This topic contains 11 replies, has 5 voices, and was last updated by aamir543 12 years, 1 month ago.
- AuthorPosts
- Posted on: Fri, 02/24/2012 - 7:34pm #36803
whateverParticipantThis is an statistical rating Ive been working on. It is based on different statistical data. Since it is completely based on stats it doesnt compete with any Rookie ladder to the ROY or anything like that. My objective is to give an alternative rating to indexes like the PER to evaluate rookies. I know is not perfect but the rank doesnt seem to be that off. It is a relative index so the player ranked number one will always have a rating of a 100. Also it gives the same weight to games played on opening night and games played yesterday. This means the rating is based on the players overall performance during the season, so a player having a good couple of weeks is not expected to jump a lot of positions, like it usually happens in the rookie ladders. Finally the ranking only includes players with at least 15 games played and averaging over 15 minutes during those games, so players like Enes Kanter or Leuer are not included. Hope you enjoy.
Rank Player Rating
1 Irving (Cle) 100
2 Rubio (Min) 99,37654
3 Brooks (NJ) 88,80776
4 Walker (Cha) 86,92405
5 Knight (Det) 86,8407
6 Shumpert (NYK) 85,1737
7 Leonard (SA) 85,007
8 Faried (Den) 84,97366
9 Parsons (Hou) 84,60692
10 Vucevic (Phi) 83,57338
11 Morris (Phx) 82,78989
12 Ayon (NOH) 81,45629
13 Harrelson (NYK) 81,07288
14 N. Cole (Mia) 81,0062
15 I. Thomas (Sac) 80,80616
16 L. Allen (Phi) 80,48943
17 D. Williams (Min) 79,58925
18 T. Thompson (Cle) 77,80556
19 K. Thompson (GSW) 76,35527
20 Fredette (Sac) 74,40488
21 Biyombo (Cha) 73,57138
22 Singleton (Was)} 71,27092
23 Vesely (Was) 69,43722
24 W. Russell (Det) 67,820230 - Posted on: Fri, 02/24/2012 - 7:48pm #641054
Mr. 19134ParticipantHow in the world did you get these numbers tho?
0 - Posted on: Fri, 02/24/2012 - 8:23pm #641069
whateverParticipantWell I used several other indexes plus other important stats. Some of the indexes I used were the efficiency rating, fantasy rating, PER. Also used stats like average pts, rebs, assists, blks and turnovers both per game and by 48 minutes. Once I got all those stats I used an statistical data program called Eviews to help weigh the importance of these stats in determining what makes a succesfull rookie. This by comparing the previously selected rookies of the year in a time span (which right now i dont remember since I programed it a while ago) and after doing that I just got to introduce the stats in the regression which gave me a coefficient. Then I took that coefficient and made it a relative index so that it would assign a 100 to the player with the highest coeficcient. I hope it clarifies things and sorry for the bad grammar Im not a native English speaker.
0 - Posted on: Fri, 02/24/2012 - 9:05pm #641084
tli232Can you provide the actual formula? that’s the only way we can evaluate the legitimacy of your indexing system against established stats like Hollinger’s PER or EEF.
Also, give us your logic, I’m fascinated as to how other people conceptualize basketball stats.
0 - Posted on: Fri, 02/24/2012 - 9:33pm #641088
whateverParticipantPlayer Coeficcient
1/2 Efficiency Rating + 1/10 Efficiency Rating per 48 min + 1/5 (Pts+rebs +assists) + 1/2 Fantasy Rating +1/2 Team Record
Player Index
50+ (+50(Player Coeficcient))/(highest Coefficient)
That’s how the final formula looked like. To get to those estimated parameters obviously I had to make several models and this actually was one of the simplest ones. The Per and other stats weren’t included since their associated p value was too high so they weren’t of significance for the model. The actual parameters weren’t expressed in fractions. The actual ones had decimals on them but were kind of close to the fractions and numbers used. For instance the intercept was 48.21 instead of 50 however since Im just doing this for fun I tried to make it as simple as I could even if I lost accuracy during the process. The parameter associated to efficiency rating was 0.41 I know its probably closer to 2/5 however I just decided to leave it as simple as I could. If u have interest in any other estimated parameters I can send them to you or something. However the actual formula I used is the one provided before, I think is the simplest one, I wouldn’t used that one if I was paid to make some real data analysis though.0 - Posted on: Fri, 02/24/2012 - 9:36pm #641089
whateverParticipantIm not a professional sports analyst though so Im not trying to compare myself to any of those people I just used simple linear regression tools to make an interest post, Im not planning to make my living out of this
0 - Posted on: Fri, 02/24/2012 - 10:14pm #641095
tli232The covariance on a lot of those inputs are rather high, if I were to take a guess. Mathematically, that’s kinda iffy.
However, I do really like how there’s Efficency / 48 in addition to straight efficiency. so many people look at them independently.
0 - Posted on: Fri, 02/24/2012 - 10:37pm #641103
whateverParticipantActually u r right the data base and the regression had some problems however i didnt mind them, my main objective wasnt to develope a perfect index I just wanted to provide a different focus from indexes such as the PER which sometimes i not a very good parameter to evaluate rookies and players in general. That becomes obvious when you look at them and find the highest PER among rookies belongs to faried while Rubio is ranked 10 among rookies.
0 - Posted on: Fri, 02/24/2012 - 11:43pm #641116
GronounoursParticipantI think you could create thousands of those formulas; you tweak them a bit and you get a decent result. It’s very subjective.
Why is Team record so heavy (1/2)? That doesn’t make sense at all to me. Irving is much better than Goudelock, why should he be punished for his teammates?
0 - Posted on: Sat, 02/25/2012 - 8:33am #641188
whateverParticipantIt is actually 10/2 times the record so for instance Cleveland record is .419 so that would represent a 2.095 which is less than a tenth percent of Irvings Coefficient which is 29.98. So the team record at least for him just represent a 7% of his coefficient and for me thats not very high at all. Even if u compare the difference among this two in their teams record once u put it in the formula it gives just 0.84 edge to Goudelock which represents just a 2% of Irvings coefficient. So in my opinion thats not high at all, actually one might think the weight of it is rather small. But its weight is not very significant in the formula since for a rookie historically team record hasnt been as important as for example an mvp to actually being considered a ROY.
0 - Posted on: Sat, 02/25/2012 - 10:13am #641217
GronounoursParticipantThat makes sense, thanks for your explanation.
0 - Posted on: Sat, 02/25/2012 - 1:43pm #641304
aamir543ParticipantWow, that is very impresive. Nice job. Out of curiosity does it just work for Rookies, or other players, and does the index number represent efficiency, or production, two different things. And is it all relative to 100, or in a group where you are making the index does someone have to have 100. And how does Faried qualify, he’s barely played. And Irving Rubio are head and shoulders ahead of everyone by this standard, and on a sidenote, both of the Morris Twins have really surprised me, and look to have a bright future ahead of them.
0 - AuthorPosts
You must be logged in to reply to this topic. | Login |