APBRmetrics

[Mike G]

After 8-10 games, there are not a lot of surprises already. Garnett is back to MVP form. Ginobili is the best part-time player ever, perhaps. Chris Paul looks like the superstar his rookie season portended. Other young guys just keep getting better: Dwight and Josh Howard, David West, Josh Smith...
Minimum 84 minutes.

[supersub15]

Thanks for this Mike. Sorry to sound like a complete newb, but what does T stand for. Can you share the formula?

[NickS]

Both TJ Ford and Calderon in the top 50, does Toronto have the best total 48 minute play at the PG position in the league?

I'm probably most surprised by Jameer Nelson. Bendan Haywood and Sam Cassell are surprises, of course, but they've always been good players, it's just surprising to see them that good, but I was starting to think that Nelson's 05-06 season was a fluke.

[Mike G]

T = Sco + Reb + Ast*1.33 - PF*.25 + (Stl+Blk-TO)*1.5

PF, Stl, TO, and Blk are per-36 minutes.

Ast = Ast/36 * TmPPG/OpPPG * 200/(TmPPG+OpPPG)

Reb = Reb/36 * (TmRPG/OpRPG)^.5 * 44/OpRPG

Sco = Pts/36 * (Eff%/.527)^.5 * 100/OpPPG

Scoring is further adjusted by estimated % of FG unassisted.
All categories are then adjusted by MPG and % of games started; which yields a difficulty-of-opponent factor (% vs starters).

100 pts and 44 reb, per team per game, represent somewhat historic averages. Since I like to compare across eras, these serve pretty well. Both are a bit high for this era, making current players look generally better.

In the past and in playoffs, I've tinkered with the multipliers and exponents in order to get better fit between eWins (a function of T) and pyth-exp wins.

(This is all from memory, so I may have to correct myself later.)

Nelson showed a downward trend coming into this season. I ignored that when predicting W-L marks, but I didn't foresee an upturn. Haywood had seemed to be in a longer-term stall. But his fans have been waiting for a bustout. Cassell is good when healthy.
_________________
40% of all statistics are wrong.

[THWilson]

[Mike G]

uA% = (.47*FTA/Min + 2.06*Ast/Min + .77*Stl/Min + 2.9*TO/Min - .103*3FGA/FGA) * 1.2*(TmAPG-APG)/TmAPG

Scoring rate is then multiplied by 1.07*(uA%^.07)
_________________
40% of all statistics are wrong.

[supersub15]

[Mike G]

Yes, but it doesn't appear in the uA% formula. OpPPG represents the 'scoring pace' of the league in the games in which a player appears. By scaling to 100 OpPPG, my Sco rate is basically the % of points a player scores, scaled to his minutes. If he goes 48 minutes, 20 is average. So for 36 min, 15 is avg.
_________________
40% of all statistics are wrong.

[thref23]

[Mike G]

Dougstats lists Paul (35 mpg, missed 1), Jackson (21 mpg), and Pargo (14 mpg, -1) as PG; only Mo Pete as SG (26 mpg). Meanwhile, Pargo is 2nd in assist rate, about 1/3 higher than Jax.

Paul and Pargo total just about 48 mpg, as do Pete and Jax. P/P total 25.3 pts and 12.3 ast, as of a few days ago.
_________________
40% of all statistics are wrong.

[Mike G]

Nash and Paul have identical Ast rates. Nash is a much more effective scorer. But Paul's Stl-TO = -0.3 -- to Nash's -3.6. (I weigh these by 1.5)

7 of the top 10 players here are in the West; 19 of the top 30.
Minimum 118 minutes.

[Mike G]

Holy cow, the season is almost 1/4 gone.
LeBron, missing games, has slipped a few notches.
Minimum 178 minutes. Ranked by eWins projected to 82 G.

[Rasta978]

Is "e82" your projected equivalent of Player Wins (or Wins Produced, etc)?

If you total e82 for all of a team's players, does that equal the team's estimated win total (using pyth, for example)?

[Mike G]

Yes, 'e82' assumes players go the same minutes/team-game the rest of the season.
No, they don't add up to projected team wins. Expected team wins are:

xW = 2*eW - 41.

So a team with 51 eW (10 above average) should win 61 games (20 above avg.) With 41, you get 41.
_________________
40% of all statistics are wrong.