NBA

[tomverve]

Goofy and bare bones as it might seem to you, it correctly classifies 96.5% of players with at least 400 NBA games as being in or out of the Hall. But maybe a 3.5% error rate and a perception of goofiness is sufficient reason to not put too much stock into it.

[bodiddley]

Weren't the co-efficients tinkered with to get those results?
Do you want to defend subtracting the most basic stats and accomplishments from height?
Unimpressive, but I suppose it is only trying to be a rough indicator.

[tomverve]

Weren't the co-efficients tinkered with to get those results?

http://www.basketball-reference.com/about/hof_prob.html

The author claims to have tried several models, and evidently this is the one that worked the best. The co-efficients weren't tinkered with, they were set by a regression method that maximized the predictive ability of the model on the tested data set, standard procedure for this sort of analysis.

Do you want to defend subtracting the most basic stats and accomplishments form height?

What am I defending? It doesn't seem particularly relevant to me how conceptually neat the model may seem, nor whether it matches my intuitions for who deserves to be in the Hall.

What matters is just that the model does a good job of characterizing which players actually do and do not have a good chance for the Hall (not whether they deserve it), based on the way players have been inducted into the Hall historically.

The procedure is irrelevant; if the method only classified players by height and weight and was able to correctly classify 96% of players as being in or out, that would be just as well. Whatever works works, and this method seems to work pretty well.

Unimpressive, but I suppose it is only trying to be a rough indicator.

I wouldn't put too much stock in Rick Barry's ability to hit a free throw. He only hit 90% for his career, and he had a really goofy form. Unimpressive, but I suppose he was only trying to be roughly accurate.

[bodiddley]

The co-efficients weren't tinkered with, they were set by a regression method that maximized the predictive ability of the model on the tested data set, standard procedure for this sort of analysis.

Layman's translation: tinkering.
Or, the weights for each factor were adjusted to get good results.  
Surprise, it gets good results ... mostly.
But the info used is too limited so that it also misses badly.

I wouldn't put too much stock in Rick Barry's ability to hit a free throw.
He only hit 90% for his career, and he had a really goofy form.

And neither does this formula, except in the broadest sense of points scored.

Actually, Barry was using the tried-and-true method.
This formula seems more to me like George McGinnis pumping them in at 90% with his one-handed form.

[tomverve]

The co-efficients weren't tinkered with, they were set by a regression method that maximized the predictive ability of the model on the tested data set, standard procedure for this sort of analysis.

Layman's translation: tinkering.

Layman's translation: you don't know what you're talking about.

Of course you give the model its best chance to succeed. What, they should assign the weights randomly? Or maybe based on how goofy they seem. If the model were trash, it wouldn't give good results no matter what weights were used. But it does, so it's not.

Bottom line is that you wouldn't put too much stock in a method that gets it right 96% of the time because you don't like its aesthetics. You spend a lot of time dissecting some of the 3.5% or so of the time the model is wrong and conclude that it's unimpressive. You might call that tinkering for dummies.

[bodiddley]

When you devise a cockeyed system to fit the results, you shouldn't be too impressed when ... the results are largely right.

It's Aristotelian.  Ptolemaic.
You would have been a big supporter of planetary epicycles, no doubt.

I would be surprised if that's the best HOF formula out there.
It's a good road ... except for that one trouble spot:

[tomverve]

Thanks for bringing up the epicycle analogy, as it clearly demonstrates your misunderstanding of the core issue here.

The epicycle model does a fine job of predicting where and when planets will be located in the night sky. If all you are concerned with is just this issue of predicting planet locations, then it doesn't matter one whit whether your model posits epicycles or not, as long as it does the predictive job well.

The utility of the HOF model is in its ability to predict who will and will not make the hall, and so should be judged on that basis alone. It may not accurately reflect the processes that govern how voters actually decide who gets into the Hall, just as the epicycle model does not accurately reflect the processes governing the changing locations of the planets in the sky. But who cares? That issue is irrelevant for the purpose at hand.

A final irony is that you are put off by the simplicity of the HOF model, but simplicity is a virtue for model fitting. When you add complexity you increase the likelihood of overfitting the model to the data. The very simplicity of the model is the best way to guard against the possibility of overfitting. Your critiques are at cross-purposes; add complexity, increase overfitting. Decrease overfitting, reduce complexity. Can't have it both ways.

[TrojanHorse]

I prefer my models to be at least 5'10"

[TrojanHorse]

and "simpler" is definitely better...most of them are way too complex as it is...

[bodiddley]

I'm hijacking this forum for some World Cup chatter:

US just dropped down 2-0 against Slovenia, after missing on a few nice chances.
I like that the US has a defender named Demerit.
I like that Slovenia has a Charlie Brown squiggle on the front of their jerseys.

Best teams I've seen:
Maradona's Argentina; runty Spain (in a loss), disciplined Ivory Coast (in a tie), Brasil. 

Only caught a few minutes of both games for Germany.
The 10-man loss to Serbia makes things interesting.
Kind of a tough break for Ghana, who would have just needed to beat Australia to go through if Germany beat Serbia.
Now Ghana v. Germany becomes a big later game.
And Serbia and Ghana need to rack up big goal scores v. Australia.

[CaptainCargo]

Nice game to watch last night.

I heard some Celt fans bitching about the foul difference in the fourth quarter. Was it the refs?






No.




The plain fact of the matter is that the Celts committed more fouls in the fourth quarter. What are the refs supposed to do? Call some phantom fouls on the Lakers to "even it up"??

Ya got beat Celt fans. Deal.

[bodiddley]

Nice 2nd half comeback for the US squad.
No clue why that 3rd goal was disallowed.
All I saw were three US players being held, one in a headlock.
Unfortunate.
Also, the one US defender will miss the next game after being carded for a handball that clearly went off his cheek.
Mostly good reffing from what I've seen, and most of the calls in this match were the right ones, but tow blown calls hurt the US.

[bankshot1]

Nice 2nd half comeback for the US squad.
No clue why that 3rd goal was disallowed.
All I saw were three US players being held, one in a headlock.
Unfortunate.
Also, the one US defender will miss the next game after being carded for a handball that clearly went off his cheek.
Mostly good reffing from what I've seen, and most of the calls in this match were the right ones, but tow blown calls hurt the US.

there was no foul on the dis-aallowed "3rd" US goal.

A ref can ref a game one way for for the first 85 minutes, or 3 quarters, and then make a call or calls, or call the game in a different manner, than they had, that decide game outcomes.
Damn shame too. 

[CaptainCargo]

LOL

[TrojanHorse]

I heard the ref was now "dis-allowed" ...from any further world cup competitions...

I guess some higher powers agreed that there was no foul.