APBRmetrics

[Mike G]

Minimum 50 total minutes to qualify. Per-36 minute rates, scaled to team and opponent averages. Starter/sub factor in place, otherwise no sliding scale for Min/G.

Not yet incorporating strength of schedule, nor separating FGReb from FTReb. Rebuilding the spreadsheet, so some bugs may need to be got out.

EDIT: Player scoring columns are affected by (estimated) % of their FG which are assisted. Shot-creation is thus rewarded.

[jkubatko]

[Mike G]

In this semi-linear weights scheme,

T = Sco + .98*Reb + 1.335*Ast - .25*PF + 1.5*Stl - 1.5*TO + 1.75*Blk

Last season, these weights did a good job correlating team eWins to team pythagorean-expected wins. The correlations became especially sharp in last year's playoffs.

The sum rating is comparable to those of other players, but probably it has more meaning when compared to the same player last year or other years. The weights above could be thought of as 'typical' values, across several hundred players. But each situation is unique; so players can still be over- or under-rated by them.
_________________
40% of all statistics are wrong.

[jkubatko]

OK, but that doesn't really answer my question. By itself, does Yao Ming's T rating of 45.2 have any meaning? For example, with John Hollinger's PER I know that 15 is league average, 20 is an All-Star caliber player, 25 is an MVP candidate, etc. PER also follows a scale similar to that of points per game, which is somewhat nice.
_________________
Regards,
Justin Kubatko
Basketball Stats!

[davis21wylie2121]

Just for comparison, here are the best players by DeanO's system:

[Mike G]

[Mike G]

[davis21wylie2121]

That's an interesting point. Dirk's ORtg is higher than all of the top Sonics (except Wilcox), but his low W% comes from that horrid 115.6 DRtg -- worse than last year's Sonics, oddly enough... The defensive discrepancy has to do with Dallas' terrible D so far, and while Seattle's defense has been bad as well, Dallas is really stinking it up. Also, Rashard and Allen have used a lot more possessions/stops than Dirk (Ridnour's about the same), as a consequence of Seattle playing 8 games vs. Dallas' 7. Since WARP is cumulative based on playing time, possessions used, and stops, that's a major factor (although Allen has a higher Usage Rate than Dirk anyway).

Early-season samples are weird, so I wouldn't make much of either list. And perhaps I shouldn't have phrased it "best players," but rather "most valuable players". Value is cumulative. "Best" implies ability, regardless of playing time. And, of course, WARP was neither outlined nor advocated in Basketball on Paper, so I could be twisting Dean's method in a way that he would not even approve of.

[davis21wylie2121]

Current (as of 11/25) leaders in WARP:

[Mike G]

Wow, the season's basically 1/4 done already.
Now I'm scaling player Scoring rates by (TmPts/OppPts)^1.4 . The exponent will shift as the season progresses;
but players for teams with good point-differential will get extra credit for the points they score.
Assists are also scaled by points-ratio, but the exponent is lower (1.15 now).
Rebounds are scaled to (TmReb/OpReb)^.25. These values currently fit team eWins to PythEx.

I find a number like my Scoring Rate (per36) to be much more graspable than Usage or Floor%.
It includes a player's points and efficiency, scaled to a reasonable MPG (36), and accounts for the scoring pace of the games he's in.

Since Tim Duncan plays for the league's best team, his rates are boosted accordingly. On cellar-dwellers, stars lose their luster.
Minimum of 80 minutes played.

[Nikos]

Seems like guys such as Zach Randolph are underrated in this scheme due to his lack of defensive stats and weak supporting cast. As a scorer he has been about as good as anyone in the league so far, but here he looks like a normal or fringe all-star. Wade seems penalized a little too much as well.

[Mike G]

This scheme may rank some guys lower than other schemes. I've got Wade and Zach in the #12-13 slots.
DW21's list (2 weeks old) had them #14-15.
Earlier, I had Wade at #23, Zach #4. At the same time, he had Zach #4;
and Wade near the bottom (between Laurence Roberts and Nate Robinson).
Maybe my system is more consistent, as well as less penalizing for one's teammates.
So if you're going to suppose someone is 'underrated', you may have to explain what the player's 'absolute' rating is.
_________________
40% of all statistics are wrong.

[davis21wylie2121]

Instead of Mike's system underrating Randolph, do you think it's possible that PER overrates him? Portland's defense has been brutal, and it looks like Randolph is one of the prime offenders. Since PER doesn't capture defense other than blocks, steals, and fouls, it's possible that it is missing a huge chunk of Randolph's valuation.

[davis21wylie2121]

DeanO stats as of 12/17 are now up.

[KD]