Saturday, January 5, 2013

Going against Jim Mora's wishes

I'm going to talk about playoffs.  Sorry coach.

As someone who has a forum through which to make such claims, I'm always in the market for predicting the outcomes of seasons, postseasons, and... well... not preseason, actually.  Normally I just go with whatever feels right, but I'm feeling frisky today, so we're going to go a little (Football) Outside(rs) the Box.  That's right, we're entering the realm of DVOA.

Defense-Adjusted-Value-Over-Average, or DVOA, is a proprietary statistic put out by Football Outsiders as a metric of a player or team's performance, adjusted for opponent, in each game situation they face, as compared to an average player or team's performance in that same situation.  For example, the computed value of a team gaining 10 yards is dependent on the down-and-distance (if it's 3rd-and-17, it's not nearly as valuable as 1st-and-10), how well an average team would perform in that situation (it's a lot easier to gain 10 yards on 3rd-and-17 than 1st-and-10 because the defense is playing softer), and the opponent (it's a lot easier to gain 10 yards on the Eagles than the Seahawks).  DVOA is represented as a percentage increase in value against an average player or team, and does not take into account volume (so CJ Spiller can lead the league in rushing DVOA despite getting 140 fewer carries than Adrian Peterson).

So how did I implement DVOA in playoff predictions?  Well, Football Outsiders has computed the DVOA of each team's offense, defense, and special teams for the entire season, which is a more reliable tool for judging teams than simple win-loss record.  Here are the data I used:

SEA 38.30% 18.50% -14.10% 5.70%
DEN 36.60% 22.10% -13.80% 0.70%
NE 34.90% 30.80% 1.30% 5.50%
SF 29.90% 17.00% -14.30% -1.50%
GB 26.60% 19.50% -7.30% -0.20%
BAL 9.80% 3.00% 2.20% 9.00%
WAS 9.60% 15.30% 1.80% -4.00%
ATL 9.10% 6.10% -2.90% 0.10%
HOU 6.60% 0.10% -14.20% -7.70%
CIN 6.10% -1.80% -3.80% 4.10%
MIN 2.00% 0.30% 3.10% 4.70%
IND -16.00% -2.90% 14.00% 0.90%

(Note: A good defensive DVOA is negative, indicating the opponent's offensive output is reduced by a percentage)

I've come up with a crude way of comparing teams'  DVOA for each phase of the game, which gives a rough estimate of which team should perform better on offense, defense, and special teams.  First, I ignored the actual meaning of DVOA for the moment and just use the raw number.  For each phase of the game (home team on offense, home team on defense, special teams) I subtracted the offense's DVOA from the defense's, and then aggregated them while weighing the phases by average number of plays per game (64 offense, 64 defense, and 7 special teams).  If the result is positive, I expect the home team to win, and if the result is negative, I expect the away team to win.  In the interest of respecting home field advantage, I gave the home team the edge in games where the result was close to zero.

Since the actual result of this calculation is meaningless with respect to anything tangible, I'm just going to go round-by-round with what I found, with big wins in bold and close wins in italics:


As those who have watched the whole season closely might have guessed, DVOA doesn't think very highly of the respected Texans, Ravens, and Colts (the latter of which has the worst DVOA of any 11-win team since 1991).  However, the Bengals aren't any great shakes either, so Houston should be able to squeak this one out.  The Seahawks are very highly rated in DVOA in all three phases, so it's no surprise that even on the road they're expected to beat Washington, since they run very similar offenses but Seattle has a much better defense.


The Falcons have looked like frauds for much of the season, and since DVOA loves the 'Hawks, it's no surprise that they are expected to pull another upset.  The Broncos and Patriots have been dominant for much of the season, so they should roll over the overrated Texans and Ravens.  The Niners eked out a win by my calculations without adding any home-field advantage, so I feel a little better about that pick.


Now we have the cream rising to the top, so it's all close games from here on out.  The Niners ride home-field to a really close win (on a neutral field I would have to take Seattle), while the Broncos should use their clearly superior defense to take care of the Pats.


Given that this is a neutral field (and is indoors), I feel much better about picking the higher-flying offense in the Broncos, and with their 11 straight wins, it's hard to argue with.  Do you think the Colts regret releasing Peyton?  Serious question.

Using DVOA gives a very intuitive result for the playoff bracket (with the possible exception of the Seahawks, but they have the look of the late-surging wild-card team that's done so well of late), and that's heartening, because sometimes advanced statistics produce some odd predictions because they're allegedly independent of situation or luck.  I feel pretty satisfied with making this bracket my pick for the actual results, although I'd really prefer the Colts beat the Ravens and the Texans beat the Bengals so that the Manning-Colts matchup can happen.  Although I bet it's hard for DVOA to incorporate the whole coach-coming-back-from-cancer thing anyway.

No comments:

Post a Comment