Before getting into the trenches, let's establish a few things. Metrics like PER, Wins Produced, and HDR are attempts to measure who the best players are in basketball. The problem is, "best" is subjective. Some metrics like PER and Wins Produced, believe "best" to be most efficient, and thus attempt to construct an accurate measure of efficiency. Wins Produced does this by determining, through regression, which boxscore stats correlate to winning, and then attributes value based upon that.
I believe the best players aren't necessarily the most efficient, but the most versatile. What do I mean by that? The best players are those who can perform at a high level, regardless of role or system. The trick, is attempting to quantify and remove the "system" from a players raw production. This is difficult to do, but there are a few guidelines we can follow, some of which I utilized in my formula.
The following are components in the HDR formula that differ from Wins Produced. I'm not going to post the entire formula until I can further test its success rate. Keep in mind this is all free flowing. If I see something wrong with the results, I'll tweak and keep you updated.
1.) Divide the credit for assisted shots between the passer and scorer. Players who have a large percentage of their shots assisted tend to be dependent on the team/system for their raw production numbers. To get a better idea of how such players would perform with/without good looks, it's necessary to adjust for assisted shots.
2.) Don't count missed shots as turnovers (like Wins Produced does). Although doing so correlates to wins, that isn't the goal of HDR, and thus would be counterproductive. There is a roughly 25% chance that a missed shot will be offensively rebounded, and thus not result in a lost possession. I divide that by two, giving "half" the credit to the shooter, and "half" to the rebounders. Therefore, I subtract .87 for every shot attempt, not 1.
3.) Slightly devalue offensive rebounds. Intuition, and David Berri, would say that an offensive rebound counts as +1, since you took away/gained your team a possession. However, after accounting for diminishing returns and shot location impact, I feel that an offensive rebound should count as +.8 for an individual. Not a huge difference, but noteworthy.
Let me stress that this isn't an attempt to 1 up Wins Produced. Their goal is to measure productivity through contribution to wins, and their formula clearly achieves that. My goal is to try and gain a feel for a player's "base" talent, and not be overly swayed by raw production, or lack-thereof. Neither metric is right or wrong, as both are attempting to measure different things.
That's the gist of HDR Player Ratings. I am still experimenting, but appreciate the feedback I've gotten so far.