KenPom

[Jerry1714]

Another update. Taking my cue from Jerry and limiting it to the Top 40 in both offense and defense, here's the latest list (average number, followed by offense and defense numbers, then current AP poll position):

1. Louisville 11 (15/7); AP No. 7
2. Florida 11.5 (13/10); 1
3. Wichita State 12 (10/14); 2
4. Villanova 13.5 (7/20); 8
5. Arizona 15 (29/1); 3
6. Syracuse 15.5 (22/9); 4
7. Kansas 16 (5/27); 5
8. Michigan State 24 (14/34); 18
9. Kentucky 24.5 (18/31); 17
10. Pittsburgh 25 (27/23); NR
11. North Carolina 28.5 (38/19) 19
12. Iowa State 33 (28/38); 15

Down to only 12 teams. Four of these teams are in the Top 20 in both:

Louisville
Florida
Wichita State
Villanova

Using the metric identified by champs99and04 (Top 40 in Offense, Top 21 in defense), the list expands to seven teams:

Louisville 15/7
Florida 13/10
Wichita State 10/14
Villanova 17/20
Arizona 29/1
Syracuse 22/9
North Carolina 38/19

The champion would probably be coming from this list, if the season ended today. Notably, Virginia just barely missed the cut, as they have the No. 3 defense, and the No. 41 offense

UConn has the 22nd ranked defense, but the team's offense has sunk down to 57. So while the defense has even improved a little bit since a few weeks ago, it's the offense that has been letting the Huskies down.

No team that was not in the Top 21 in BOTH offense and defensive has won the title.

 

[Karmacop]

No team that was not in the Top 21 in BOTH offense and defensive has won the title.

Only four teams currently qualify:

Louisville (15/7)
Florida (13/10)
Wichita State (10/14)
Villanova (7/20)

Syracuse (22/9) is one good offensive game away from joining that list.

 

[tcf15]

UConn's offense and defense ratings are going in opposite directions right now. Offense down to 76 and defense up to 11.

 

[Tenspro2002]

UConn's offense and defense ratings are going in opposite directions right now. Offense down to 76 and defense up to 11.

Didn't need fancy computer numbers to tell us that.

We won today entirely because of our defense; offense has been lousy for 2 weeks. But that formula worked in 2011, so maybe it'll be good enough this year.

 

[Karmacop]

A team in the top 10 in both Offensive and Defensive Efficiency would be a favorite to win it all.

There is one team that qualifies now -- Florida. They are now the clear title favorites, using this metric.

 

[Waquoit]

Can't get past the fact that Ken Pom had UConn at #9 AFTER they won 11 in a row (7 against ranked teams) and the title. He couldn't pick the winner after the race was run.

 

[Leebo]

Can't get past the fact that Ken Pom had UConn at #9 AFTER they won 11 in a row (7 against ranked teams) and the title. He couldn't pick the winner after the race was run.

You do understand how the stats work, right? Because if you do, it's obvious that there are going to be conflicts between statistical performance and actual results.

It's predictive not descriptive. As in, the day after the the championship game, you scheduled a game between UConn and Ohio St, the team that finished #1 in kenpom, UConn would be the underdog. The kenpom ratings are for predicting the odds of future games, not describing the outcomes of past games.

 

[nelsonmuntz]

Good thread.

The problem with this analysis is there are so many bad teams on the top teams' schedules that it is hard to get an accurate assessment of efficiency. Louisville is the most extreme example. Pitino has been running up the score on bad teams all year but only has 5 wins against Top 100 competition. Louisville has the third worst SOS of any team in the Top 35 RPI (Wichita and SDSU have worse, Gonzaga's is better than Louisville's). Louisville beat Rutgers by 48 points, I am sure that game did wonders for the Ville's efficiency numbers.

Florida's numbers are legit. The SEC sucks, but there are still a few NIT caliber teams in that league in addition to Kentucky and Florida played a solid non-conference schedule. I would agree that they look like the favorites right now.

 

[Jerry1714]

Good thread.

The problem with this analysis is there are so many bad teams on the top teams' schedules that it is hard to get an accurate assessment of efficiency. Louisville is the most extreme example. Pitino has been running up the score on bad teams all year but only has 5 wins against Top 100 competition. Louisville has the third worst SOS of any team in the Top 35 RPI (Wichita and SDSU have worse, Gonzaga's is better than Louisville's). Louisville beat Rutgers by 48 points, I am sure that game did wonders for the Ville's efficiency numbers.

Florida's numbers are legit. The SEC sucks, but there are still a few NIT caliber teams in that league in addition to Kentucky and Florida played a solid non-conference schedule. I would agree that they look like the favorites right now.

SOS is taken into effect on the numbers as are factors like "luck". There was a point where we were one of the luckiest teams in country. Now that belongs to our new friend Tulane.

 

[Scot-e]

I know Tulane sucks but I am really looking forward to going there for a game- that city is awesome!

 

[Waquoit]

You do understand how the stats work, right? Because if you do, it's obvious that there are going to be conflicts between statistical performance and actual results.

It's predictive not descriptive.

Predictive? That post-season KenPom would have been wrong about UConn beating Pitt, Ville, Cuse, SDSU, Zona and Kentucky at a minimum. That's less than 50%. OK, then.

 

[Jerry1714]

Predictive? That post-season KenPom would have been wrong about UConn beating Pitt, Ville, Cuse, SDSU, Zona and Kentucky at a minimum. That's less than 50%. OK, then.

No, that is not accurate. Going into the game the Kenpom predictions had UConn beating Zona, Ville and Cuse. So it was actually at 70% in pregame predictions through the post season, much better than vegas or any analyst.

