KenPom 2015-2016: Back on Top: #1 in Luck (1/11)

[Deacsfan27]

So "lucky" teams win close games and lose by being blown out? Is that correct?

Yes, because close games are more often determined by instances that defy the offensive and defensive efficiencies of the database.

It is basically shrinking down the sample size to 1-10 possessions instead of hundreds to thousands that have been charted for the year/recent past.

 

[deacvision7]

They were the 28th most experienced team last year, and the 237th most experienced this year. That may explain Harvard in this instance (as well as the fact that they have some freshmen on their team playing serious minutes).

How does that explain Wake though? We are ranked 320th in experience and were 304th last year, and ranked 241st in luck.

I actually have no idea if there is a relationship between luck and experience. There may be a brief positive correlation (the older you are, the luckier you are), but I doubt it.

This season is far too limited from a sample size perspective to make any rational conclusions from the data. I would be interested to seeing this "luck" number in March, however.

 

[Deacsfan27]

So they wouldn't be a dominant UNLV circa 1991 (until Duke) team. I agree. But I think that hypothetical team would be more skilled than lucky. Otherwise, they wouldn't have been good enough to have been in close games versus every team they played in the first place.

That's assuming they are playing good teams all the time.

Wake has been in a 5 point or less game with the following KP ranks: 326, 25, 44, 250, 111. That doesn't mean that we are good just because we played them all close.

 

[deacvision7]

Yes, because close games are more often determined by instances that defy the offensive and defensive efficiencies of the database.

It is basically shrinking down the sample size to 1-10 possessions instead of hundreds to thousands that have been charted for the year/recent past.

So I would just argue that hypothetical team played above their means a few times for their close wins but really showed their true capabilities when they got beat handily. It's sports. Weird shit like that happens all the time.


If anybody really thinks a system like KenPom should bat 100% on its predictions then that's just foolish. You cannot predict sports with anywhere near perfect accuracy.
ETA ^^^ is basically me saying I don't think there is any real need for this "luck" analysis if what you guys are saying is true and they are just trying to explain why their ranking system isn't perfect.

 

[deacvision7]

That's assuming they are playing teams all the time.

Wake has been in a 5 point or less game with the following KP ranks: 326, 25, 44, 250, 111. That doesn't mean that we are good just because we played them all close.

please see below

This season is far too limited from a sample size perspective to make any rational conclusions from the data.

 

[wsc8521s]

Last year 12 teams in the KenPom final top 50 finished in the top 100 in luck, or one less than the number of teams ranked 301-351st in the KenPom final rankings last year (13).

Sure, but it's self perpetuating -- if you blow teams out Kenpom will like you, if not it won't and it will be harder to finish in the top 50

 

[Deacsfan27]

This season is far too limited from a sample size perspective to make any rational conclusions from the data. I would be interested to seeing this "luck" number in March, however.

Sure, I mean I can find (literally) hundreds of data points for you to look at that indicate that there probably isn't a strong correlation. A year isn't even really a good sample size.

Here is last years top ten (KP ranking, experience ranking):

1. Texas Southern (202, 14)
2. Maryland (32, 228)
3. Wofford (86, 154)
4. Oregon (38, 267)
5. Eastern Illinois (216, 76)
6. North Dakota State (138, 282)
7. High Point (126, 95)
8. James Madison (203, 317)
9. Chattanooga (158, 241)
10. Tulsa (84, 103)

 

[Deacsfan27]

So I would just argue that hypothetical team played above their means a few times for their close wins but really showed their true capabilities when they got beat handily. It's sports. Weird shit like that happens all the time.


If anybody really thinks a system like KenPom should bat 100% on its predictions then that's just foolish. You cannot predict sports with anywhere near perfect accuracy.
ETA ^^^ is basically me saying I don't think there is any real need for this "luck" analysis if what you guys are saying is true and they are just trying to explain why their ranking system isn't perfect.

Nobody thinks that it should bat 100%. It should bat the percentage that it awards to win. If he gives Wake an 80% chance to beat UNC-G, in theory, Wake should win 80% of their games against UNC-G. Doesn't mean that there isn't one game that Wake will lose by 20, and it could be next Tuesday.

