Tuesday, October 16, 2007

Win Probabilities and Expected Wins 2007 - Week 6

Sunday night game and commentary (you'll be real surprised) added
Monday night game added (another surprise) and expected wins table added

The following table lists the probabilities that the home team would win each game given the rushing averages, passing averages (includes sack yards lost and sacks as pass attempts), sack rates, third down conversion rates, penalty yards, and interception and fumble rates (by number of completions and rushing plays) for both teams. A logistic regression model trained on 1996-2006 data computes the probabilities. The final score margin is Home team points - Away team points. The sum of these probabilities of victory form the expected win totals in the second table to be added after the Monday night game. The first table will be updated as more box scores come in.

Cleaning up from last week, where clearly I was not on the ball, I had to re-update the week 5 article with the P(win) for DAL@BUF, where the probability of Buffalo winning given their stats was 46.9886% despite the 6 turnovers Dallas coughed up. Without further ado, here are the week 6 numbers.

Home TeamAway TeamP(Home Team Won) (%)Final Score Margin

Teams that "should have" won but didn't
Washington Redskins (@ Packers), Win Prob = 77.6814% Favre threw 2 interceptions; Campbell threw only one. Favre averaged 4.33 yards a pass; Campbell averaged 5.25 yards. Green Bay was lucky enough to recover all four of their fumbles, while Washington recovered all of theirs. Washington really did outplay Green Bay, but they seem to have lost it on a couple of unlucky bounces.

Seattle Seahawks (vs. Saints), Win Prob = 55.3389% I was at the game, and I can tell you the Seahawks were never in it. Holmgren essentially forfeited the game when he decided to punt down 3 scores with 11 minutes left in the game. Hmm, think you'll need to score there sometime, Bubba? That said, this might reflect garbage time stats. Seattle managed 4.4 yards per carry, while the Saints gained only 3.7. Shaun Alexander was getting booed because he was slow and couldn't find holes. Leonard Weaver, the FB taking over for Mack Strong, is inflating this average (3 for 40) as is Hasselbeck's scramble. The Saints had 5 rushes (4 by Brees, 1 by Karney) of 0 yards to deflate their average. Seattle also managed to gain 6.9375 yards per pass versus 6.833 for the Saints. On the other hand, the Seahawks allowed sacks on over 10% of their pass plays. The Saints O-line pulled it together and gave Brees all the time in the world. The reason this game is in the Seahawks favor is probably the Saints' 2 fumbles, only one of which was actually lost. Remember, going by the FO philosophy of fumble recovery being random, I count all fumbles. In the end, however, take out the botched punt and give the Seahawks that blocked field goal, and it probably would have been 21-13 at the half and a very different game, considering that the Seahawks shut out the Saints in the second half. Yes, the Saints were likely trying to run out the clock, but there were a couple drives where the Hawks D prevented a couple of what would have been surefire scores.

So what does this mean? Except for Hail Mary passes, I don't remove any garbage time data from the system, so maybe this means that the system is wrong. On the other hand, if the stats were so close before garbage time that some ultimately meaningless drives led by Hasselbeck could tip the stats in the Seahawks' favor, maybe the game was closer than everyone thinks. Again, despite the Saints' great pass protection, which had been a problem in their first four games, that game could have easily been 21-13 at halftime. And remember, the Saints' offense had only one productive quarter. Nobody goes 0-16, and the Saints finally got some luck bouncing their way. Why the hell can't the Dolphins get some of that?!

Atlanta Falcons (vs. Giants), Win Prob = 54.1972% Manning threw two picks, and I threw out Harrington's Hail Mary interception at the end of the first half. Manning was also probably a little lucky that the picks he threw were around midfield rather than backed up in his own territory. Otherwise, it looks like the Giants did outplay the Falcons.

TeamActual WinsExpected WinsProjected Win Total
AFC East
AFC North
AFC South
AFC West
NFC East
NFC North
NFC South
NFC West


Jordan said...

Probably a headache and a half to keep track of, but stats should be weighted differently depending on the score of the game. Like a 10 or 14 point differential cutoff line. Stats accumulated while the game is still close are more effective representations of a teams ability.

Derek said...

That's what footballoutsiders.com is for. I am using only box scores from jt-sw.com, and I can still classifies 85% of games correctly. But even FO's stats showed that Campbell outplayed Favre.

Anonymous said...

Either that is a misprint or your model really dislikes the Bears. They are projected to not win any more games rest of season while the currently winless Dolphins will win more than them?


I love football and statistical modeling but maybe football is just too short to get anything useful until the point anything it can tell us is already painfully obvious to many diehard fans anyway.

Free agency, injuries and emotion mess with models too much to get any kind of reliable forecasting based on numbers. Although 75% win IS pretty impressive! Regression due?


Derek said...

The model really dislikes teams that are ranked 30th, 26th, 26th and 26th in yards per attempt on rush off., rush def., pass off., and pass def. respectively. The Dolphins are 8th and 19th on rush off and pass off respectively. Yes, I'm a Dolphins fan, but the stats show they're not even the worst team in that division. They're not the 0-6 team people, including many Dolfans, think they are.

And I would expect some regression in accuracy. I've found accuracy to be highly dependent on home field advantage (which is really large for interconference games and nonexistent in divisional games this season so far).