FO on FOX: Guts and Stomps

Like you said, teams have to PLAY bad teams in order to STOMP bad teams. I'd be interested to see STOMP%, or the percent of games against under-.500 opponants that ended in a STOMP win. Likewise for the other categories.

If it turns out that STOMP wins has a higher correlation with postseason success than STOMP%, then you're drawing the wrong conclusion from the stats. It's not STOMPING bad teams that makes a team more likely to succeed in the postseason, it's FACING bad teams that makes a team more likely to succeed in a postseason (starters are fresher, fewer injuries, etc).

For instance, you mentioned the 2003 Patriots several times as a team that won the SB despite a surfeit of STOMP wins. However, if the 2003 Patriots only played 1/3rd the league average number of games against sub-.500 clubs, that might explain why they had so few STOMPs.

Well this was just about the best single piece of football related writing I have seen in my entire life. It tells us something interesting, bursts some well loved myths, and its concise and to the point. More importantly it has data to back up its interesting claims (who would have thought...).

I am serious this is a great article IMO, and the best thing I have read on FO thusfar (and I really enjoy the site).

Of course, part of my love for thie artcile may stem from its agreement with my longstanding opinion that most games within 7 pts are more or less statistically speaking a tie (or at least tell us little to nothing about which team is better).

#2 -

It is, of course, something that stat-minded baseball fans know well. Neyer and Epstein wrote a book on it, in fact. "Winning the close ones" is a BS dump, like chemistry, character, "clutch", etc., that "analysts" like to use when the results don't agree with their guesses (hypotheses is much to strong a word...). Great teams have great numbers kicking the crap out of bad teams, not squeeking past anybody.

Related, and simple example: Hitter one has better numbers, in a tougher ballpark and plays a moderate defensive position brilliantly, while going through a much tougher season of media scrutiny. But you like hitter #2, who is also a very good hitter, a likable teddy-bear of a guy and a good quote and story, but who doesn't play defense at all. How to make your case for voting for him? A-ha! He's much more "clutch"! Remarkably, the "right" guy won the award this year in the AL (and I hate the Yankees, so that's tough to admit...), so maybe we are making a little bit of progress after 100 years of MVP voting. This is something that has appeared throughout the history of statistics (definitely not just sports history).

-Sean

I think there are two kinds of stomp teams. The kind where they get lucky breaks and where the luck eventually catches up to them and they lose, and the kind that are good at exploiting matchups. Obviously most stomps would be the former. But I thought the pats advantage was always their superior coaching and matchup play and preparation. I think it would be hard to tell the difference between one and the other just by analyzing win/loss and point differential.

I think that means I agree with this article, but I'm not sure about that. heh. But I do believe that intangibles like momentum and confidence really do exist and have effects on game outcomes. It's like willpower - you can control it in the short term even if it isn't sustainable.

Is there any correlation between when the stomps and/or quality wins occur during a season, and postseason success? I remember the Ravens' wins kept getting bigger and bigger as the season went on, so that the Tony Banks-era, field-goal squeaks were mere memories by the time the playoffs arrived.

These stats do highlight just how good were those 2003-04 Patriots. Who are 2005's leading stompers at this point?

Great article Aaron. One question, though: what about total wins? I'm curious how that compares to the other categories you carved out. Thanks.

#6 - I meant in terms of winning record in super bowls and conference championships.

Post 3:

Well, to give credit where it's due Neyer/Epstein were merely following through on the work of Bill James.

And very interesting article.

Thank you.

the correlation coefficients are nice, but i'd like to see how those categories correlate with postseason performance, rather than regular season winning%. after all, each of those categories is actually included in regular season winning% (especially the last two). the correlation between two overlapping sets of data doesn't actually have that much to tell us about cause and effect. but correlation between regular season stomps/guts/etc vs playoff performance would be really interesting--in some sense, the postseason is a "testing ground" for theories about how a championship team performs in the regular season. most importantly, it's an independent data set.

To look at the AFC so far:

Ind: 6 Stomps; 2 Guts; 2 Doms; 1 Skate
Den: 3 Stomps; 3 Guts; 1 Dom; 1 Skate
SD: 1 Stomp; 1 Gut; 2 Doms; 2 Skate
Jax: 1 Stomp; 2 Guts; 0 Dom; 4 Skate
Pit: 2 Stomp; 1 Gut; 1 Dom; 1 Skate
Cin: 2 Stomp; 0 Gut; 2 Dom; 3 Skate
NE: 0 Stomp; 2 Gut; 0 Dom; 3 Skate
KC: 2 Stomp; 0 Gut; 0 Dom; 3 Skate

So in the AFC, Ind and Den are the clear favorites, and NE and Jax don't look so hot - nothing surprising.

