Location-based Retrodiction Analysis

[Sef Dresslar]

Just thought I'd share something I just looked at.

For the 2000-2011 seasons I looked at each game result and broke up a retrodiction analysis based on game location.  I used the simple power rating for each team and the FBS universe only.  The results:

Road win: 2518-579 (0.813)

Neutral win: 394-141 (0.736)

Home win: 3683-1004 (0.786)

It seems like it would be better to pull the game result in question out of the power rating calculation in each case, but that would take way too long for the level of info I was looking for.  Anyway, what I thought was interesting was that neutral site upsets were more common than upsets pulled off at home, and pretty significantly so.  My only thought is that most neutral site matchups obviously happen in bowl games so the extra preparation time helps the underdog, evens teams up through injuries healing, etc.

Also posting in case I totally screwed up the analysis somehow and any of you have better info on this.

[Sef Dresslar]

Took out bowl games.  Neutral site then was 148-38 (0.796), fitting nicely between road and home.

[Matthew Smith]

Have you adjusted for spread and/or difference in team quality?  I'd ask this because bowl games and CCG's (which make up majority of neutral site games) generally tend to be pretty even matchups, while home/away games (both in and out of league) have a great many mismatches.  Assuming that your analysis was basically "what is the record of the higher rated team", I'd assume this would then explain it.

That's probably easy to check; you could bucket road/neutral/home games into buckets based on how different your power ratings are for the two teams (near tossups, mild mismatches, major mismatches, expected blowouts, etc.), and then see what the varying win %'s are inside of each bucket.

I did something a while ago that wasn't QUITE like that, but was along a similar vein, basically using league records as a proxy for team quality, bucketing games by how different their final league records were, then looking at home/road splits (neutral site league games are rare enough to not be worth adding to that analysis) for intra-league matchups only
cfn.scout.com/2/1059791.html

I also did a follow-up on that, looking specifically at the impact of distance on home-field advantage, basically concluding that HFA is in fact dependent at least to some degree on distance.
http://cfn.scout.com/2/1063535.html