This topic contains 3 replies, has 3 voices, and was last updated by AvatarAvatar dmo21 11 years, 11 months ago.

  • Author
    Posts
  • #38256
    AvatarAvatar
    dmo21
    Participant

     Okay, so I’m trying to form my own stat which combines regular stats with weightings, FG%, and how much their respective team wins. As I think this brings in attributes from each area as in the stats there is offense and defense, the FG% shows efficiency, and winning percentage shows if they can win or not, as it is key in comparing players. The stat is broke down for each position, as I think statistical weighting is different for each position (importance of an assist is different for a poitn guard than a center). I have come up with my own weightings (just from my head) and was wondering if they seem logical and how should I change them if I need too? The weighting system I used was from .5(least important) to 2(very important). Numbers can be reused and they don’t have to add to anything significant.

    This is what I have so far:

    PG- ppg(1), rpg(.5), apg(2), spg(2), bpg(.5)

    SG- ppg(1.5), rpg(.5), apg(1), spg(1.5), bpg(.5)

    SF- ppg(1.5), rpg(1), apg(1), spg(1), bpg(1)

    PF- ppg(1.5), rpg(1.5), apg(.5), spg(.5), bpg(1.5)

    C- ppg(1), rpg(2), apg(.5), spg(.5), bpg(2)

     

    0
  • #658530
    AvatarAvatar
    tli232

    The only statistically sound way to find weightings is to do a logistic regression that uses the variables you specified as betas to predict 1 of: PER/ EWA/ VA. That is the true way statisticians come up with weightings for factors to predict future performance.

    Using one’s intuition to determine the weightings is absolutely arbitrary, even though it may work for most players.

     

    0
  • #658716
    AvatarAvatar
    TallmanNYC
    Participant

     Hmm, ambitious to create your own stat. And juding my Til’s response above you are opening yourself up to a lot of second guessing. 

    I won’t go into if a logistic regression is what you need to do because (a) I don’t really know what that means and (b) I really doubt you have the data or the time to do it.

    I will throw out two things though. You are assuming that stats are more important for different types of players because of what you want them to do. I think you should think about what opportunity the player has to do the stuff. Namely steals, rebounds and blocks. I think it is great if your point guard can get blocks. It is tremendously helpful, so why weight that value down low. Even the best blocking point guard of all time (who knows who that is) isn’t going to have 2 blocks a game. But if one does, lets give them credit. 

    I think Centers and Powerforwards get a lot of rebounds just because of where they stand on the court. A center who gets 6 rebounds a game is probably not very good. 2 of those probably come off of missed freethrows. While a guard who averages 6 rebounds a game is probably at least a near all star. But in your stat, the 6 rebound center is getting 12 points and the 6 rebound PG is getting 3. Frankly, I think you have it backwards as I bet that PG (who would be a guy like Jason Kidd or Magic Johnson or Oscar Robertson (only PGs I can think of who rebound like mother-f@ckers)) is a much better player that the 6 rebound center (who would at best be a guy like Broke Lopez).

    And I would flip things around the same way for assists. A PG getting 5 assists a game is a problem. A center getting 5 assists a game is a unique talent that is probably winning you a lot of games. 

     

    0
  • #658732
    AvatarAvatar
    dmo21
    Participant

    Well first off, the stat is more for comparing players at the same position, not comparing a point guard with a center and such. Secondly, the weighting was for what I thought a prototypical player at that position to do. As I would much rather choose a C that averages 15ppg and 10rpg rather than a guy who averages 10ppg and 5apg, because that’s not really what I want them to do. And they still do get a higher stat with doing the extracurricular stats for that position, its just diminished a bit. The stat is more to look at who is pure at their position, point guard being the best example, as a guy who gets more assists will usually end up higher than a point guard who is purely just a scorer or chucker that shoots bad percentages.

    Thanks for the imput though!!

    0

You must be logged in to reply to this topic. Login