I do not think you understand how it works. Looking at the team ranking is a very poor way to use KenPom, just because one team is rated higher does not mean they predict them to win the game.

If you want it to be 100% good luck finding that, you will be a very rich man.

 

[Waquoit]

No, that is not accurate. Going into the game the Kenpom predictions had UConn beating Zona, Ville and Cuse. So it was actually at 70% in pregame predictions through the post season, much better than vegas or any analyst.

I do not think you understand how it works. Looking at the team ranking is a very poor way to use KenPom, just because one team is rated higher does not mean they predict them to win the game.

If you want it to be 100% good luck finding that, you will be a very rich man.

That's interesting and will make me rethink to some extent. I'm not afraid of numbers, I still have a couple of the original Baseball Abstracts. And I like to go to the track where numbers are big. No, I'm not totally aware of Ken Pom's methodology. But in my brief exposures I've found him to be of the "fun with number's" variety.

 

[Leebo]

Predictive? That post-season KenPom would have been wrong about UConn beating Pitt, Ville, Cuse, SDSU, Zona and Kentucky at a minimum. That's less than 50%. OK, then.

I thought my example would make it clear. If UConn played Ohio State the after the championship game, Ohio State would be favored. Could UConn still win? Obviously. But the site uses statistics to give you the odds when two teams match up, not a description of how they fared in wins and losses directly.

Also, imagine this... #1 in kenpom faces #100 in kenpom. #100 wins.

One week later #1 is now #5 and #100 is now #80.

Who would be more likely to win the rematch? It's still the higher ranked team, even though they lost. That previous game doesn't actually impact any future game.

One more thing... it's not actually possible for a site like kenpom to be "wrong" with regard to any individual game, unless the site says one team is expected to win 100% of the time. The site gives percentage chances, not absolute predictions.

 

[Waquoit]

One more thing... it's not actually possible for a site like kenpom to be "wrong" with regard to any individual game, unless the site says one team is expected to win 100% of the time. The site gives percentage chances, not absolute predictions.

It's pretty great when you can never be wrong. Let's just say you have greater faith in his numbers than I do.

 

[Leebo]

It's pretty great when you can never be wrong. Let's just say you have greater faith in his numbers than I do.

I don't know what you mean by that. You seem to be misunderstanding me still.

If I have a 6 sided die, and I tell you "There's a 5/6 chance I'll roll higher than a 1", am I "wrong" if I roll a 1? No. What I gave you were the odds. Not a prediction.

That's what kenpom is doing.

 

[Walker11]

I don't know what you mean by that. You seem to be misunderstanding me still. If I have a 6 sided die, and I tell you "There's a 5/6 chance I'll roll higher than a 1", am I "wrong" if I roll a 1? No. What I gave you were the odds. Not a prediction. That's what kenpom is doing.

I think it's obvious that he is giving probabilities. The possible issue is with the probabilities, and with the stats which are the foundation for the probabilities.

 

[Leebo]

I think it's obvious that he is giving probabilities. The possible issue is with the probabilities, and with the stats which are the foundation for the probabilities.

I'm not sure it's obvious, because Waquoit has been describing it as though kenpom is "wrong" for ranking teams UConn beat above UConn. When that happens, it means that UConn was an underdog statistically when they won, and they still are.

 

[Jerry1714]

I think it's obvious that he is giving probabilities. The possible issue is with the probabilities, and with the stats which are the foundation for the probabilities.

If the numbers are so funny, then why have they been historically been so accurate?

No one is saying the #1 team is the best team, neither does he, but if you look at the efficiency ratings through history there is no denying how accurate they are at predicting success in the tournament.

 

[Walker11]

I like looking at his website and I think they are helpful. I just thought 2 people were arguing 2 different things but perhaps I'm incorrect.

I do think he often over ranks teams with very efficient offenses which perhaps don't hold up as well vs. elite competition.

 

[Leebo]

Well, to be fair, I'm not quite sure what Waquoit's specific concern is. But the fact that kenpom has UConn as #9 in 2011 doesn't bother me the way it bothers him apparently, simply because that #9 refers to performance in future games (which is odd to talk about once the season is done).

 

[Jerry1714]

I like looking at his website and I think they are helpful. I just thought 2 people were arguing 2 different things but perhaps I'm incorrect.

I do think he often over ranks teams with very efficient offenses which perhaps don't hold up as well vs. elite competition.

History says you are wrong:

http://www.burntorangenation.com/ba...istics-of-ncaa-champions-basketball-analytics

 

[Walker11]

History says you are wrong:

http://www.burntorangenation.com/ba...istics-of-ncaa-champions-basketball-analytics

Of course the champion will have an elite offense. Maybe I wasn't clear. My issue is with Duke being #4 with a 2/55 split and Creighton at #9 with a 1/102 split. These types of teams are often over ranked IMO. I think his system over ranks Duke almost every year.

 

[Jerry1714]

Of course the champion will have an elite offense. Maybe I wasn't clear. My issue is with Duke being #4 with a 2/55 split and Creighton at #9 with a 1/102 split. These types of teams are often over ranked IMO. I think his system over ranks Duke almost every year.

Gotcha, I misinterpreted what you meant.

 

[HuskyHawk]

Of course the champion will have an elite offense. Maybe I wasn't clear. My issue is with Duke being #4 with a 2/55 split and Creighton at #9 with a 1/102 split. These types of teams are often over ranked IMO. I think his system over ranks Duke almost every year.

I think the problem with teams that rely too heavily on offense, like Duke this year, is that they have less of a margin for error. A bad shooting night and they are done. I think this described UConn early this year as well, and since then, KO has changed the focus to defense, which is more reliable and less vulnerable to luck.

Great D is a prescription for winning...until you run into team with great D and great O.