College Football Prediction Models
After many, many months of waiting for this weekend to come, it’s finally here; the real opening weekend of college football after the appetizer that is Week 0. We’re in for a treat with some juggernauts doing battle here in the non-conference with Texas visiting Ohio State, LSU vs Clemson in the battle of “The real Death Valley”, and Alabama playing a true road game at Florida State.
For the sickos (me), there’s also Temple vs UMass and Kent State vs Merrimack on the card, so plenty of action to go around. More importantly, how can we bet these games and what do the numbers say? As part of my weekly contributions here at VSiN.com, I’ll be analyzing and writing about how sophisticated projection models like my T Shoe Index, SP+, FPI and Sagarin all align or differ on a handful of games each week. This analysis should help confirm your previous betting leans (when the models align) or perhaps keep you off of a play and help you whittle your card down (when the models disagree).
Note: SP+ Rankings are the work of CFB analyst Bill Connelly (ESPN), FPI is the Football Power Index from ESPN Analytics, and Sagarin is Jeff Sagarin, formerly of USA Today.
Week 1 College Football Model Alignment:
Clemson vs. LSU (+3.5)
It just worked out that this week, one of the biggest model agreements was on a marquee game, though I don’t always expect that to be the case. There is only a 0.3-point variance from the largest projection to the smallest projection on this game, with an average of Clemson -2 being the consensus. TSI, FPI, SP+ and Sagarin all agree Clemson should be favored by 2.0, 1.9, 2.1 and 2.2, respectively. Talk about consensus. Interestingly, we’ve seen the market move towards Clemson, so this will be a nice test case of models vs. the market. But, if you want to back LSU on the road, the models agree there’s data to support that bet.
Washington vs. Colorado State (+21.5)
In a more under-the-radar game this weekend, the Huskies will host Colorado State as 21.5-point favorites, although there are some 22.5s still out there. Our model consensus has a variance of 1.6 points with an average of Washington -20.5, which seems small but is on the other side of a relatively key number (21). SP+ projects the closest game, with Washington pegged as a 19.4-point favorite, while FPI is right on 21, with the other models falling in between. This isn’t a huge edge, but if you’re thinking about pulling the trigger on the Rams, the models indicate you’re getting a good number at 21.5 or better. If you like Washington, you’d want to get them at 19 or better to have model consensus on your side.
Tulane (-6) vs. Northwestern
This is one of the most intriguing models vs market games of the week. The market has been all about Northwestern, driving them from +7.5 all the way down to +4.5 at one time, and it’s now back to 6. The models, however, indicate this should be a comfortable win for Tulane. With a variance of 3.2 points and an average projection of Tulane -13.3, the models are loud and clear that Tulane should be a much heavier favorite, which makes me wonder why the market has been so bullish on the Wildcats. The lowest projection is FPI’s Tulane -11.4, while Sagarin has the most Green Wave-friendly projection at -14.6. I bet Tulane -6 over the summer, and the models only give me more comfort in that position, as a data guy, but the market is really weird on this one. I’m fascinated to see how it plays out.
Week 1 College Football Model Disagreement
Southern Miss (+13) vs. Mississippi State
I played Southern Miss +13.5 here, for full transparency, but our models have a wide range of predicted outcomes here. TSI leads the more optimistic view for the Golden Eagles, projecting Mississippi State -4.5, while SP+ conversely has the Bulldogs by 16.4. The model average is Mississippi State -10.2, so still comfortably under the 13, and Sagarin and FPI also align with TSI on the underdog side here, but I thought it was worth noting there’s such a broad range of outcomes, likely due to the huge roster and coaching staff turnover from last year to this year for Southern Miss.
Wyoming (-5.5) vs. Akron
Another broad-spectrum of projections here, and truthfully it makes me happy to see TSI being the most conservative on one game (USM vs MSU) and the most liberal on another (this one). The TSI projection here is the most Cowboy-friendly, projecting a 16-point Wyoming win; however, FPI has this game projected as just a 4.6-point difference. Our model average is Wyoming -8.5, which is where this game opened over the summer. SP+ leans underdog with a projection of 5.4 and Sagarin is in line with the favorite, projecting Wyoming -8.4. I personally bet Wyoming -7, so I did not get the best number but the model average gives me some hope I can still cash this ticket.
Texas State (-13.5) vs. Eastern Michigan
I told you these wouldn’t all be marquee games, and nothing screams “sicko” like betting Eastern Michigan vs Texas State, am I right? There’s a 10.9-point variance in projections on this game, with an average of Texas State -14.8, so getting on the other side of the key number of 14 is crucial. SP+ is actually the most conservative projection here, with just a 10.6-point edge in favor of the Bobcats, while Sagarin is calling for the blowout with a 21.5-point projected win for Texas State. TSI and Sagarin indicate the favorite is the play, SP+ and FPI are on the ‘dog. It will be interesting to see how it plays out, but with mixed data and the market moving towards EMU, I’m personally staying away.
One note one this segment: FPI seems to really underestimate larger disparities between teams near the top and teams near the bottom (i.e. Penn State vs Nevada), so I’m going to try to keep this segment to spreads of 14 or less, where there’s less room for huge discrepancies.
Keep an eye out for this weekly feature on Thursdays and also my Monday early-week college football bets.
