Sunday, December 17, 2006

NFL Referee Statistical Tendencies (thru 12/11)

In this article, I examine which types of penalty statistics by referee seem to be based on the referee (and not just something random). The way I go about this is to see whether there is a correlation between the statistical breakdown by referee from weeks 1-7 and from weeks 8-14.

To put it another way, let's pretend that there are only three referees in the NFL and we compare the statistics for how often the home team wins. If we see a correlation of 1, then it seems like the statistic does relate to the referee. For example, if in quarter one 100% of ref A's games have home team wins, 50% of ref B, and 0% of ref C and we see the same statistics in the next time period, then it seems like it is related to the ref and not purely by chance.

Included are also the correlation values between the 1Q and 2Q of the NFL season (weeks 1-4 and weeks 5-8).

Let's see for a variety of statistics:
Largest correlations:
Penalties against visiting team (0.318) (1Q-2Q correlation was 0.237)
Total penalties (0.262) (1Q-2Q correlation was 0.376)
Penalty yards against visiting team (0.274) (1Q-2Q correlation was 0.188)
How often the visiting team has more penalties (0.242)
Total penalty yards (0.165) (1Q-2Q correlation was 0.363)

Very interesting -- there is a hugely negative correlation for the rate at which the home team wins. (-0.407) This suggests that the rate that a home team wins the game is not correlated to particular refs.

What does this suggest? Maybe the percent that a home team wins is not across-the-board affected by all referees. The variable that has a bigger correlation across-the-board to referees is the total number of penalties accepted in the game and the number of penalties accepted against the visiting team.

What does that mean? Some refs call more penalties than others and the discrepancies among referees' tendencies seems to exist throughout the season. So this suggests that some refs call the game differently than others in a consistent way through the season (or some believe in letting the players play while others tend toward enforcing the rules more).

On Replay Challenges: Using the statistics at Mike Sando's excellent blog at The News Tribune, I have run the statistics to see if there is a correlation between 1Q replay challenge reversal rate and 2Q replay challenge reversal rate. There is a correlation between 1Q and 2Q at 0.163 -- and see Mike Sando's blog for the full statistics (although he has not done correlation analysis).


Blogger Mike said...

Rex ... Nice work. Please email me if you get a chance. mike.sando(at) Thanks.

Mike Sando

11:35 PM  
Anonymous Anonymous said...

Eagles colts game nov 7 refs SUCK! Bad calls bad calls just give players flags or play two hand touch!

7:35 PM  
Anonymous Anonymous said...

Nobody wants anyone hurt but knee jerk reaction or reaction with no common sense will ultimately damage the way the game is played. Ie eagles colts game

8:31 PM  

Post a Comment

<< Home