RealGM

Why is it that when figuring out efficiency, generally ALL stats are taken into account? In trying to figure out the general efficiency of a player, wouldn't you want to do like any other statistical survey and only go out one or two standard deviations to eliminate outliers?

Just wondering..

What do you mean? Are you suggesting that we discount career high games when calculating PER? It wouldn't change it that much.

No, probably more like stats in the 3rd, and possibly even second standard deviations for any given stat.

For figuring out how they HAVE performed, trying this would be absurd. For figuring out how they're LIKELY to perform, it would seem more useful. A 5th year career 20ppg scorer isn't likely to score either 3 points or 44 points. He's capable of either, but putting them outside that range gives a more accurate indication. It's done with case studies with regard to future expectations, why not sports stats for the same thing?

moofs wrote:No, probably more like stats in the 3rd, and possibly even second standard deviations for any given stat.

For figuring out how they HAVE performed, trying this would be absurd. For figuring out how they're LIKELY to perform, it would seem more useful. A 5th year career 20ppg scorer isn't likely to score either 3 points or 44 points. He's capable of either, but putting them outside that range gives a more accurate indication. It's done with case studies with regard to future expectations, why not sports stats for the same thing?

You're asking that question HERE?

You're too smart to be this stupid. Go find Larry Coon and buy him dinner.


Just don't forget to make your OOOO face when he gives the answer.

Assuming I read your response correctly, I did have a very small intended target audience (actually it was pretty much only KevC and maybe one or two others, but PMs are disabled, so..)

Moofs, your question is too advanced for me. Honestly, I'd like for someone to tell me how calculating efficiency works. Calling KevC...

I don't think it matters. They're outliers for a reason. One 81 point game isn't going to raise the average 5 points. I can't remember but I think it was four tenths of a point?


If you wanted to do the expected value, it's really easy.

Sum of % of score * score = E(x)


By percentage of score I mean chance that the player get that score. So for kobe it would be 1/1000 * 81 for his 81 point game.

moofs wrote:No, probably more like stats in the 3rd, and possibly even second standard deviations for any given stat.

For figuring out how they HAVE performed, trying this would be absurd. For figuring out how they're LIKELY to perform, it would seem more useful. A 5th year career 20ppg scorer isn't likely to score either 3 points or 44 points. He's capable of either, but putting them outside that range gives a more accurate indication. It's done with case studies with regard to future expectations, why not sports stats for the same thing?

Okay, then what are you considering an "outlier"? A single game? A season's worth of a fluky stat (like someone gets a very high TS% for a certain season compared to his career)? If you mean the former, a single game is too insignificant to ever count as an outlier. If you mean the latter, most projections DO take fluky stats into consideration. I know for a fact that Hollinger does this to project PER for the next season.

I don't understand any of this..

kevC wrote:Okay, then what are you considering an "outlier"? A single game? A season's worth of a fluky stat (like someone gets a very high TS% for a certain season compared to his career)? If you mean the former, a single game is too insignificant to ever count as an outlier. If you mean the latter, most projections DO take fluky stats into consideration. I know for a fact that Hollinger does this to project PER for the next season.

No, stuff in the second and third SDs (4.6%, 2.3% on each end of the scale), possibly even first, although that should end up being excessive.

http://en.wikipedia.org/wiki/Image:Stan ... iagram.svg

Maybe just third (which in the NBA would likely only be a game or two at most, so I'd expect you need the second as well.

This is where I wish I knew either more statisticians (none) or math-prone basketball fans About the only two people I know IRL right now that are into basketball are my dad and someone from Clutchfans, and neither of them know jack about stats. I failed it (calculus based) the first time through myself (made an A the second time because english speaking professors help you pass classes, who knew? it's still an evil, evil course)

I don't think Hollinger takes any of that into account.
((in fact I have serious doubts Hollinger understands this stuff either))

King Roosk wrote:Moofs, your question is too advanced for me. Honestly, I'd like for someone to tell me how calculating efficiency works. Calling KevC...

Just numbers plugged into an equation that Hollinger made up. Check his wikipedia page.

http://en.wikipedia.org/wiki/John_Hollinger

(Points x 1.0) + (FGM x 0.4) + (FGA x -0.7) + ((FTA-FTM) x -0.4) + (OREB x 0.7) + (DREB x 0.3) + (STL x 1.0) + (AST x 0.7) + (BLK x 0.7) + (PF x -0.4) + (TO x -1.0)

I've seen a lot of people shooting holes in his equations too, he just happens to work for a huge sports magazine, which helps him be taken as "more authoritative". (whether he actually should be or not I have absolutely no idea, I do know that I've written some long posts regarding TS and eFG and how accurate his equations are. Those two are easy to hit since he uses those fairly arbitrarily chosen constants)