In my last post, I introduced the idea of a “convex hull” for the usage vs. efficiency relationship. If we agree that the upper-right edge of that relationship represents “greatness” (in terms of scoring), then it is a simple matter to quantify the relative greatness of all other player-seasons by measuring the distance from each point to that edge (see the plot below).
The distance between a data point and the edge of the USG/EFF relationship is a measurement of relative offensive value.
You may already be familiar with my marginal scoring metric, PSAMS (if not, see here). The basic idea with that metric is that I try to take into account the volume and type of shots (inside, mid-range, 3-pt, and free throws) for each player and calculate an “adjusted” scoring metric. For example, I give more credit to players who generate a high volume of inside shots and debit players who don’t. I take into account the fact that some players are responsible for taking more than their fair share of mid-range shots (which tend to be lower efficiency), while others take less, thus placing the burden of taking those bad shots on their teammates. And so on… Continue reading
UPDATE: Figured out how to do this using the corrplot package which add a lot more options. Check it out for GSW:
Click to enlarge.
This post owes its genesis to Alex Konkel who blogs at Sports Skeptic. He asked if I could calculate something called the variance inflation factor (VIF) for the adjusted +/- regressions I’ve been doing. This would apparently enable us to examine the collinearity between variables (i.e. players). We’re actually trying to work out some kinks in that analysis, but in the meantime, it gave me an idea. Why not just calculate the correlation matrix for all players? Continue reading
It’s always a hot topic in Warriors land to debate whether we should have drafted Ekpe Udoh over Greg Monroe. Monroe and Udoh couldn’t be more different in terms of box score stats (Monroe gets them, Udoh doesn’t) — as Kevin Pelton pointed out just today (and you, dear reader of my blog, know I’ve been on the case for a long time already).
Box score stats are nice, and sometimes they lineup with team-level results, but the latter are what we really should care about the most. How does a player impact team-level results? To answer that question, advanced stat guys like myself look at team-level metrics, such as adjusted +/- (APM/RAPM) and my new A4PM (adjusted four factor +/-). With that in mind, today I wanted to take a look at how Udoh stacks up against some of the alternative draft choices the Warriors might have made. I’m focusing solely on the players that were talked about leading up to the draft as being possibilities for GSW at the #6, and the ones that have had enough playing time by now to have some confidence in their +/- data. These two criteria narrowed the list to 5 players: Greg Monroe (#7), Al-Farouq Aminu (#8), Paul George (#10), Cole Aldrich (#11), Ed Davis (#13), Patrick Patterson (#14). I also threw in Gordon Hayward (#9), since he was in the middle of that pack (but not ever rumored to go to GSW, as far as I can recall). Continue reading
Nate Parham (@NateP_SBN) over at Golden State of Mind recently replied to my post on adjusted turnovers:
Curious: could you use these numbers along with Hollinger’s pure point rating to make an adjust pure point rating?
The reason I like PPR is that it effectively accomplishes what people miss when people talk about a point guard’s turnovers: how well he balances the harm of creating turnovers for the team with the benefit of creating a scoring opportunity for others.
Hollinger’s PPR formula just uses assist and turnover rates (not adjusted for anything except pace). Continue reading
You know Blake Griffin. Get to know Ekpe Udoh.
There are four factors of an offense or defense that define its efficiency: shooting percentage, turnover rate, offensive rebounding percentage, and getting to the foul line. Striving to control those factors leads to a more successful team. (Dean Oliver, “Basketball on Paper”)
A while back I did some work regressing the four factors (FF or 4F) on point differential at the *team* level (Part I and Part II). The result was the following equation:
- effective FG% (eFG):
- foul rate (FTR):
- turnover rate (TOR):
- offensive rebounding rate (ORR):
Since the time I wrote that post, I’ve thought it would be useful to translate the team level FF relationship to point differential (and winning) down to the player level. The way to do this (or, at least, one way) is to calculate adjusted versions of each of the four factors (i.e. APM-style), and then regress those adjusted factors onto player-level APM or RAPM. (It should be noted that Joe Sill calculated adjusted FF a few years ago, but those data and the articles have been taken down since he started working in the NBA.) Here, I’m using data from 2009-2010 through last Thursday’s games to calculate my own version of RAPM and adjusted four factors for each player. Continue reading
For my version of adjusted assists (not sure if anyone has done it before), I distinguish between assists to 2-pt and 3-pt field goals. To do this, I simply multiply the AST3 totals for a stint by 1.5. So, for example, if there is a 10-possession stint which has two assisted 2-pt field goals and one assisted 3-pt field goal, that would be 3.5 “equivalent” assists. The logic here is obviously that 3-pt buckets are worth more than 2-pt ones.
When looking at these numbers, you may at first be surprised to not see the top of the list jam-packed with a bunch of point guards (although Kidd, Nash, and DWill are all in the top 10). This is because *any* player on the floor that is a great spot-up shooter (especially 3-pt shooter) is going to naturally raise the number of assists dished out. Also, this metric (in theory, anyway) should be able to pick up on players who generate a lot of the mythical “hockey” assists (the pass before the pass before the shot). Using similar reasoning, one also realizes that point guards who have very high USG and look for their own shot (e.g. Rose and Westbrook) are going to appear lower in the rankings. Remember, adjusted assists is telling us how many assists were dished out at the *team level* while a player was on the floor. I have an idea in mind that might be able to adjust adjusted assists to account for these issues, so stay tuned for a future update, if I can work out the details. Continue reading