Aaron, this is a nice article, but you left out any analysis of SWAGGER. Clearly, all of those Super Bowl winners had a lot of swagger, and I'm sure you'll find the correlation coefficient between swagger and Super Bowl wins quite high.

(But seriously, great article.)

#2 Perhaps we've seen progress in the MVP voting, but the Cy Young voting is apparently still in the Stone Age.

I'd really be interested if DVOA was used instead of won-loss records to determine who the better team is. A team could get to +.500 by playing a lot of bad teams, then when a team DOMINATES them it looks good but in reality it wasn't the same quality. It also applies to teams with sub .500 records especially since those teams had that extra loss by the team that beat them.

Maybe use positive overall DVOA instead of +.500 and negative overall DVOA for sub .500? The piece looks like it's on the right track but it just seems that something is missing.

Great article Aaron, thank you.

In response to #4 and #11, regarding intangibles such as confidence and momentum.

I know this is a bit off topic, but related none the less. I have a friend who is a professional online poak-er player (don't want my comment held up as spam). As such, he keeps a huge database of statistics on his play to help maximize his winning percentage. Over the years he's come to realize that confidence, swagger, momentum, whatever you want to call it, has the biggest outcome on whether its a winning session or a losing one.

If he sits down at a table and immediately starts winning, he stays. If not, he gets up and moves on to another table, where he is a new face and thus back on even keel. If you sit down and start losing, everyone looks at you as beatable and vulnerable. If you sit down and start winning, everyone shakes in their boots at the prospect of getting involved in a hand with you.

Interesting lesson in psychology, but I'm sure this holds true on the football field as well, at least to a certain degree.

Good Stuff.

I'm just confused - how does this correlate with respect? Because I'm pretty sure there's a very strong inverse correlation.

Great article,

But here's a question: A good team that, by the vagaries of the schedule, happens to play a lot of bad teams will end up with a lot of STOMPS. Hence it will tend to end the season with very good record and will get a 1st round bye. We already know that getting a 1st round bye is one of the biggest determining factors that decides which team goes to the Superbowl, so is it any wonder that there is a high correlation between STOMP wins and SB winners?

I always believed that getting the first round bye doesn't make a huge difference in making it to or winning the Super Bowl. I believe that the very good teams get the first round bye, and that is why they get to or win the Super Bowl. If you put Denver and the Colts in at wildcards, wouldn't they probably meet in the AFCC?

A few years ago, I saw Dr. Z quote a stat on the '70's steelers. I cannot remember the EXACT quote, so I apologize, but the gist of it was those Steelers went like 75-1 against teams that ended up below .500. The message was that you HAVE to beat the teams you're supposed to beat. I've long believed that was a very important key to being a great team...your correlation confirmed that.

Re #18: ITA. Teams with first round byes go to the Super Bowl because they are better teams, not because they had a week off.

I've decided to use this article as part of my argument for why the 1998 Tulane football team should have gone to a BCS bowl.

Thank you.

Some quick research indicates that over the past 10 seasons:

16 of 20 Super Bowl teams have had a bye
8 of 10 Super Bowl winners had a bye

Clearly its important to get that bye, nobody needs to do research to know that. What isn't clear is whether they fair better because of their week off or because of their superior talent. I'm guessing a little of both.

On the other hand, only one of the four teams that get a bye win the superbowl. Look at it this way. Points differential (points scored - points allowed) coorelates better with superbowl victories than having the best record does. That would indicate that stomping opposing teams is more important than just beating them.

#19 Looking at www.profootball-reference.com, I was able to verify the statistic. Just hunting and pecking, the Steelers from 1972-1979 went 49-1 against teams with losing records (they did lose some games to .500 teams). The only loss was in 1979 to the Cincinnati Bengals. I find that stat utterly amazing.

Re #22: What's important is to be the better team.

The teams with the byes also play at home. So for a non-bye team to make the Super Bowl, they (almost always) have to win two games on the road against better teams. I think the off-week is fairly irrelevant.

Re: #12

This article was in part researching/defending the legitimacy of DVOA, since it automatically credits STOMPS more than GUTS. So to compare them to DVOA would wind up being kind of circular.

(That very question was asked by one of the Outsiders when Aaron first showed us the numbers.)

Also, it was supposed to be very acccessible to the wider audience. to specifically counter a complaint related to wins, the idea was to present the research solely in terms of wins.

Re: importance of bye weeks in the playoffs

What we need to do is to get the NFL to change the playoff format for the next decade or so, giving the byes to the wild card teams. Then we can compare that decade to the previous one, and see if it's the byes or being the better team that is important.

Aaron et al, you work for Fox Sports. I'm sure they have the connections to get this done. Let's see this happen.

