Sagarin, Pomeroy, Lunardi, ESPN BPI Updates 2/4/2013

[GatorLurker]

Originally Posted by Matthanuf06

I can guarantee its mean reversion. That doesn't mean we aren't the best team.

I am not so sure.

I think that he sees a team putting up ungodly numbers and when he looks at the personnel he doesn't see 3 or 4 lottery picks.

And I don't either.

But what I do see is an amazing team playing hoops at an amazing level.

[GatorLurker]

And there is probably some flaw in his algorithm. There is something about the way Billy D's teams play that just crushes so-so teams and that probably inflates their "scores" with kenpom and Sagarin Predictor.

[Osiris_DPM]

Originally Posted by GatorLurker

I am not so sure.

I think that he sees a team putting up ungodly numbers and when he looks at the personnel he doesn't see 3 or 4 lottery picks.

And I don't either.

But what I do see is an amazing team playing hoops at an amazing level.

Pomeroy had a good explanation for this, and it had nothing to do with the eye test or lottery picks. It had something to do with the numbers exceeding historical norms over numerous categories, and in the instances where this has occurred in the past, certain factors that cannot be predicted by these metrics (the human element) inevitably resulted in scenarios where the predictive effect based on these algorithims was inapplicable.

In short, he thinks Florida is the best team, but that because their adjusted metrics are so far outside the normal distribution even historically, they are likely to encounter a scenario where a variety of factors come into play that cause them to lose a game despite these metrics predicting otherwise.

You should also read his blurb in defense of Florida losing to Arizona:

1) Arizona 65, Florida 64 [59]. (Saturday) ‘Twas a low-scoring game, but not necessarily a defensive battle since there were only 59 possessions. For the AP voters that are going to put Arizona above Florida this week (which is almost all of you, I’m guessing), it’s worth reading about the head-to-head fallacy. You’re judgment is basically the opposite of history and you’re only defense is that the AP made you sign an oath stating “I don’t believe in home-court advantage”.

Here's the article on the "head to head fallacy" he references

http://kenpom.com/blog/index.php/web...o-head_fallacy

If you read a lot of his rationale in categorizing creating these algorithims, he actually makes a lot of sense, unlike the talking heads at ESPN, and he even said that if Florida loses a game, it does not change their status as the top team based on these metrics, just falls in line with similar historical trends based on unquantifiable factors.

I like his attempt to quantify the unquantifiable in his luck category. That factors in those intangibles like having a clutch go-to player, an automatic FT shooter, experienced players who do not panic, etc, and attempts to assign a statistical value to it. There is soooo much more that influences those "unpredictable" outcomes, but they are on the right track

[GatorLurker]

Originally Posted by Osiris_DPM

Pomeroy had a good explanation for this, and it had nothing to do with the eye test or lottery picks. It had something to do with the numbers exceeding historical norms over numerous categories, and in the instances where this has occurred in the past, certain factors that cannot be predicted by these metrics (the human element) inevitably resulted in scenarios where the predictive effect based on these algorithims was inapplicable.

In short, he thinks Florida is the best team, but that because their adjusted metrics are so far outside the normal distribution even historically, they are likely to encounter a scenario where a variety of factors come into play that cause them to lose a game despite these metrics predicting otherwise.

You should also read his blurb in defense of Florida losing to Arizona:



Here's the article on the "head to head fallacy" he references

http://kenpom.com/blog/index.php/web...o-head_fallacy

He actually makes a lot of sense, unlike the talking heads at ESPN, and he even said that if Florida loses a game, it does not change their status as the top team based on these metrics, just falls in line with similar historical trends based on unquantifiable factors.

I have read all of that before, but it doesn't really answer the question. In his algorithm this year's Gator team is above historical norms and so he says "BS". I say "they are". So, is his rating system flawed?

I have been vilified on NBN for saying that this year's team is better than either the 2006 or 2007 NC teams, but I still hold that belief. They are almost as good on O and better on D. It is a joy to watch.

