It's hot as the dickens here in Wisconsin, which I am not a fan of. The fiancee gripes about our air conditioning keeping the apartment too cold, but in my opinion: put on a dang sweater. I can only strip off so much before indecency and/or flaying occurs.
RECORD COMPARISON, 2006-2015
Record vs Top 25
Win Lk Comparison
|Conf||Record||C-Rec||Tot Rat||rk||Adj Rat||rk||Res Rat||rk|
|# Tms||Prj Rec||Prj C-Rec||Adj Off||rk||Adj Def||rk||Prj SOS||rk|
|Perf W||1 Loss||Bowl||Ret Off||rk||Ret Def||rk||Rec 3yr||rk|
|Gm||Wk||Date||H/V||Opponent||Opp Rk||FCS||C-Gm||Prj Lk|
|Opp Rk||Team||B10 Avg||Opp %ile|
One of the things I've been trying to highlight during these low-quality previews is a team's record in close games (defined as games with a final margin of 7-pts or less). It's more-or-less accepted wisdom that close games lend themselves to more danger in the result, since it is harder to lose by virtue of a missed field goal when you are already up 21 pts. Generally, the rule-of-thumb is that teams should go 0.500 in the close games they find themselves in, and (while that can be a bit simplistic) justified skepticism can be given toward those teams that far outpace that assumption. Given that it is justNorthwestern week, this is where I note that they went 5-0 in close games during the 2015 season and got apple-cored in their three losses (including the bowl game). That bowl game loss was to Tennessee, a four-loss team that went 2-4 in close games. Just putting that out there for y'all to chew on.
While the 0.500 assumption meets the "eh, good enough" threshold to which I usually aspire, I've tried something a bit different when looking at close games and W/L records. Let's be honest, when Ohio State found themselves in a close games against Indiana and their backup QB, most of us were still betting on the Buckeyes. Putting themselves in that close game territory certainly gave them less room for error (see: pass-interference in the end-zone that went uncalled), but they're still the better team. We'd still expect them to win most of the time.
So consider a game where one team is a 7-pt favorite. I'd say they're probably 70% likely to win (and considering historical Vegas results, that's about right). But that's considering all results, including the 59-0 blowouts (as an example). Limiting the results to those that would be considered close (7-pts or less, in this case), the odds are a bit more compressed, and given a large enough sample size we can come up with a decent model for comparison. Below is a model of how we might apply this (using game results since 1990):
Since a team favored by 30-pts doesn't usually find itself in a close game, the data ges a little wonky at the edges. But we still get a decent enough picture of how to model "close game win likelihood." In case you didn't read the primer above (and I know you didn't, it's okay), the model we're using is a logistic model, using the following form:
W-Lk = 1/(1 + exp(-cf*ExpMOV))
A smaller coefficient lends itself to a flatter curve, hence the relative values above.