TEAM: DOM-STOMP-GUT-SKATE
-------------------------
NYG: 0-5-1-1
SEA: 0-4-3-2
CAR: 2-1-1-4
CHI:   0-2-2-4
ATL:  1-1-0-5
DAL:  0-2-2-3
TB:    0-1-1-5
MIN:  0-1-1-4
-------------------------
Favorites: NYG, SEA
Contenders: CAR, CHI, ATL, DAL
Pretenders: TB, MIN

At least the pretenders have new QB's to give them a sliver of hope.

"The off week is fairly irrelevant"

I strongly disagree. It may be irrelevant with respect to how much of a benefit a week's rest gives (although I'm not sold on that). But the key is that, in a 1 and done format, you have to win one less game than teams that don't have a bye.

Even if you're a fantastic team and have a 70% chance of winning any game you play (which I think is a little high, given the caliber of teams that generally make the playoffs), and even if you neglect the advantages of having home field in at least one game and the advantage of a week's rest, then your odds of going to the SB are .7*.7 = 0.49 if you have a bye, and .7*.7*.7 = 0.343 if you don't. That's a 1 in 2 chance versus a 1 in 3 chance, which is very significant.

Some more accurate numbers, taking into account that opponents faced tend to be better later in the playoffs, would probably be:
No bye: .8*.7*.6 = 0.336
Bye: 0.7*0.6 = 0.42
which is still very significant.

Re: 26/27

Would it be a lot of work to run it again with DVOA?

I was trying to compile teams' 2005 records against winning and nonwinning opponents a couple days ago, and I noticed that I kept mistakenly giving a "win over winning team" credit to teams that had beaten Washington. DVOA is taking over my brain.

Oh, and excellent article. I'd seen similar stuff for baseball, but it wasn't at all obvious to me that the same would be true for football.

#14 Or, it could be that winning [the game you mentioned] is somewhat dependent on having an early lead. If only there were a statistic that gave a boost to teams that did well in first quarter offense...paging Forrest Gregg!

Re 14:

The're a fundamental difference between that card game and sports in that regard. In that card game, you win by convincing your opponent to make bad choices from incomplete information; intangibles make up much of that incomplete information.

Also, quickly leaving a table you're losing at has a direct benefit: if that brief session is statistically accurate, you're avoiding future losses. Not all tables are created equal, after all. (If only a football team could call a Mulligan after the first quarter and change opponents :)

Re: the bye week. This is one of TMQ's favourite topics. Teams that had a bye (ie, home teams) are 49-11 (.817) in the divisional playoff round. But home teams are only 17-13 (.567) in the next round, the championship game.

Given that the home team almost always has a better record than the visitor, you'd expect the home teams' winning percentage to be a little higher than in the regular season (about .600). The divisional round home winning percentage is significantly higher, however - which is difficult to attribute this to anything other than the bye.

Great article. I haven't yet had time to dig through the tables, but I'd be interested in the correlation between average margin of victory in Nomember and December, against all teams, and post-season success. When I think of the the truly great teams, I remember them bludgeoning everybody like the proverbial rented mule, right through the playoffs. As good as the Pats have been in this most recent era, they have not absolutely crushed the competition, as some other previous champs have done.

They only won one Super Bowl, but I still think of the '85 Bears as the standard against which all others are measured. None of their playoff games was even within shouting distance of being a contest. Total and complete domination, no matter the opponent.

But home teams are only 17-13 (.567) in the next round, the championship game.

I'm sure a large part of that is Bill Cowher going 1-4 in AFC Championship games at home...

Great article. I had the same question as Kibbles in #1 on %. So, I ran the percentages. Also, in regard to #17, because I was curious as to effect of home field advantage on the numbers, I separated into championship games where the home team had the higher percentage in a category, versus where the home team did not. The results are even stronger in support of Aaron's hypothesis.

I used same categories. The "Domination" percentage is total number of Dominations divided by Total number of games played against opponents over .500. Others are similarly calculated. I did not count a game where two teams were within 5% of each other in a category (2 of 9 vs 2 of 10 in "stomps", for example), so the records don't add to 20 games. Here are results:

Championship Game
Home Team w/ Higher "Domination" %: 7-2
Home Team w/ Lower "Domination" %: 1-5

Home Team w/ Higher "Guts" %: 3-3
Home Team w/ Lower "Guts" %: 6-2

Home Team w/ Higher "Stomp" %: 10-3
Home Team w/ Lower "Stomp" %: 1-4

Home Team w/ Higher "Skates" %: 1-6
Home Team w/ Lower "Skates" %: 5-3

Those are some resounding numbers. While it is true that most home teams in the Championship games are going to have a higher % of stomps and dominations than the opponent, it is not always true. And in those cases, the road team generally wins. Conversely, when the home team has the higher percentage of skate games (close wins over .500 or less teams), it was 1-6 in the title game.

