Revisiting Close Games

by user Leftyloon

Several months ago I blogged about winning close games. The results show that a team's record in close games is not consistent from year to year. There does not seem to be an ability to win close games that carries over from season to season. However, there is another side to that coin, or more specifically two other sides (if that makes any sense). Do teams in a given season possess a certain ability to win close games for that year? To answer this question I perfromed two seperate regression analyses.

The first regression analysis involved seperating the 2005 season into two halves. The first half was each team's record in the first half of its schedule. The second half is the teams record in the second half of its schedule. If a team played an odd number of games, I included the extra game in the teams first half. For example, if a team played 11 games, the first half would be games 1-6 and the second half would be games 7-11. I then looked at every Division IA team's record in the first half of its schedule. I them compared this winning percentage in close games to the team's winning percentage in close games in the second half of their schedule to see if a team's record in close games in the first half of the schedule has any predictive value for the team's record in close games in the second half. Teams that did not play a close game in either the first or second half of their schedule (or both) were not included. With a sample size of 100 teams, the r squared value for this analysis is .0023. This means that a teams record in close games in the first half of its schedule has almost no value in predicting its record in close games in the second half of its schedule.

The second regression analysis involved separating each Division IA team's schedule, but in a slightly different manner. Instead of dividing the schedules into first and second half, I divided the games into even and odd. Game one goes into the odd pile, game two into the even, game three into the odd and so on. I then looked at each team's record in the odd games and compared it to their record in the even games to see if a team's record in close odd games was of any use in predicting its record in close even games. Teams that did not play a close game in either any odd or even games (or any at all) were excluded. With a sample size of 106, the r squared value for this analysis is .0007. This means that a teams record in close games in odd games has no value in predicting its record in close games in even games.

There is no ability to win close games (at least not in college football) whether it is across seasons or within seasons. Close games are like coin flips. Its not uncommon to get 3 or 4 head tosses in a row, but eventually things will even out and a few tails will be come up consecutively.