The line for FSU vs UofL has had a rocky fall since the opening line was released. The original line released on BOL was FSU -8.5. Since then, there have been 14 line changes over the past 36 hours, causing the line to bounce all the way down to -3, before rebounding a bit and currently sitting at -4. The fact that there have been 14 line changes shows that there is a huge disagreement in the betting marketplace about this matchup. So what is everyone looking at? We’ll use three of Sagarin’s rating systems to take a look.
Predictor – What It Does: The predictor system essentially does a nonlinear regression in excel to solve for power ratings. It essentially retrofits power ratings to minimize the sum of squared errors between each game’s expected outcome and actual outcome. It also uses a few tactics like taking the natural log of the sum of squared errors, rather than simply the absolute value. As a result, this gives diminishing marginal returns to teams who run up the score.
Predictor – What It Doesn’t Do: Predictor cares more about this season, so it doesn’t give much weight to how your team did last year. Sure there is some base starting value from last season, but the further into this season we get the less last season matters. This is honestly the way that it probably should be.
The predictor actually has UofL as a slight favorite at home against Florida State, so we can see why tons of bettors were hopping onto the Louisville +8.5 bet.
Elo – What it Does: I grew up with the Elo system from playing chess. It follows a very logical method of looking at who you won and lost against and adjusting accordingly. Beat a good player? You go up. Lose to a bad player? You go down. Lose to a good player, or a beat a bad player? You probably won’t change much in the rankings. While predictor looks more at margin of victory, Elo looks more at wins and losses. For football versions it has been adapted to incorporate margin of victory to some extent, but the bulk of it looks at wins and losses.
Elo – What it Doesn’t Do: Elo doesn’t understand the concept of a “season” as chess doesn’t have seasons. As a result, the only way to adjust for a new season is to do something like 2/3*Last Year Elo + 1/3*Average Elo for All Teams = New Starting Season Elo. This is a pretty bad method as it doesn’t take into account factors like returning starters, coaching changes, etc. It also allows teams who were undefeated last year to sustain their high Elo rating, as long as they keep winning, regardless of their poor margin of victory this season. (For the record, BOL often uses this method the most for their opening lines).
Under this method, Florida State is almost a 10 point favorite. Again, you can see why there is some huge disagreement on this one.
Sagarin Rating – The Sagarin rating is a combination of multiple handicapping methods, most notably the two listed above. As you can see, combining the two gives us a line of -4.34 which is around where the line has currently stabilized.
Florida State is 1-6 ATS, with an ATS margin of -8.4. The only other team with a worse record ATS is UCLA, but they only have a margin of -4.9. In other words, Florida State has been one of the most unprofitable team to bet on against the spread.
Louisville has been a pretty average 5-3, with an ATS margin of +3.4.
What about Thursday nights?
Underdogs of 3.5 to 10 points playing in Thursday night games have covered over 55% of the time, with a sample size of more than 700 games. There’s something about playing on a different night that makes favorites more prone to upsets. A 55% cover rate is about equivalent to 3 points on the spread.
Depending on what method you use though, Louisville may not be a 3.5 point underdog, or even an underdog, so it’s hard to really make an adjustment for that.
How would I handicap this?
Predictor is a better rating, but the Elo is so high that an adjustment must be made. Let’s take the predictor line as-is, and then let’s try to adjust the Elo for FSU’s poor cover rate. Then finally we’ll slant the line a few points in Louisville’s favor for the Thursday night game. Not 3 points because they might not really be 3.5 point underdogs, but mentally they’re underdogs, so let’s take a 0 to 3 point range and use that as our expected range for the closing game line.
1. Predictor = Louisville -0.7
2. New Elo = -9.79 – ln(8.4 ATS margin) = FSU -7.66
3. Combined Rating = FSU -3.5
4. Adjust for Thursday = [ FSU -3.5 to FSU -0.5 ]
5. FINAL LINE RANGE: FSU -3.5 to FSU -0.5
6. Win Probabilities:
High – FSU 65.3%, UofL 34.7%
Low – FSU 51.7%, UofL 48.3%