As always, great and interesting article; love FO. However, several things come to mind. What are that stats of STOMP+DOM (i.e. all wins over 14 pts), wouldn't that lessen the effect of playing no bad teams? Of course, that brings up the point of #1, maybe you want to face bad teams.

RE 35, I wouldn't expect the better teams winning percentage (w%) in the playoff games to exceed their w% in the regular season. They are playing better teams than the avg team they faced in reg season. Thus, just being over .5 means something, like the better team usually wins. There are must be ways of better deciphering the stats (like they do with twin studies to tease out environ v. genetic factors). Somehow I don't see the NFL changing how it runs things merely to satisfy our statistical curiosity though. However, it looks like we've got people willing to tease out even minutae from the data available. One just has to be careful since some statistical results are counter-intuitive and your reasoning will lead you to the wrong conclusion.

My last question has to do with the correlation stats. They're against regular season wins (not independent, as pointed out in #9). What are the correlations to post season win %? Otherwise, qual wins (and even better other wins) is the best predictor (of reg season wins). However, what one really wants is correlations to post season wins (or better SB wins), since that's the goal.

confirms my long term suspision(sic) that Minnesota and Denver in 1998 would have been one of the best superbowls ever. Instead Atl earned another "skate" and promptly got demolished.

Re #34: The’re a fundamental difference between that card game and sports in that regard. In that card game, you win by convincing your opponent to make bad choices from incomplete information; intangibles make up much of that incomplete information.

Oh? And in football, you don't win by convincing your opponent to make bad choices from incomplete information? What's a Draw Play, then? Theoretically, the draw play should be the easiest play on earth to defend, since it's so slow developing and the linemen aren't run blocking. In practice, the draw play is often deadly, since the defense makes a bad choice (defend the pass) based on incomplete information (linemen are pass-blocking, QB dropped back and held the ball).

I think football is very much a game of trickery, of one team trying to use incomplete information to guess what the other team is doing, and the other team trying its hardest to vary up their patterns and keep themselves unpredictable.

I also think that a lack of confidence can have a very big impact on winning and losing. If a defense is normally stout against the run, and they start giving up rushing yardage in big chunks, they're more likely to lose confidence in their run defense and start stacking the box, which opens things up against the pass, etc. I think the Denver/KC game from earlier in the season was a perfect example of a team losing confidence and getting hammered. Denver wasn't doing anything special, but they were moving the ball at will doing it, because KC (which had been decreed a "stout" defense before that matchup) just couldn't figure out what was going on. They were shell-shocked.

Another great example was the Pitt/Indy game last monday night. Cowher opens up the second half with an onsides kick. The only explanation I can think for why a normally conservative coach would do such a risky play is that he felt like he needed a lot of help to pull off the win- quite an odd position since Indy was only up by 9 points at the time. He just didn't have confidence in his team, whether it was because they were trailing or because Indy was undefeated, and as a result, he gave up an easy 7 points to Indy.

Falco #38:

GREAT comment, great additional stats. The results could not be more clear. Great teams beat their opponents by a lot. Close games are a sign of weakness.

Besides N.E 2003(4), the most gut wins of any superbowl winner has been 2.

Now that we have this information, we can filter it for this year. For instance, we'll know better than to count against Indy if they get some guts or skates from resting starters. We would also know better than to give any doms should they lose badly. Knowing circumstances will help us a ton.

don't mean to be redundant, but great piece! its also a great piece to introduce to people who are skeptical of the type of football stats found on this site. but to reiterate what a couple people said earlier, it left me thirsting for a DVOA version. i feel as though a stomp can too easily be turned into skate with a meaningless last second touchdown, or vice versa. here is a memory: BAL vs CLE last year, CLE down by 7, drives down inside BAL 10yd line with under 2 minutes to go. Ed Reed then picks a pass off and a kneel down would end the game, but he takes it 100yds to the house! a fond memory for me, but that was no stomp.

Okay, Aaron, now that you have just proven that STOMPS are more important to determining championship caliber than GUT or SKATE wins, I believe a little rewriting of history is in order (along with massive doses of memory-erasing drugs). As a result, my Colts are better than your Pats and have been for some time. So... anyhoo, please return the Lombardi's now, and I won't have to send you to a time-out corner or send a note home to your mom. Please.... Hello? Aaron, you there? Hmmph!

Okay, so having shown that the Pats of the past few years are pretty much outliers, how the hell did they do it? Can it simply be timing (peaking at the right time)? It's too easy an answer and impossible to measure (though it probably works for the 2000 Ravens as well).

bobman -

I would guess the Pats phenomenon is a combination of a few things:

1) Being good at the things (low DVOA variance, long kicks, 4th quarter offense) that tend to be more of a factor in close games than in blowouts.

