Need some help making a stat

Hendrix · Post #1 » by **Hendrix** » Thu Nov 25, 2010 12:44 am

I was thinking about making a stat, but I'm having difficulty getting from the idea to having a working stat.

Basically I want to use adjusted +-, and convert that into a ranking out of 82 (games a year). So for example if you're adjusted +- was = to the mean adjusted +- of the league, then you would have a ranking of 41 wins. Or for example if you had 5 players playing 48mpg, with a 35 win ranking then your team would end up with 35 pyth wins, or if your team ended up with 48 wins, and "player x" had a rating of 30 wins then he was hurting the team. etc.. etc...

Also I was thinking if you took

(players minutes played/minutes available to play) X Players win rankings = Stat "Y"

Then you could add up all the "stat Y's" on your team and hopefully it should equal out to somewhere around your teams pythagorean W-L.

However I did a quick attempt at this, and the Raptors sums of "stat Y" = 33 wins. However their pyth w-l was 36 wins.

Anyways, I think the problem I'm having is with converting the adjusted +- #'s properly into a # out of 82. What I originally did was find the range of highest and lowest adjusted +- #'s. (Wade=18.644, Boykins= -13.4713, Range= 32.1153). Then calculate where "player x's" adjusted +- resides in that range as a proportion. So Bosh's adjusted +- was 5.738. Ranking him @ .598 in the range. Multiply that by 82 games, and his value = 49.04 wins

But I think this is wrong, and there's probably a better way to do it. But I don't know what it is. That would probably work if the range was normally distributed, but I'm 3 wins off still so it seems like it isn't, and there's possibly some outliers or something throwing things off. For example if 1 player had a +20 rating. And everone else in the league was between -10 and 10, then a player with a 10 value (2nd in the league) would get .6666 wins X 82=54.667 wins. If the range, and values where the same however the players were normally distributed with a whole bunch with values higher then 10 then that player would still get a 54.667 wins. Not right.

So basically I was wondering if anyone had an idea how I would get a more accurate assesment of what the % is to multimpy by 82.

mysticbb · Post #2 » by **mysticbb** » Thu Nov 25, 2010 4:14 am

I have the impression that you don't really understand APM in the first place. A +5.7 in APM means that a player is 5.7 points better per 100 possessions than an artifical average player in the league. The average APM for the league is 0.

Using Bosh's number you can just devide the 5.7 by 2 and get a + 5.7/2 for the ORtg and a -5.7/2 for the DRtg. Last season the average ORtg/DRtg was 107.6. Thus you would get 110.5 as Bosh's ORtg and 104.8 for his DRtg. Now use the pythagorean formula (exponent 16.5) and you will get 0.706 Win% or 57.9 Wins.
If you want to add those numbers up for a team, you need to adjust that by minutes played, after that devide that by the total minutes and you should get pretty close to the pythagorean expectation.

Well, I checked it for the Raptors by using the APM from basketballvalue.com and I get 39.2 Wins, which is pretty close. The difference might be due to the fact that the minutes played for a couple of players wasn't enough to have a APM value for them. It is also a fact that APM has a pretty big error with about +/- 4 for each player. Bosh, btw., had +7.0 last season which equals to 61 wins (if you would have all players with +7 on your team).

Chicago76 · Post #3 » by **Chicago76** » Thu Nov 25, 2010 5:05 am

mystic nailed it.

Some of the deviation betwen mystic's computation and the actual team pythag wins could be due to simple rounding. In the Bosh example, his Ortg and Drtg rounded to one decimal each can support a win total of 57.6 to 58.1 wins (pre adjustment for player minute/total team minutes).

Hendrix · Post #4 » by **Hendrix** » Thu Nov 25, 2010 5:24 am

Lol, that'll work too. Thank mysticbb

mysticbb · Post #5 » by **mysticbb** » Thu Nov 25, 2010 5:38 am

Chicago76 wrote:Some of the deviation betwen mystic's computation and the actual team pythag wins could be due to simple rounding. In the Bosh example, his Ortg and Drtg rounded to one decimal each can support a win total of 57.6 to 58.1 wins (pre adjustment for player minute/total team minutes).

Well, you will not get any more informations out of it by using more "accurate" numbers. The error is large, thus using more than one decimal is pretty much useless.

@Hendrix

Try to make stuff like that as simple as it gets.