dorsey Levens, Larry Johnson, Derrick Blaylock, and T. J. Duckett. Doesn't that look strange, that none of the top 6 DVOA were considered starters really, just good backups? Does their reputation help their performance? Do they only get called on the field in favorable situations? (Teams against which there have been a lot of run plays, tiring out the starter. Or many runs in a row, tiring out the starter. Or when grinding out the clock? Many of the players saw whole games at a time at starter.)
A perfect example from last year would be LaDainian Tomlinson, -5.0% DVOA, compared to Jesse Chatman and Michael Turner, who put up about 18% DVOA between them. Any thoughts on what's going on here?
I think the reason for the phenomenon is pretty simple: sample size. The fewer attempts by a player, the wider the range of DVOA. Here are the five worst backs last year with at least 75 carries, according to DVOA: Quentin Griffin, Anthony Thomas, Maurice Hicks, William Green, and Eddie George. Green with 162 carries is the highest. The wide receivers are an even better example: The DVOA top five includes Brandon Stokely (102 passes), Lee Evans (75 passes), and Plaxico Burress (60 passes) while the bottom five includes five guys with 86 passes or fewer.
This is why I set a minimum for the DVOA/DPAR rankings. I set the minimum to try to include roughly three receivers per team and, for RB, to include both backs when a team has a committee. But the lower I set the minimum, the more of these very high and very low DVOA ratings will make it into the main table for each position. It's basically a subjective decision how many players to rank, and I have no problem with people who refer to our numbers making their own decisions about minimums. (For example, you could write: "According to FO, Curtis Martin has the highest DVOA of any running back with at least 200 carries.")
The other issues you brought up are hopefully already corrected in the DVOA formulas, except for "backup comes in after starter has tired down defense," that one isn't in there at this point. As for Tomlinson...
Tito: I have been doing Fantasy Football since I was 12 years old, having won two Blauvelt Bowls (II and VI). I am also an avid baseball statistician in my free time. Despite this excellent background, I do not understand why LaDainian Tomlinson's DPAR is as low as it is. Is it simply due to his sub-4.0 YPC average? His five fumbles? Enlighten me.
I assume you are talking about 2004, as Tomlinson was second in DPAR for 2003 and sixth for 2002. Tomlinson's rating by our metrics went down last year for the same reason that his conventional ratings (except for TDs) went down last year. He was hurt and didn't play up to his usual standard. It is pretty clear that he was better by mid-season. Check out Tomlinson's regular season numbers before and after San Diego's Week 10 bye:
|LaDainian Tomlinson 2004||Yards/G||Yd/Carry||DPAR||DVOA||Suc. Rate|
Yep, he's fine. There are a lot of places where our projections for next year disagree with conventional wisdom, but we have Tomlinson rated the #1 fantasy football player just like everyone else.
Finally, here are two emails I received that have nothing to do with the book. They are requests for information that we don't have, so I figured I would toss them out to the readership to see if anyone can help.
John Grenci: hello, I was wondering if you had the pre-season Las Vegas lines for seasonal wins in past seasons?
Norman Marty: Dear Aaron, we are trying to find a stat that nobody seems to know: How many times on average per game are the ten yard marker chains brought onto the field during a game, and how long does it take to bring them out?
If you know the answer to either of these questions, post it in the discussion. Thanks!
102 comments, Last at 25 Dec 2012, 5:05pm by Geschenke