2) Pre-2004, having a great defense and a merely good offense, which tends to create low-scoring games.

3) Just dumb luck/statistical variance.

It seems that in 2003 no team dominated. It was a skater's paradise. All 4 teams who made it to the championship round were vulnerable with at leat 4 skates a piece.

Re: #40

you misunderstood. .600 is the approximate winning percentage of the home team in all regular-season games. you would think that this percentage would be even higher when the home team is also better than their opponent, which they are in the postseason (usually). but in CCGs the home team wins less than 60% of the time, despite being at home and (usually) being better. thus HFA means less in the postseason than in the regular season. that's step 1. step 2 is to go from there and look at the second round playoff games. there, the winning percentage of the home team is over .800, despite the fact that we have shown that HFA itself means less in the playoffs. the conclusion: the bye week itself matters a lot.

now, it matters even more than this because teams with a bye don't have to win as many games, as #30 pointed out. of course, this analysis may be somewhat suspect; after all, the skill differences between the teams are higher in the divisional round (1 seed playing 4/5/6, 2 seed playing 3/4/5) than in the CCGs (usually 1 vs 2, given the dominant record of higher seeds in the divisional round). also, just because HFA means less in the CCGs doesn't mean it means less in the entire postseason. nevertheless, that is the gist of TMQ's argument.

ammek #35:

Teams that had a bye (ie, home teams) are 49-11 (.817) in the divisional playoff round. But home teams are only 17-13 (.567) in the next round, the championship game.

Since 1990, teams with a bye not named the Kansas City Chiefs in the Divisional Playoff: 49-8 (86%). Chiefs: 0-3.

Since 1990, home teams not named the Pittsburgh Steelers, San Francisco 49ers, and Philadelphia Eagles in the Championship Playoffs: 14-4 (77.8%). Steelers (20%), 49ers (25%), and Eagles (33%): 3-9 (25% combined).

To understand the 2003 & 2004 Pats "Stomp" and "Guts" factors, you have to look at a few things. First, the 2003 team had a great defense, a slightly above-average passing game and a horrid running game. That is the main reason that team had so many "skate" or "gut" wins, much more than would be expected of a superbowl champion. Also, the 2003 & 2004 teams had a tendancy to play down to thier opponents. Lastly, in those two years, they faced 15 quality opponents. Most teams that face that many quality opponents in a year (7+) tend to falter in the post-season, possibly because they get worn down by facing a tough opponent week after week. The Pats, for whatever reason, never hit that wall, well untill this year.

B,

Have to point out that 2003 was the only year where all 4 sb contenders had alot of skates relative to the other year. Another conclusion might be that NE won in a year of stastical outliers. Whoever won that year would have been the exception and not the rule.

I added guts + skates and found that only two years(2001, 2004) did the team with least (guts +skates) not win the SB.

I see a potential problem with continuity of data. A potential problem could be the way the league was realigned 4/5 years ago. Alot of the outliers have occured in the 2000's. It could be the desired parity or it could be just a couple of outliers bunched together.

It'd be nice to figure a couple of rules. I'm looking at basketball right now, to see if the findings hold true in other sports

Hey Aaron, I've got a stats question for ya. Can we safely assume that you check for autocorrelation, heteroskedasticity, and multicolinearity?

Any win-loss measurement in small numbers is mostly homoskedastic: the error primarily comes from the variable quantization.

There's clearly an autocorrelation ("Strangely, all of my red socks are red!") component on the four categories listed above, because they're measuring wins versus win percentage. It's not goiong to be a big deal because each of the components is essentially affected in the same way.

The four categories are mutually independent, so there's no a priori multicolinearity. Any Given Sunday, and all that. (Whether some of those categories are practically colinear is another thing).

This ends your yearly quota of random large statistics words that no one outside of finance knows about. Thank you! :)

Incidentally, I know I'm now over the quota, but: the first point (autocorrelation) is actually the interesting one. Basically, if you group wins of any sort and check for a correlation versus winning percentage, you'll find a correlation (this is autocorrelation).

In other words, if you plotted, say, "Wins versus the San Francisco 49ers" vs "Winning Percentage", you'd probably find out that teams that beat the 49ers 0 times had a lower winning percentage (on average) than teams that beat the 49ers 2 times. This is just because teams that beat the 49ers twice have a minimum winning percentage of 0.125, whereas teams that didn't beat the 49ers at all have a maximum winning percentage of 0.875. Poof, autocorrelation.

Then again, over the past two years, if you couldn't beat the 49ers, maybe you are a really bad team.

RE 51, You're right I did mis-read.
Let me understand, because I find the foolish game titles more confusing than useful.

