## Thursday, July 13, 2006

### It's How You Finish

How important is ending the season on a winning note? It seems logical that teams that finish the season strong would have momentum in recruiting and a general good vibe among their fans for the following season. These factors combined with the standard maturity of college athletes after another semetser at school should equate to marked improvement for the team the following season. But is this logical progression true? Let's find out. To answer this question, I looked at how each BCS conference team finished the season in 2002, 2003, and 2004. I then looked at their record the following season (2003, 2004, 2005). To define finishing the season, I used each game played on or after November 1st. For most teams this was a sample of between 3-6 games. Of course better teams play more games after November 1st thanks to bowl games and conference championships. I then ran a regression analysis to determine how well the winning percentage in these closing games predicted each teams winning percentage the following year. Here are the R squared values for each of the three seasons.

End of 2002 Predicting 2003: .3432
End of 2003 Predicting 2004: .2976
End of 2004 Predicting 2005: .1445

There seems to be something to this logic. The way a team finishes the season does have at least a limited impact on their performance the following year. Is there a better way to judge how a team finishes a season? In order to access whether there is, I will run the same analysis, but instead use net points instead of winning percentage to predict the subsequent season's winning percentage. Net points are simply the amount of points a team has scored more or less than their opponents. A positive differential indicates the team in question has outscored their opponents, while a negative differential means they have been outscored. For example, if a team goes 3-1 from November first onward they have a winning percentage of .750. However, if their 3 wins are by 7, 3, and 4 points, and their loss is an annihilation by 40 points, then their net point total is -26.

End Net Points 2002 Predicting 2003: .3297
End Net Points 2003 Predicting 2004: .3256
End Net Points 2004 Predicting 2005: .2515

Again their appears to be at least a marginal relationship between a team's closing net points and their winning percentage the following year. This relationship is much more consistent than actual won/loss record in the season's final few games. Further validation that teams are a function of their points scored and allowed more than their won/loss record. However, the effect of finishing a season strong should not be overstated. As noted in an earlier post, a team's record for the entire year and Pythagorean record for the whole year are better predictors of their record the following year (at least for 2004-2005).