Thursday, October 22, 2009

NFL after week 6: Over/Under And Referees

After week 6 (through games of October 18), let's take a look at whether there is a correlation between NFL referees and the over/under result of the games.

I split up the season so far into two groups: the statistic for weeks 1-3 and the statistic for weeks 4-6. With this extremely limited data, let's see whether the distribution among the referees for weeks 1-3 have a correlation to the distribution for weeks 4-6.

The correlation for this limited data is 0.642. So there is a strong correlation with this very limited data. The referees who mainly had higher than expected scores in their games in weeks 1-3 for the most part also had high-scoring games in weeks 4-6. Those who had low-scoring games in weeks 1-3 kept up that trend in weeks 4-6.

Looking at this another way, if you bet that (a) the referees who had 100% over/under in weeks 1-3 would always be over for weeks 4-6 and (b) those who were 0% for weeks 1-3 would always be under in weeks 4-6, you would have gone 9-2! An 81% betting success rate over three weeks looks pretty good.

Maybe the people who set the over/under lines are not taking into account the referees in the game. Maybe this is a hidden advantage for those keeping track of it.

But watch out -- this analysis is based on only six weeks of games. Maybe it is just a statistical anomaly because of the small sample size. Or maybe there is a correlation in the early part of the season but referees have less influence over the points scored as a season progresses. Or maybe some referees will hear of this analysis and actively work to have less unconscious influence on the total points scored per game.

Post your comments about whether you think this is a statistical anomaly of no significance or if you suspect it might be a valid theory...



Post a Comment

<< Home