The CCG is the final pre-SB game and that is played between two teams with equal rest, .567 w% for HFA "better" team, worse than that's team w% for the regular season, as to be expected, since the winning team is beating a better than avg opponent in the CCG, but not as easily as it did in the reg season.

However, in the DPG, the HFA (better and with a bye) team wins @ .8, which is significantly higher than the .6 reg season avg, where I would have expected the number to be lower than .6 (not factoring in the bye effect).

How about the stats for the "Wildcard playoff game", where again neither team has had a bye, are those HFA's
It would also be interesting to chart regular season games to see if the bye effect holds there also (v. the teams non-bye performance). If the "bye effect" is real, it should be verifiable against other comparisons (i.e. the WPG should be like the CCG and the bye effect should also appear in reg season games also).

Teams that had a bye (ie, home teams) are 49-11 (.817) in the divisional playoff round. But home teams are only 17-13 (.567) in the next round, the championship game.

Of course the caliber of the opposition is better in the championship game so that has quite a bit to do with it as well.

The bye effect wouldn't have to appear in regular season games. If its main benefit is to allow injured players to heal, it might not be present in the regular season games as the byes are all early in the season, and injuries accumulate.

Excellent. This is the kind of stuff that makes me continue to frequent this site.

Re #56:
That reminds me: I need to get my own multi-colinearity checked soon.

Does anyone else think that data on how teams coming off the bye week do in the regular season might be relevant to this discussion?

chris clark #59:

The Wild Card Game was instituted in 1978, when the 16 game schedule was adopted, and passing rules liberalized to open up the game. A second Wild Card Game was added in 1990 to allow a 6th seed into the playoffs.

Here are the results for the home teams from 1978 to 1989 and from 1990 to 2004. I'm seperating them because the results vary wildly due to the introduction of a 6th team in the playoffs and the removal of the bye for the 3rd seed, but the granting them a home wild card game. These figures exclude the strike season of 1982.

Wild Card Game
78-89: 13-9, 59%
90-04: 42-18, 70%

Divisional Game
78-89: 30-14, 68%
90-04: 49-11, 82%

Championship Game
78-89: 16-6, 73%
90-04: 17-13, 57%

The new playoff format with 6 seedings appears to make it more likely that the #2 seed survives the divisional round, since the #3 seed is now tired out by a wild card game and no longer has a bye, resulting in increased competition against the #1 seed in the championship round, since it is usually #1 vs. #2.

What is especially interesting is the split in results between conferences since 1990. In the NFC, for the home team:

Wild Card 3 vs. 6: 7-8
Wild Card 4 vs. 5: 11-4
Divisional 1 vs. 4/5/6: 15-0
Divisional 2 vs. 3/4/5: 12-3
Championship 1 vs. 2/3: 9-6

In the AFC, for the home team:

Wild Card 3 vs. 6: 14-1
Wild Card 4 vs. 5: 10-5
Divisional 1 vs. 4/5/6: 10-5
Divisional 2 vs. 3/4/5: 12-3
Championship 1/2 vs. 2/3/4/5: 8-7

The difference that really jumps out is the propensity of the #3 NFC seed to lose, and the propensity of the #1 AFC seed to lose. Since 1990, 9 NFC #1 seeds have gone to the Super Bowl vs. just 6 AFC #1 seeds.

On the bye week thing, we could look at the records of teams coming off byes and getting to play at home versus teams who played previous weekend. Now, this would include games that would not be like the playoff scenario, where the worst team was also playing at home.

Here are 2005 (limited sample size) results: 8-3 (73%), with week of game in parentheses.

BALTIMORE 13, NY Jets 3 (4)
WASHINGTON 20, Seattle 17 OT (4)
KANSAS CITY 28, Washington 21 (6)
San Diego 27, OAKLAND 14 (6)
ARIZONA 20, Tennessee 10 (7)
PHILADELPHIA 20, San Diego 17 (7)
CAROLINA 38, Minnesota 13 (8)
NEW ENGLAND 21, Buffalo 16 (8)
San Diego 31, NY JETS 26 (9)
BUFFALO 14, Kansas City 3 (10)
SAN DIEGO 48, Buffalo 10 (11)
Indianapolis 45, CINCINNATI 37 (11)

(I excluded CLEVELAND vs Chicago in week 5, because both were coming off bye, but the home team, Cleveland won that game in what now appears to be an upset).

Only 2 road teams won in these situations, and they happen to be rated top 5 in DVOA (San Diego and Indianapolis). In those 3 wins, the home team was the underdog in all 3, which rarely if never happens in the divisional playoff round. The 4 situations that I subjectively believe would be most similar to divisional playoff round discrepancy between home and away, with home as above .500 team and favorite, are Carolina-Minnesota and San Diego-Buffalo, Kansas City-Washington, and New England-Buffalo, and the results are in line with the Divisional Playoff Round effect. It merits further investigation of earlier years to see if this holds up. I suspect there is some evidence there to mirror playoff percentages with increased sample size.

