The Allstate Playoff Predictor at ESPN is pretty cool. You can set the results of future games and get immediate feedback as to the team's odds of making the Playoff based on the FPI's projection (we'll just call it Allstate's projection from here on, as if they had anything to do with it other than stamp their name on it as a sponsor). We're looking at the version from October 26 and we'll review its conclusions, which are mostly on track but sometimes leave us scratching our heads.
By comparing two teams at a time and showing their odds if they win out, we can see what the projection is thinking. Remember, these are not odds to win out, these are Allstate's odds to make the Playoff given that the team wins out.
Georgia (13-0): >99%
Alabama (12-1): 99%
If Georgia wins out and goes 13-0, they have a >99% chance of getting to the Playoff. That's not a tough conclusion, in fact I wouldn't blame them if they just said 100%, as no 13-0 Power Five team has ever missed the Playoff and Georgia would be one of the strongest candidates yet. Alabama winning out, on the other hand, yields a 99% probability. I guess that's accurate, though I think >99% would be fine, too. SEC champ and beating Georgia ==> you're in the top four, guaranteed.
Above: The AllState Playoff Predictor had a really high opinion of the Alabama State Hornets, giving them a 50% chance to win the SEC. Just hours ago they decided the Crimson Tide were a better fit.
Georgia (12-1): 90%
Alabama (11-2): 59%
Here are the odds for each if they LOSE in the SEC title game. Georgia at 12-1 still has a 90% chance of making it, which might be a touch high. They might not have the normal 70% for a 1-loss Power Five team since they wouldn't have a conference championship (only 37.5% of 1-loss non-conference champs have made it). Alabama's 59% odds at 10-2 seem a bit high since no 2-loss team has ever made it. I'd give Georgia the 59%, Alabama closer to 39%.
Michigan (13-0): > 99%
Michigan St (13-0): 95%
Michigan at >99% makes sense if they finish 13-0, but 95% for a 13-0 Michigan State is too low considering they have exactly the same path to the title as the Wolverines. There's no 5% chance a 13-0 Michigan State ends up outside the top four; there's just no conceivable case where that happens. Both results should be >99%.
Oklahoma (13-0): 99%
Cincinnati (13-0): 88%
Now we compare two teams finishing 13-0. If Oklahoma wins the Big 12 they're in 99% of the time. I'd call it >99% but maybe there's a chance a 13-0 Oklahoma still looks bad enough to be left out for the very first time for a 13-0 Power Five team. Cincinnati's 88% is correct in being lower than Oklahoma's odds, but probably still too high. No non-Power Five team has made it yet, so I'd even suggest 50% is on the cusp of being too high.
Ohio State (12-1): 95%
Oregon (12-1): 28%
This one's a kick in the pants. If Ohio State wins out they get in 95% of the time; Oregon, who beat Ohio State, only gets in 28% of the time. Both of these numbers are too low: the Buckeyes at 12-1 with a Big Ten championship should be >99%, and Oregon would at least be the basic 70% shot a 1-loss teams has, even if the Pac-12 championship doesn't carry much weight. Add to that their win over Ohio State and 28% is a joke. They'll get a lot more consideration than that, probably 60%. I don't doubt Ohio State finishes higher at 12-1 than a 12-1 Oregon, but the difference in making the top four is not this great. Remember: 87.5% of 1-loss Power Five conference champs have made it in the past.
Iowa (12-1): 44%
Iowa St (11-2): 40%
Here's another one that's hard to explain. A 12-1 Iowa, winning out and winning the Big Ten, sits at just 44%? That's almost exactly half of the historical 87.5% rate for 1-loss Power Five conference champs. They do have a bad loss to Purdue and they wouldn't have the same tough schedule the teams in the Big Ten East did, so their odds wouldn't be >99% but would be pretty close considering they have to beat the East champ. Meanwhile Iowa State already has 2 losses; how does a Big 12 championship give them a 40% chance? When no 2-loss team has ever made it? It's not like there's a dearth of teams that could finish with 1 loss or fewer, and with Cincinnati around how does Iowa State ever get it? Also, Iowa beat Iowa State and it looks like only Iowa State is benefitting from that game.
Oklahoma St (12-1): 30%
Baylor (12-1): 31%
Here's more insult for the Big 12: Oklahoma State and Baylor, as 12-1 Power Five conference champs, don't even match the odds that Iowa State pulls as an 11-2 team. It's not like there's a big gulf separating them. Yes, Iowa State just beat Oklahoma State. But Baylor beat Iowa State and Oklahoma State beat Baylor. The only difference in these teams is that Iowa State lost to Iowa; it seems the SOS boost from that game outweighs the additional loss. Both Oklahoma State and Baylor should at least be above the 70% chance that a 1-loss Power Five team has, and Iowa State's odds should be closer to 25%. But Allstate is right in that these teams, who have basically cloned schedules and results, would have almost the same exact odds at 12-1. It's just that Power Five conference winners with 1 loss make the Playoff at an 87.5% clip, not 30.5%.
Pittsburgh (12-1): 70%
Notre Dame (11-1): 41%
Mississippi (11-1): 39%
Mississippi (12-1): 85%
Finally a grab-bag of three teams in different situations. Pitt for some reason has really high (70%) odds if they win out and take the ACC, much higher than comparable 1-loss teams from the Big 12, a much tougher conference. Did Pitt win a big non-conference game? No, they beat Tennessee, and lost to Western Michigan. They also beat UMass and New Hampshire so it's not their SOS giving them a boost. A 1-loss ACC team should probably be about the same as a 1-loss Big 12 team, but it looks like Pitt's high FPI is giving them an "eye test" dividend. But 70% isn't too high, as that's the rate for ALL 1-loss Power Five teams, so even if you give them zero credit for winning the ACC they're in the right ballpark.
Notre Dame has no conference title game, so if they finish with 1 loss it's all about their schedule, and this year it's turned out to be a bust. Counting them as a 1-loss Power Five team without a conference championship which carries a (small sample) 37.5% success rate, 41% seems about right. Having Cincinnati as their only loss helps, but it means Cincinnati is ahead of them for certain at 13-0.
And Mississippi probably has no conference title game to play (Alabama would have to lose again), so 11-1 is their best finish. 39% seems a bit high at first blush as they're behind Georgia for certain and probably an 11-2 Alabama too, but really they're in the same boat as Notre Dame so 39% probably isn't far off. If they make the SEC title game and win it their odds increase to 85%, which is rather low for a team that would have to beat Georgia, with Alabama as their only loss. How would you keep the SEC champs out of the Playoff under those circumstances? Maybe there's a 1% chance the Committee would find four more deserving teams, but not 15%. The only 1-loss Power Five conference champs that haven't made the Playoff were Baylor in 2014 (a confusing 11-1 with no title game win and TCU declared co-champ) and Ohio State in 2018 (suffered a 49-20 loss to Purdue).
=======================================================================================================
There you have it. From what we can tell, schedule strength is overvalued by these projections, overriding actual wins and losses. Hence Iowa State's boost from playing Iowa despite not beating them. And it looks like it's using the FPI as the "eye test" which overrates the odds of 12-1 conference champs like Pitt compared to a similar 12-1 Oklahoma State or Baylor, and gives a wholly undeserved edge to a 13-0 Michigan over a 13-0 Michigan State. At 13-0 and winning the Big Ten there are no "style points", you're just in, period.
It's too bad Wake Forest wasn't included (ESPN only includes the 15 teams with the current highest odds per FPI) because I would love to see how poorly a 13-0 Wake Forest does, given their low FPI and weak schedule. Something tells me it would be around 35%. This is roughly consistent with their FPI odds to win out (0.6%) vs. their odds of making the Playoff (0.2%). No 13-0 Power Five conference champ has missed the Playoff, so the number it returns should at the very least be over 50%.
Of course, this doesn't mean Wake Forest will show up prominently in the first Committee rankings. Their schedule and performance puts them below all the other undefeated teams, and probably several (or most) of the 1-loss teams. That will change the more they win.
Comments