Production You Can’t Count On

This Project is on the Backburner

This post is going to be big.

From here on in, I think every single post I make will probably be based on this post. It’s so big, I think a question/answer format is the best way since linear progression is the best way to explain everything.

  1.  Why Was Bballpants Created?
    To show the forums what I thought was right. So they could see the proper methodology.
  2. What is the proper methodology?
    The proper method is to be able to test your predictions and generate a % correct. Reproducible, easily understandable and testable.Ed Weiland got it right with Lin (small stat sample). He focuses on positions. A Point guard should finish at the hoop, and steal. A Power Forward, should finish, get to the free throw line, Block and Rebound. He cherry picks those stats and ranks them accordingly.

    Wagesofwins, D. Berri, Arturo Galleti, James brocato, nerdnumbers, thenbageek, miami-heat-list , they get it done with Wins Produced/Win Score (large data sets). Wins Produced focus on Efficiency, points, steals, rebounds, assists, regardless of position. As such, it is much better than Ed Weiland’s method when you have to consider multiple players of different positions who CAN play multiple positions.

    Ian Levy of uses brute force per stat renderings. He’ll take the entire college stat line and compare it stat by stat to another person’s college stat. His similarity scores are wrong but useful. His main goal isn’t prediction but understanding which stats are important. The very wrongness of the Similarity Scores, and the accurateness of and Ed Weiland means that some stats ARE MORE important!

  3. What Stats are More Important?

    Ian Levy of did an amazing article for Probasketball.
    Read the article. It’s good. The main chocalate pudding of his article lies below:

    1. Ast/40 _ 0.877
    2. Blk/40 _ 0.875
    3. OReb/40 _ 0.859
    4. FGA _ 0.838
    5. DReb/40 _ 0.836
    6. 3PTA/FGA _ 0.827       Translation: How many times do you attempt a three or a fg?
    7. Stl/40 _ 0.750
    8. FT% _ 0.714
    9. PF/40 _ 0.656
    10. 3PT% _ 0.622
    11. FTA/40 _ 0.544
    12. TO/Pos _ 0.530
    13. Pts/40 _ 0.370
    14. TO/Pos _ 0.530
    15. USG% _ 0.276
    16. Min/G _ 0.118

    That’s the transferability rate of NBA skills. I’m going to group them by highest to lowest probability of transfer.-Production You Can Count on, the title of Ian Levy’s article, means Assists, Blocks, Rebounds, Steals and FT% are all highly transferable positive skills.

    The shooter’s mentality transfers very well too. People that like to shoot threes, shoot a lot of threes! The corollary to this is that shooter’s mentality and behavior transfers over to the NBA for good or bad.

    Production YOU CAN’T Count on! is the inference I made from his USG, Min, Pts, and 3pt%

    You can’t count on scoring. You can’t count on shooting. it’s barely 50%. If you apply usage and minutes, you get an even lower scoring coefficient. It turns into 33% or some such. You have to be really great at shooting before anyone can consider you to be a shooting guard. If you can’t count on points, so why include them so prominently in top algorithms?
    Why does Ed Weiland prioritize it in his 2PP% favoring?

    Maybe we should de-prioritize scoring! Counter-intuitive isn’t it? After all, the NBA is a ‘make buckets’ league. Maybe the reason why Ed and Wins Produced are so prognosticatory is because they prioritize the right stats. Rebounds, Steals, Assists, and Blocks are 75%+ transferable!

  4. What Are You Suggesting?

    I suggest we extend Ian Levy’s complete work to its logical conclusion. Before we apply the Wins Produced formula to a college player’s stats, we adjust those same stats for NBA transferability. Only 75% of draftees keep their FT%  in the NBA. So why don’t we multiply there FT% by 75%? We also do the same for Rebounds, Assists, and [Blank] as well!

    The real doozy would be the points. We’ll have to halve it.

    A Win Score (simplified Wins Produced) ranking list would also be incredibly interesting.

So, What Are The Conclusions?

How do we test this theory? I’m going to use this and map the first 4 NBA years of every draft from 2000 to 2008. Then I’m going to record their career NBA average in terms of WP48 and WP. Then I’ll take  those college stats and apply the appropriate % decreases so as to mimic their stat decreases when they reach the NBA. Then I’ll rig a WP/WP48 matrix with those college stats and see how closely it fits with their career.

Hopefully, this will be as accurate as predicting college to NBA WP48 as WP48 is to NBA wins.

*** The immediate conclusion from this is that Schedule Strength is meaningless. Your points are going to increase due to a weaker schedule  but only 33% of the draft will bring their 2PP% to the NBA. So why bother with schedule strength. You might get better stats across the board? The free throw line is the same in Division I as it is in Division 2. There are still 100 Possessions per game. And if the pacing is slower in the lower conferences, doesn’t that counteract the increased scoring?
I’ve actually done a pace adjusted win score list and a totally unadjusted win score list. The difference (eyeballing) seems negligible. We’re talking decimals.

*** Ian Levy’s isn’t working on his NBA Skill Probability work until next year, I think. He has to complete the entire stat gamut before Wins Produced can be calculated.
*** I’m not sure I should have  posted this without the completed data. But my memory is bad, and posting up the idea might actually help someone else do it. Or light a fire up Ian Levy’s ass to complete his stat projections. 🙂


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: