CBB: KenPom question

[crimson23]

Hi everyone - longtime reader, first time poster.

Has anyone ever done a systematic comparison of KenPom's win percentages against Vegas moneylines? I am not talking about KenPom's projected point spreads. Instead, I'm referring to when KenPom projects a particular team has an X% chance of winning, if anyone has systematically compared that win percentage to the implied win percentage of Vegas MLs.

E.g., on 11/23, KenPom gave Loyola Marymount a 69% chance of beating Idaho St. That translates to a "breakeven" ML of -222.58. Vegas had a ML of -130 for Loyola, or an equivalent win percentage of 56.52%. This is a theoretical "edge" of 69% - 56.52% = 12.48%. I've spent the last few days pulling all of KP's projections into a spreadsheet and comparing them to Vegas MLs, and most of them are fairly close, but there are a few games each day that have had double digit differences in a team's projected/implied win percentage. Sample size is obviously too low at this stage to see if this theoretical edge is actual +EV, but I thought I'd ask if anyone has done any research into this.

I'm not looking to debate the underlying assumption of whether KP is better at calculating win percentages than Vegas -- I'm not blind to the reasons why he wouldn't be. But for now, just asking to see if anyone has looked at this angle, and maybe see if we could compare notes. Or, even better, whether there's an easy way to test this hypothesis against past data.

Thx!

[Romanov]

Hi everyone - longtime reader, first time poster. Has anyone ever done a systematic comparison of KenPom's win percentages against Vegas moneylines? I am not talking about KenPom's projected point spreads. Instead, I'm referring to when KenPom projects a particular team has an X% chance of winning, if anyone has systematically compared that win percentage to the implied win percentage of Vegas MLs. E.g., on 11/23, KenPom gave Loyola Marymount a 69% chance of beating Idaho St. That translates to a "breakeven" ML of -222.58. Vegas had a ML of -130 for Loyola, or an equivalent win percentage of 56.52%. This is a theoretical "edge" of 69% - 56.52% = 12.48%. I've spent the last few days pulling all of KP's projections into a spreadsheet and comparing them to Vegas MLs, and most of them are fairly close, but there are a few games each day that have had double digit differences in a team's projected/implied win percentage. Sample size is obviously too low at this stage to see if this theoretical edge is actual +EV, but I thought I'd ask if anyone has done any research into this. I'm not looking to debate the underlying assumption of whether KP is better at calculating win percentages than Vegas -- I'm not blind to the reasons why he wouldn't be. But for now, just asking to see if anyone has looked at this angle, and maybe see if we could compare notes. Or, even better, whether there's an easy way to test this hypothesis against past data. Thx!

No, but last year I did something similar and it won. Could've been variance:

Whenever Pinny and matchbook favored a team by 5% more than ken pom I bet the team on the moneyline. Only bet teams in major conferences. The little guys were killers (not enough info in market maybe?) Ended up 30 or 40 units, just casually doing it whenever I wanted action.

Honestly this doesn't belong in the think tank but w/e

[crimson23]

Appreciate the feedback. Apologies if I posted this in the wrong forum -- my sense of the NCAA forum was that it was more about individual games/matchups rather than system-based discussion. Sorry, I'm new

[Romanov]

Appreciate the feedback. Apologies if I posted this in the wrong forum -- my sense of the NCAA forum was that it was more about individual games/matchups rather than system-based discussion. Sorry, I'm new

Systems are usually avoided here. System usually fail.

[Pokerjoe]

A system/angle/trend is a data-mined pattern, devoid of a number useful for comparing to a line. Does not belong here.

KenPom is not a system. It's a rating/model/projection, compared against the line in a hunt for value. Belongs here.

imo.

[RickySteve]

lulz, this thread obviously belongs here.

The market will chew you up if you bet these numbers.

[podonne]

A system/angle/trend is a data-mined pattern, devoid of a number useful for comparing to a line. Does not belong here.

What if the system/angle/trend includes the line as a parameter? Like, bet all home dogs at +200 or higher?

[byronbb]

Hi everyone - longtime reader, first time poster.

E.g., on 11/23, KenPom gave Loyola Marymount a 69% chance of beating Idaho St. That translates to a "breakeven" ML of -222.58. Vegas had a ML of -130 for Loyola, or an equivalent win percentage of 56.52%. This is a theoretical "edge" of 69% - 56.52% = 12.48%. I've spent the last few days pulling all of KP's projections into a spreadsheet and comparing them to Vegas MLs, and most of them are fairly close, but there are a few games each day that have had double digit differences in a team's projected/implied win percentage.

Thx!

I highly doubt you can get 12% edge on ML against Vegas books near close. The logical conclusion then is that there is information about the game which KenPom's projection does not incorporate.

[Pokerjoe]

What if the system/angle/trend includes the line as a parameter? Like, bet all home dogs at +200 or higher?

No, still no good. A rule "home dogs +200 or higher" doesn't help you determine if a dog of +250 should be +200.

But this is J7's forum, he's the judge on what belongs here.

[crimson23]

Yes, Pokerjoe is correct as far as what I am hoping to do -- i.e., look for differences between the Vegas ML and what the ML "should" be based on KP's independent model, in search for +EV. But in any event, my question was only if someone had previously embarked on a empirical analysis of the implied ML's of KP's model, and not whether the system in theory holds water. If the question is no, then no need to further clutter the forum. Thanks, and sorry!

[Justin7]

This thread is fine here.

[offshoregenius77]

Ken Pom doesnt incorporate injuries or back talent levels like the current pinny linesmakers do, therefore this "system/model" would probably take the short end of the stick in the long - run. Would be interesting to see however.

[durito]

lol at pinny linesmakers

[Joe Dogs]

So are the Kenpom ratings in anyway a useful tool for wagering?

[LT Profits]

Ken Pom doesnt incorporate injuries or back talent levels like the current pinny linesmakers do, therefore this "system/model" would probably take the short end of the stick in the long - run. Would be interesting to see however.

Are you sure? KenPom has player ratings, so don't team ratings go down to a player level? In other words, if a key player is out, isn't it reflected in that night's game prediction?

[TPowell]

I dont believe so LT, would be very interested to get a definitive answer though

[durito]

the team ratings are not done at a player level. the player stats on the site are just an added feature.

[LT Profits]

the team ratings are not done at a player level. the player stats on the site are just an added feature.

Ah, thanks Bud.