Winning the NC is not verification of being the best team that year, BTW. Villanova vs Georgetown is the obvious example. Duke over UNLV in the 1991 semi-finals is another.

[gatorbogey]

i agree, these are legendary numbers;
i'd like to get some hardware first though before anointing this team. still, we're showing right now we're a pretty darn good team. metrics are in the legendary category.

[tommyuf21]

Originally Posted by GatorLurker

Yes.

Nerlens Noel will disrupt our O. That is a big problem. He pretty much erases the low post game unless we have refs that will make calls against him. That is not a given.

And for some reason opposing players rush shots against NN instead of first faking. That ain't working. It is time to try something new.

Our D will be fine. UK is actually easy to guard if you have the one-on-one defenders that Florida has.

If we are shooting OK from 3 we will be fine, but if the team takes an off night it could be a nail biter.

The nice thing about playing UK later is that you get to see what everyone attempts to do against them that doesn't work, especially when it comes to attacking Nerlens.

On the other hand, they are getting better and will present a much tougher challenge in a week than they would have last month.

However, I don't see our guards driving on Noel one on one without some kind of other off ball action going on. If you go back to last year, Patric and Murphy had good games against some outstanding UK bigs and Davis and MKG are gone. The rest of our team is much better and obviously the Cat's are not as good.

I can see a split, with each team defending their home turf. I know that would be considered a failure by some here, but after what has happened over the past three years, I'd take it.

[qwghlmgator]

Originally Posted by GatorLurker

I have read all of that before, but it doesn't really answer the question. In his algorithm this year's Gator team is above historical norms and so he says "BS". I say "they are". So, is his rating system flawed?

I have been vilified on NBN for saying that this year's team is better than either the 2006 or 2007 NC teams, but I still hold that belief. They are almost as good on O and better on D. It is a joy to watch.

Winning the NC is not verification of being the best team that year, BTW. Villanova vs Georgetown is the obvious example. Duke over UNLV in the 1991 semi-finals is another.

He's saying they are exogenous variables that are either unknow or unquatifiable - and hence not included in his model. This is not a flaw and is true of virtually any predictive statistical model ever devised.

Or, more prosaically, he's saying it's a statistical model not a crystal ball.

[socraticsilence]

Originally Posted by GatorLurker

I have read all of that before, but it doesn't really answer the question. In his algorithm this year's Gator team is above historical norms and so he says "BS". I say "they are". So, is his rating system flawed?

I have been vilified on NBN for saying that this year's team is better than either the 2006 or 2007 NC teams, but I still hold that belief. They are almost as good on O and better on D. It is a joy to watch.

Winning the NC is not verification of being the best team that year, BTW. Villanova vs Georgetown is the obvious example. Duke over UNLV in the 1991 semi-finals is another.

See, I don't get how you think this team is better than 2007- 2006, okay we came together late- I could see this years team taking the 2007 team in the regular season but c'mon in a game that mattered? How?

[SmootyGator]

Originally Posted by socraticsilence

See, I don't get how you think this team is better than 2007- 2006, okay we came together late- I could see this years team taking the 2007 team in the regular season but c'mon in a game that mattered? How?

I think the more important questions is: Could this team beat themselves?

[GatorLurker]

Well, Ken wrote me back and said that this year's UF team is putting up numbers in his system that are "greatest of all time" like numbers. He thinks that somehow their ranking is inflated, but gave me no explanation of why that is.

If I created a model and it was giving me unexpected numbers I would be looking for the reason.

He also said that even if Florida's numbers were shrunk a pretty fair amount they would still come out as the overall #1.

Then I was a bit surprised that he thought that UF and Indiana were about even. I guess that he doesn't believe his own model.

My view is that this UF team has been playing D in a legendary manner coupled with a very efficient O. How often does that happen?

I understand the "love" shown to the Championship teams on NBN. It was a fantastic ride and they are awesome Gators. And the 2007 team was probably the team of the decade. But this team is such a joy to watch because they play the game almost to perfection. They are not as physically talented as the NC teams, but their execution on O and D is a marvel. I just hope that they keep it together for a little more than 2 more months.