The AllState Playoff Predictor

[Perd Hapley]

I’ve made some previous references to this, but the next time you see it pop up on ESPN (or anything related to % chance if making the playoff on ESPN because that’s what they use)…..just point and laugh. It’s totally meaningless, and is a completely broken algorithm. Same predictor had FSU at 95% or something last year.

But the most damning evidence is below. You can go onto ESPN’s website and choose any team, and fill out the rest of their schedule and get their final % chance of making it. For UGA, it gives them a 64% chance of making it…..at 9-3, and with a loss to UMASS.

Don’t think I’ve ever seen any predictive model miss something this badly. AllState should stick to the Sugar Bowl and cheating people out of valid hail damage claims. They straight up suck at this.

 

[patdog]

GA is in assuming they win out. Mississippi looks to be either the last team in or 1st team out.

never mind. Reread your post & realize this assumes GA loses. Yeah. No chance they make it with 3 losses.

 

[ZombieKissinger]

I could build a better predictor in 30 minutes with Google sheets

 

[85Bears]

FSU was 95% lock for the playoffs until Jordan Travis suffered a gruesome left ankle injury in week 12, Tate Rodemaker their backup was then injured and FSU started third string freshman Brock Glenn in the ACC championship.

So that’s kind of an anomaly. FSU was a lock with Travis but they weren’t one of the playoff teams with third string freshman Brock Glenn at qb.

 

[Perd Hapley]

FSU was 95% lock for the playoffs until Jordan Travis suffered a gruesome left ankle injury in week 12, Tate Rodemaker their backup was then injured and FSU started third string freshman Brock Glenn in the ACC championship.

So that’s kind of an anomaly. FSU was a lock with Travis but they weren’t one of the playoff teams with third string freshman Brock Glenn at qb.

They were 95% after the Louisville win….mere hours before the selection show. They never changed it to account for the injury. And that part is fine, because that had nothing to do with why they were left out.

This predictor is trash. The above scenario proves it. Nothing it says can be taken seriously….primarily because you can’t write an algorithm to account for human inputs and contingencies based on common sense.