Why would the NFL schedule the vast majority of bye teams on the road the next week otherwise, than to mask this effect? And who did San Diego piss off in the off season?

Re #48 and related posts.

Actually, the Patriots generally prove the rule, rather than being the exception. Pats were the best in "DOMINATE" percentage (4/8), and were one of the road teams to win championship game, with higher DOMINATE and STOMP and lower SKATE percentage than Steelers. In 2003, the numbers favored the Pats slightly, and not the Colts. (The Colts actually SKATED at a higher percentage). The lone exception is 2001 in Superbowl (but not at Pittsburgh, where numbers pointed to NE).

I think the perception may be due to the GUTS wins. The GUTS category, which the Pats did well in, is not in and of itself a negative. (In other words, Seattle is not automatically going to lose to NY Giants in postseason merely because they beat them in close game in regular season) It just has little correlation either way with whether answers the "gut check" correctly when the game is again on the line in the playoffs in the 4th. The other 3 categories seem far more important. DOMINATES tells you that a team is good enough to win comfortably against good opponents, a sign of dominance. STOMPS tells you that a team tends to take care of business against weaker opponents, and SKATES tends to tell you the opposite, that a team struggles but managed to beat weak opponents. It is no wonder that a high SKATES percentage correlates with Championship game losses, and represents a home team who built a better record on close victories over bad teams, and thus likely is not as good as the visiting opponent.

I think the same thing applies to college basketball.

Even though it's early in the year, you see the teams that dominate at the end of the year winning by 40-50 against bad teams.

Just winning by 20 agains Appalachian St. may signal doom for you favorite team. :(

I tried using this method to pick NBA games the past two nights and I've only picked one wrong out of 6 so far, but used the spread. With 3 more tonight.

I used the last 10 games and counted close games(within 5 points) and blow outs( more than 10 points).

I will let everyone know what else I've found but the hypothesis that close wins(in a short time period) eventually lead to close losses or worse.

Great Article FO.

I'd love to see a piece on the existence of "momentum." I believe that there is no such thing as football "momentum." If a team takes the opponent's lead from 14 to 7, they've increased their chance of winning, but I don't think that being the most recent team to score really has an effect on the game. I believe the illusion of momentum is created only by the team's increased likelyhood of winning. Has there been, or will there ever be such an article? Your thoughts...

NCAA

Team: DOM-STOMP-GUT-SKATE
--------------------------
Texas: 7-4-1-0
USC:   3-6-2-0
--------------------------

Hmmmmm

My only complaint about this analysis is the definitions used. I don't necessarily disagree with the overall concept, or the proof of concept. But it seems to me that 14 points is not a stomp or domination. 14 points is 2 scores, which lately is not an uncommon thing to produce in a short period of time (3-4 minutes) given a turn of events.

I think there needs to be a redefinition and analysis. I think a stomp or domination would be 3+ scoring advantage (17+ points), a shutout of any nature, or a 14+ point win with few than 7 points by the opposition.

The other definitions would have to be altered dramatically, of course, but I don't think the current definitions are adequate. Technically, the Giants win over Dallas would qualify as a "guts" win, but I hardly classify it as such. A blown 17 point lead, a poor call that reversed an interception that would have radically altered momentum, and a number of other items only point to the Giants' win as adequate...hardly gutsy...perhaps lucky.

That said, while the numbers say something, how far does the back testing support the hypothesis? It seems to me there is too much emphasis on points, too. The one thing that the Giants' win indicated was that they did dominate in yardage, but that they are vulnerable because they beat themselves...not the sign of a particularly good team. In addition, their win over Philly 2 weeks ago would qualify as something, but not really domination (300 yards by a 3rd string QB?, against a team with a losing record and being completely reworked?).

I think overemphasizing the points in this (admittedly simplified) analysis tells us only that winning teams win. Which isn't saying much at all.

Interesting article. It (and my stubborn belief that the Patriots are better than they've shown) got me thinking about the goals of a football team in the regular season - I believe they are:

a) make playoffs
a.1) win division or
a.2) win wild-card
b) have a bye in playoffs
c) have home field advantage in playoffs
d) deceive other teams as to true strengths and weaknesses
e) be as healthy as possible during playoffs

Clearly (a) is the most important of the above, but what is more important of b, c, d and e? (b-e) are all aspects of 'Positioning for success in playoffs'. Could it be that one team values (d) more than (b,c,e) - or that they consider (b, c) harder to come by than (d) and more controllable than (e).

