So, as I said in a few places, I'm working to correct Hollinger's Pure Point Rating. Pure Point Rating, PPR, is this according to Hollinger:
"Coaches and personnel people almost instinctively look to a player’s assist/turnover ratio to check how he’s doing. But ask them why they look at assist/turnover ratio, and you’ll get lots of blank stares and convoluted answers. Probe further, asking if they think Reggie Miller would make a better point guard than Steve Nash, and you’ll quickly get a series of guffaws. But guess who had the better assist/turnover ratio last year?"
"What it really means is almost nothing, because assist/turnover ratio is a flawed stat. The problem isn’t with "assist" or "turnover," it’s with the "ratio." Using a ratio is faulty for two reasons. First, it assumes assists and turnovers are equal, when in fact a turnover is more costly than an assist is helpful."
"Second, it equates very different amounts of productivity. If Player A just sits in the corner all season and finishes with three assists and one turnover, while Player B directs the offense all year and has 300 assists and 101 turnovers, then according to assist/turnover ratio, we should assume that Player A is "better" at running the offense."
"First, it adjusts for the fact that assists do less good than turnovers do harm by multiplying assists by two-thirds. There’s a factual basis in this. As I noted in a recent column, of the three acts of creating the basket (getting open, making the pass and making the shot), the passer does one. So we give him one-third of the credit of a 2-point basket, or about two-thirds of a point. Since turnovers cost almost exactly one point (teams average about 1.02 points per possession), we needn’t make any adjustments to that part of the equation."
"The second adjustment is measuring productivity, to avoid the Player A vs. Player B situation above. The way to do this is to sum a player’s accomplishments on a per-minute basis, then adjust them for his team’s pace. Finally, multiply the end result by 100 to make the numbers more user-friendly."
After the jump I'll explain what I'd like to do to improve it and where I need some help.
Here are the problems with PPR:
1) It makes every assist equal in value
2) It looks at turnovers which include offensive fouls, bad passes, ball handling turnovers, and technical/flagrant fouls
3) It is a per 100 minute stat, not a per 36 minute
To solve problem #3 I'm simply changing the multiplier to 36 instead of 100. That was an easy fix :)
To solve problem #2 I'm using 82games.com to remove offensive fouls, ball handling turnovers, and technical/flagrant fouls. That was pretty easy too.
To solve problem #1, I WAS going to use 82games.com to separate the assists into 4 categories with different values: dunk assists, close assists, jump shot assists, and 3pt assists. Each of these would have a different weighted value with more points to dunk, close, and 3pt assists than jump assists.
This solution initially really helped fix PPR. However, I found some data that was even better than the 82games.com. Hoopdata.com has assists broke up into 5 categories: at rim, 3-9 feet, 10-15 feet, 16-23 feet, and 3pt assists.
Here's where you come in. I need to decide how much of the credit does the point guard get for each assist. Hollinger said the point guard does 1/3 of the work. Obviously an assist at the rim takes a lot more work from the point guard than an assist in the 16-23 feet range. Thus, I believe that an at the rim assist should be more than 1/3, but a 16-23 foot assist should be less than that.
That being said, in the comments place how much credit you think the passer should get for each.
3pt- 25% of the credit
16-23-25% of the credit
10-15-33% of the credit
3-9-%40% of the credit
at rim-50% of the credit
Thanks for all the help guys!