We're revisiting the AllState Interactive Predictor again, after comparing it to 538's College Football Predictions in previous articles here and here. The general conclusion of those articles was that 538's engine was superior, and today's comparison doesn't change that though both have their blind spots.

All these comparisons show the teams finishing the best they can realistically finish, i.e. winning out and if that gets them to their conference title game, winning it, with exceptions as explained. The first percentage column is AllState/ESPN's odds, the 2nd is 538's. We indicate in green which projection is more realistic if there is a clear winner. First we look at Georgia and Alabama, currently #1 and #2 in the CFP Committee rankings.

**Team** **Rec** **ESPN** **538**

```
Georgia 13-0 >99% >99%
Georgia 12-1
```**>99%** 58%+
Alabama 12-1 99% **>99%**
Alabama 11-2 67% 29%+

Both sites have 13-0 **Georgia** as a lock. Congratulations! But there's a discrepancy in a 12-1 Georgia. If the Bulldogs lose to Alabama, ESPN still has them a lock, while 538 is much less certain. We can't set 538 to make sure Georgia goes 12-0 then loses to Alabama, so we use the odds that show if Alabama wins out and Georgia wins their immediate next game (Tennessee). It gives 58%, which is the minimum result they'd show for a 12-1 Georgia. To us it seems >99% for a 12-1 Georgia is a touch high, but not by much. Maybe 95%? 99%? But not a lock.

Same thing for **Alabama**, ESPN isn't willing to give a 12-1 Alabama the "greater than" sign. In other words, AllState thinks a 12-1 Georgia that just lost to Alabama is a >99% lock, but the 12-1 Alabama that just beat them is only a 99% lock. Interesting result!

An 11-2 Alabama that loses to Georgia in the SEC title game has 67% odds per ESPN (too high!) while 538 can only be coaxed into giving them 29% odds (probably about right, maybe a touch low).

The rest of the 1-loss or undefeated teams:

```
Oklahoma 13-0 >99% >99%
Ohio State 12-1 97% 98%
Michigan 12-1 96% 98%
Michigan 11-1 79% 84%
Cincinnati 13-0 89%
```**62%**
Notre Dame 11-1 48% **30%**
Oklahoma St 12-1 80% **98%**
Oregon 12-1 28% **98%**
Mich State 12-1 50% **97%**
Wake Forest 12-1 5% **63%**

Well at least they're both saying **Oklahoma** is a lock at 13-0. And the difference in forecasts for **Ohio State** is minimal and seems accurate, based on whether Oregon might hold them out or not.

The numbers for **Michigan** are solid for both projections, whether or not the Wolverines make it to the Big Ten title game.

Now the craziness starts.

AllState/ESPN gives **Cincinnati** an 89% chance if they win out. 538's 62% is much more reasonable.

AllState/ESPN gives **Notre Dame** a 48% chance if they finish 11-1. It's hard to see Notre Dame having a 50/50 shot at the Playoff at 11-1 given all the teams that could finish ahead of them, and Cincy occupying a spot in line in front of them in most scenarios. 538 is again more realistic at 30%.

ESPN also gives **Oklahoma State** only an 80% chance compared to 538's 98%. The Cowboys would have to beat Oklahoma twice to get to 12-1, and that won't leave them out of the Playoff unless they're a lot of craziness.

Here's the real kicker: AllState's widget gives a 12-1 **Oregon** just 28% chance of making the Playoff. Think about that: Oregon is #3 in the rankings right now. And we're setting the widget for Oregon to win out and win the Pac-12, and it's saying that almost 3/4 of the time, that knocks them down 2 spots or more. 538's take on this is again much better: about 2% of the time we could see Ohio State and Cincy both passing them, perhaps.

And this one might be worse: A 12-1 **Michigan** **State**, having won the Big Ten after beating Michigan, Penn State, Ohio State, and probably Wisconsin, makes the Playoff half the time, almost the exact odds they give an 11-1 Notre Dame who has Wisconsin as their sole decent win. Since this option eliminates Ohio State from the mix it's hard to see the Spartans missing out half the time.

And finally, a 12-1 **Wake Forest**. 538 correctly notes that a 12-1 Wake Forest isn't the same as a 12-1 Michigan State, Oklahoma State, or Oregon, giving them a 63% shot. Remember that Power Five teams that win their conference at 12-1 have made it 87.5% of the time in the past. But AllState/ESPN gives the Deacons just a 5% chance.

To be fair to AllState/ESPN, they do a lot better with 2-loss teams than 538. No 2-loss team has yet made the Playoff.

```
Pittsburgh 11-2 4% 25%
NC State 11-2
```**1%** **35%**
Texas A&M 11-2 **35%** **54%**
Mississippi 10-2 **1%** **10%
**

It's hard to say what the "correct" values are here since no 2-loss team has made it yet. It's like figuring the odds of a 16-seed to beat a 1-seed before UMBC-Virginia. You could say the odds are 0%, but with only six years of valid stats, and three 2-loss teams finishing at #5, zero percent is too low. But 538 assigns some pretty high values. **Pittsburgh's** 25% seems a bit high considering all the 1-loss competition, while AllState's 4% might be a touch low. But **NC State** shouldn't be higher than Pitt at 35%, that's absurd.

**Texas A&M's** 54% is too high, too. Yes, the Aggies beat Alabama and would have to beat Georgia too, but that doesn't put them above 50%. AllState's 35% is more realistic. We also don't know how a 10-2 **Mississippi** with no conference title has 10% odds to make it—even thought that's the number our own projections give the Rebels, mainly because they are forecast to finish at #8 and that puts them "close". We never said our projections didn't have kinks to work out.

The AllState/ESPN widget only allows manipulation of the top 15 teams, so 3-loss teams are out, but here's some mess from 538:

** 538**
Iowa State 10-3 **46%**
Wisconsin 10-3 **21%
**

Somehow, a 10-3 **Iowa State**, having made it to the Big 12 final, beats Oklahoma and finishes 10-3 and that gives them....a 46% chance of making the Playoff? No 2-loss team has made it so we can safely say the odds for a strong 2-loss team are still less than 50%. No 3-loss team has come close, so a strong 3-loss team should have about 10% at the very most. Even a Big Ten-winning **Wisconsin** shouldn't show 21% odds.

So once again we confirm that 538's widget is much better than AllState/ESPN's, at least for 0- and 1-loss teams. Those are the teams that matter, however. Just use 538's widget and ignore 2-loss and 3-loss teams altogether—unless Alabama finishes 11-2, then we don't know which one will be more accurate.

## Comments