"Luck" just takes into account some teams win more close games than they should, and some lose more than they should.

 

[Deacsfan27]

Sure, but it's self perpetuating -- if you blow teams out Kenpom will like you, if not it won't and it will be harder to finish in the top 50

If you blow teams out that means you have a really high offensive efficiency and a really low defensive efficiency, which will likely rank you in the top 50 of KenPom in his rankings, and somewhere around average for luck.

Don't really understand your "self perpetuating" point.

My point is that luck has nothing to do with how good a team is. Good teams, bad teams, great teams, awful teams---they all win some close games and they all lose some close games.

You don't get any additional points on KP for "winning" or "losing" a game except for the change in what your offensive and defensive efficiency rating would be if you scored one more, or one less point in the game.

I don't know how many games we would have to look at for all teams in Division 1 to get to a "smooth" sample size, but I bet the team that is the best in the past 25 years in games decided by 5 points or less is some random-ass team, and not a blue-blood like Duke or Kentucky.

 

[wsc8521s]

If you blow teams out that means you have a really high offensive efficiency and a really low defensive efficiency, which will likely rank you in the top 50 of KenPom in his rankings, and somewhere around average for luck.

Exactly. That's why I'm not sure what pointing out that the top 50 is average in luck adds. Kenpom is going to like those teams anyway because they don't play a lot of close games

 

[Deacsfan27]

Exactly. That's why I'm not sure what pointing out that the top 50 is average in luck adds. Kenpom is going to like those teams anyway because they don't play a lot of close games

Well DV7 said that being lucky was actually a skill, so it should reason that the top 50 teams in KenPom would rank highly in "luck" too because they are ranked higher overall than those that aren't as skilled, and therefore less "lucky".

 

[Deacsfan27]

Anyway, there is no consistent rhyme or reason as to why a team wins all of their close games, while another team loses all of their close games. There are certainly reasons for specific instances (age, experience, skill), but at the end of the day, sometimes weird, fluky shit happens that causes the better team overall to lose a game to an inferior opponent.

 

[wsc8521s]

Well DV7 said that being lucky was actually a skill, so it should reason that the top 50 teams in KenPom would rank highly in "luck" too because they are ranked higher overall than those that aren't as skilled, and therefore less "lucky".

Not necessarily because relatively speaking on Kenpom teams will get punished for those close games (or rewarded for losing close games). That's my point.

Was Texas really better than Maryland last year when the Terps knew how to win close ones and the Longhorns didn't? I have my doubts. That's part of basketball

 

[Deacsfan27]

Not necessarily because relatively speaking on Kenpom teams will get punished for those close games (or rewarded for losing close games). That's my point.

Was Texas really better than Maryland last year when one team knew how to win close ones and another didn't? I have my doubts

I just don't buy into the notion at all that teams "know how to win close games" or are "clutch".

They are just cliche buzzwords that don't have any statistical truth to them other than an invalid sample size.

 

[Fuck Your Couch]

Nothing is 100%. It is a fact that if you win by 1 point you easily could have lost. Whether it's a last second shot from the other team that misses or a last second shot that you had to hit or a FT you had to make or a FT they had to miss. None of those situations are close to completely controllable.

I feel like everyone gets this but have just staked out their position and now are unwilling to budge.

 

[wsc8521s]

I think doofus is probably right -- he should just use something besides Kenpom to justify Kenpom.

I do absolutely think winning or losing close games can be contagious and get in the heads of players within a single season. Whether that makes them "good" I'm not so sure

 

[Atlanta Deacon]

I think we are actually a team that doesn't know how to close games. We consistently give away leads. We've been "lucky" that the game isn't 41 minutes long.

 

[FckVwls]

Maryland was "screwed" last year getting seeded as a #4 according to the experts. Most everyone had them as a #2 or #3 based on resume. And even given that they were 3 point underdogs in the 4/5 game against West Virginia.

That's a pretty good visualization of a high KenPom luck team.

 

[SpringChicken]

Dang it I missed the part where this argument makes sense

 

[ACCOffice]

I finally get it. Give Kenpom 20 years of data with no variables, which include all teams remaining the same and the ability of the team members staying the same, and it will have a 50% chance of predicting results.

Except for luck.