For example, the '05 Patriots are a good team playing in a poor division. Could there have been an assessment early on of their chances at the above goals and a conclusion that they could easily win their division without showing their true selves - but that it would be very difficult to get a bye or significant home field advantage in the playoffs?? - I like to think so, and it wouldn't surprise me if there was more inclination to give slightly injured players time off early in the season knowing that the schedule has most of their important games (divisional) at the end of the year.

How does this tie back to the article? If the above theory has merit, then a team that places a lot of value in (d) would be less likely to STOMP an oponent because they are trying to accomplish more than just a win, or they are trying to accomplish a win only within certain parameters (i.e. not divulging something).

I would think that each head coach would have a different ranking of (a) thru (e) in their minds and that some coaches would not even have (d) on their list. Based on interviews and the way the patriots treat their injury reports, I suspect that Bill Belichick ranks (d) as fairly important.

I also believe that we'll be seeing a different Patriots team than we've seen so far this year and that they will defeat the undefeated Colts in the playoffs. (that last is more a hope than an belief).

rmonihan:

Teams that beat other teams by 14 points win when they play them again later in the season far more often than not (they beat them by 7 points the second time, on average). Less than that and it becomes more common to lose to them the next time. This doesn't even compensate for the shift in HFA in the second game.

14 points is actually a good breakpoint. If you look at the distribution in second-game point differentials for wins of 14 points or more versus wins of 7 points or more, the difference is pretty noticeable. Things don't change when you go from 14 to 17, though.

(incidentally, the main outlier in that distribution for wins over 14 is the Bills/Patriots from 2003, where the Pats lost the first game by 31, and won the second game by 31.)

Pat,
Put that way, I guess it seems like a good break point, but what is the likelihood that a 14 point win results in a win the next time around? 55%? 65%?
I don't know the answer to that, but obviously you do.
I'll stick to my point, though, because 14 points is an amount that can be generated in a short period of time due to a number of bizarre reasons. Cases in point, while the Dallas Cowboys didn't beat Philly by 14 in the second go round, they would have lost if not for a very unusual 2 1/2 minute meltdown at the end. And is Seattle really 42 points better than the very same Eagles? Most of those points were generated in a very small span, resulting in a very lopsided result.

It's just a point to say that 14 points doesn't seem like much, that's all. A team that is 8-13 points down can generate 14 in 3 minutes or less, changing the overall outcome and aspect of the game. In fact, I remember San Fran (at their peak in the 80's) generating 28 in a single 4th quarter, to win a game they were down by 21 in.

San Fran is unusual in that they were REALLY good, but I don't think that, as good as they were, that this kind of is typical. Some people would say generating those kinds of points show just how good they were. I'd say it was damn lucky, mainly because something like that (and the Cowboys over Philly win) happens so infrequently. But it happens often enough to indicate just how little 14 points represents.

Again, I'm not arguing the concept or proof of concept. I just wonder about the criteria.

Sadly, there's no way to factor luck into this kind of thing...or is there?

But I'd say a game like this weekend's Giants/Cowboys game had all the hallmarks of a lucky win on it, not a gutsy one.

They win about 60% of the time. And I should've said "win by 5" not "win by 7" before. They either win, or lose by less than 7 about 88% of the time, however.

The other reason for 14 points is the fact that it's basically the width of the distribution. Grab any random game, and something like 70% of the games you grab will be decided by less than 14 points. So one team beating another one by more than 14 points is an outlier.

It is true that you can score 14 points very fast, however. That's just because it's 2 plays, and that's all. There is a way to deal with "luck", etc., however - use DVOA instead of point differential. DVOA will be a smooth deviate, which means you won't have the weird quantization that football scores produce.

But I'm pretty sure Aaron did this simply to dispel the stupid myth that people have that beating the crap out of a bad team doesn't tell you anything, whereas playing a good team close tells you something.

Putting up an article on FOX regarding the performance of those teams who have a low DVOA against poor teams might've been a bit much. :)

Does that performance against teams with a low DVOA information exist? If so, why would it be a bit much?

How far back does that 60% figure go? Is that for all NFL games in the modern era (since the merger)?

I figured that he posted that article to dispel the argument that just winning is all you need to do. Frankly, while the information is interesting, I don't find it overwhelming.
How many mediocre teams have huge wins offset by remarkable losses?
On the basis of dominating wins alone, how do teams that lead in that category perform overall? Is it possible (or even likely) that a team that has an overall mediocre record has a large number of dominating wins comprising its wins, and a large number of gutsy losses in the other column?
If so, wouldn't that call the concept into question?

This continues to be one of my favorite articles ever from the Outsiders.

Can the same theory be applied to college basketball?

Florida's margins of victory so far: 26, 22, 4, 13 and 15.