Tuesday, February 13, 2007

The Penalty Mystery

In his “Stat Relevance Watch: Wrap –up and Analysis”, the Sunday Morning Quarterback concludes his excellent series on the relevance of various college football stats to winning.

The SMQ conducted his analysis on a conference by conference basis, and then averaged the ranking of the relevant stat among conferences to determine which was most important to winning. His results, in order of relevance to winning, were –

Yards Per Pass
Rushing Offense
Total Offense
First to Score
Turnover Margin
Yards Per Carry
Time of Possession
Home Team
Passing Offense
Penalty Yards

While the list seems counter to every time honored football belief (when was the last time you heard a commentator say “Well, I think the team with the most yards per pass will win this one Lee”), the statistical accuracy cannot be challenged.

In looking at the list, one statistic stands out in particular – Penalty Yards. In every other single statistic, the team possessing the more favorable side of the equation was more likely than not to have won the game. For Penalty Yards however, in every single conference the team with the greater number of penalty yards wasmore likely than not to have won the game.

In other words, if you were an alien, and arrived on earth to be introduced to college football for the first time, in looking at the statistics on what it took to win you would be encouraged to commit as many penalty yards as possible, or at least more than your opponent.

In fact, the average winning percentage of teams committing fewer penalties is 40.5%. If you “flip” this average percentage, you find that teams committing more penalties won on the order of nearly 60% of the time, making the commission of more penalty yards seemingly more important to winning than having a greater number of passing yards than your opponent or even having the home field advantage.

“Lee, I think the team committing the most penalties will win this game today.”

My hypothesis in an earlier post that the team committing more penalties won more often was because the team with a longer time of possession had more opportunities to commit minor offensive penalties. These offensive penalties, in turn, were less harmful than defensive penalties to overall success, as while an offensive penalty rarely ever totally ruin a drive, defensive penalties more often than not extend drives, often disastrously.

My hypothesis was based mostly on Florida’s experiences this year. Florida’s least penalized game was also its only loss (Auburn), and also the game in which Florida held the ball the least amount of time.

In doing some more statistical analysis on this idea, I have found at least some basic preliminary evidence in favor of this hyposthesis. In looking at the SEC, Florida was the league’s most penalized team, while Vanderbilt was the least. The following shows the average penalty yards per game, the average time of possession per game (decimalized).

There seems to be some correlation between time of possession and penalty yards per game, although imperfect. The two outliers – Arkansas and Alabama, are explainable in large part because of the game they played each other on September 23rd. In that game, which Arkansas won by a single point, Alabama had possession for 39:32, or nearly 10 minutes longer than Arkansas. No other SEC game I saw had such a lopsided possession time, and it skews the results here in this small sample size.

However, in examining the game, the penalties to time of possession seems to bolster my position. Alabama committed its most penalties of the year in the game in 11. Arkansas, with a vastly shorter time of possession, committed only 5. Of Alabama’s 11 penalties, 9 were on offense, with most of them being the 5 yard false start variety.

The trouble with a definite determination of why penalty yards seems to help a team win, or at the very least doesn’t hurt it from winning, is the same problem we have with all our college football statistical analysis – lack of sample size. When you have only a 12 game sample size, a single outlier like the Alabama-Arkansas game can skew the whole exercise.

So, until I have more time to do further statistical analysis, coupled with the limited sample size available, the mystery of why penalty yards don’t seem to matter will remain that for now.


Anonymous said...

I'm working on responding to this Mergz, but in your time of possession column, you have "28:66," "28:96" amd "28:87." What does that mean?


Mergz said...

When I was first working on the stats, I changed the times from raw times to decimalized times because I was thinking of searching for some sort of ratio that might be valid (I didn't find one). So, instead of 28:40 for average time of possession on Arkansas, I made it 28.66 (note the decimal) for 28 and 2/3rds minutes.

I should have changed it back - my bad. But you can see what I was thinking.

Anonymous said...

Here's a thought: Defensive Pass Interference is skewing the results. If you get called for DPI you are likely *preventing* a touchdown, right? So the team that doesn't allow the other guys to pass for long yardage is forcing them to play harder to get the W. Just a thought ...