The flaw in ESPN's BPI rating

[Aike]

Someone mentioned in another thread that we are only 18th in BPI. Even as we climb in Kenpom, Sagarin, and every other measure, we have basically stayed put in BPI.

I suspected that I knew what the flaw was, and did a little research to discover the truth.

BPI gives less weight to games where key players are missing. This sounds like a good and fair thing. But what if that key player is Alex Poythress, and you've actually played better without him?

UK's three best BPI scores have been in the past 3 games. Scores of 99.2, 99.4, and 99.5. Our 4th best score was way back in November, when we scored 99.1 against Duke.

Because Poythress has missed these past three games, they have only counted as 7/8 of a game each.

ESPN should have a mechanism to give you full credit when you perform better than usual despite a key injury. If we had been given full credit in the model for these past 3 games, we would be up around 13th by my math. That would make a lot more sense.

 

[Aike]

Here's another interesting factoid in regard to BPI.

UK is 5 - 1 vs. BPI top 50 teams. Our BPI average in those 6 games is 94.8.

First place UNC has an overall BPI of 88.8.

What that tells me is that when we play in big games, we are as good as anyone in the country. It's a lack of focus against inferior teams that has cost us.

 

[JerseyCat84]

Miami and SMU ranked above Kansas tells me all I need to know about the BPI.

 

[Smashcat]

Here's another interesting factoid in regard to BPI.

UK is 5 - 1 vs. BPI top 50 teams. Our BPI average in those 6 games is 94.8.

First place UNC has an overall BPI of 88.8.

What that tells me is that when we play in big games, we are as good as anyone in the country. It's a lack of focus against inferior teams that has cost us.

5-1 in Sagarin top 50 games too.

 

[Aike]

5-1 in Sagarin top 50 games too.

Yeah, the 5-1 is great, but that 94.8 BPI score in those games is very impressive. Bodes well for March, unless we stupidly overlook someone early.

 

[CatFanInTheBathtub]

when we play in big games, we are as good as anyone in the country. It's a lack of focus against inferior teams that has cost us.

That's possibly part of it. However it's no coincidence that 4 of our 5 losses to supposed "inferior teams" were on the road. We haven't had any problem focusing on those teams in Rupp.

 

[Aike]

That's possibly part of it. However it's no coincidence that 4 of our 5 losses to supposed "inferior teams" were on the road. We haven't had any problem focusing on those teams in Rupp.

Well, yeah. It's hard to win on the road.

 

[Teachable Moe]

Someone mentioned in another thread that we are only 18th in BPI. Even as we climb in Kenpom, Sagarin, and every other measure, we have basically stayed put in BPI.

I suspected that I knew what the flaw was, and did a little research to discover the truth.

BPI gives less weight to games where key players are missing. This sounds like a good and fair thing. But what if that key player is Alex Poythress, and you've actually played better without him?

UK's three best BPI scores have been in the past 3 games. Scores of 99.2, 99.4, and 99.5. Our 4th best score was way back in November, when we scored 99.1 against Duke.

Because Poythress has missed these past three games, they have only counted as 7/8 of a game each.

ESPN should have a mechanism to give you full credit when you perform better than usual despite a key injury. If we had been given full credit in the model for these past 3 games, we would be up around 13th by my math. That would make a lot more sense.

That seems backward. If you're at a temporary disadvantage, losses should count less, wins should count more.

 

[Aike]

That seems backward. If you're at a temporary disadvantage, losses should count less, wins should count more.

Their model just weights the game less if key players were missing. Effectively, you are penalized if you play well without key players.

 

[Teachable Moe]

Their model just weights the game less if key players were missing. Effectively, you are penalized if you play well without key players.

That makes it seem like they just threw together arbitrary differences just so they could have a model of their own. It's like Big K Cola.

 

[CatFanInTheBathtub]

Finally looked at those rankings. Took me about 10 seconds to determine how crappy they are. KU #12 and OK #2 is exhibit (A), seeing as how KU has a better record and 2 wins against the Sooners.

My question though as it relates to this thread is how do they determine a "top player". Is it a starter ? a guy who averages top 5 in MPG or PPG ? something else ? Does seem very arbitrary.

 

[Aike]

That makes it seem like they just threw together arbitrary differences just so they could have a model of their own. It's like Big K Cola.

Lol. Their basic premise is that games shouldn't count as much if key players are out. Unfortunately, that's way too simple.

 

[Aike]

Finally looked at those rankings. Took me about 10 seconds to determine how crappy they are. KU #12 and OK #2 is exhibit (A), seeing as how KU has a better record and 2 wins against the Sooners.

My question though as it relates to this thread is how do they determine a "top player". Is it a starter ? a guy who averages top 5 in MPG or PPG ? something else ? Does seem very arbitrary.

It's based on minutes played.