2012 predictions

[EvanZ]

Here are my projections in a Google spreadsheet:

[DSMok1]

How did you create the minute projections? Purely based on depth charts?

Also--what set of RAPM's did you use. Are the EZpm values projections?

[mystic]

Thanks Evan for sharing.

RAPM looks like prior informed (6 yr), the ones said to be the best for prediction.

[EvanZ]

How did you create the minute projections? Purely based on depth charts?

Also--what set of RAPM's did you use. Are the EZpm values projections?

Minutes are just educated guesses based on depth charts from ESPN, Rotoworld, and looking at pre-season box scores. I may make some changes up to the last minutes. For example, Gortat just broke his thumb last night. Curry sprained his ankle yet again last night, might be out for a while.

RAPM is the one that Jerry says is the best predictor:

http://stats-for-the-nba.appspot.com/ranking_rec

ezPM is mostly an average between the last two years, but I tweaked some ratings here and there for players based on my gut. If you look at the win totals for both sets of projections, there are not major differences. The R^2 between the RAPM ratings and my "massaged" ezPM ratings is 0.44.

Rookies were assigned a -2 rating. Replacement level was set at -3.3 for players that were not rookies and did not have over 500 possessions or so in either of the last two seasons (otherwise, I used their ezPM).

[DSMok1]

If you're interested, I generated (a while back) rookie and sophomore year APM projections based purely on pick number. Interestingly, age of draftee mattered 0, so there doesn't seem to be any "potential bonus" or anything.

Note the high Stdev, particularly for lower draft picks.

I would project all players below -3.5 projection as -3.5, or what I consider replacement level. (They won't be playing if they're below about that number).

These were generated presuming that, if some of the players didn't get playing time, that we were looking at the top portion of a normal distribution for that draft pick. So if 30% of pick #42 players got PT, I assumed they were the top 30% of a normal distribution at that draft pick.

First Rookie, then Sophomore Year:

Code: Select all

Pick	Expected	Stdev
1	0.51	1.65
2	-0.22	1.65
3	-0.67	1.65
4	-1.01	1.68
5	-1.28	1.72
6	-1.52	1.76
7	-1.73	1.80
8	-1.92	1.84
9	-2.09	1.88
10	-2.25	1.92
11	-2.40	1.96
12	-2.54	1.99
13	-2.68	2.03
14	-2.81	2.07
15	-2.94	2.11
16	-3.06	2.15
17	-3.18	2.19
18	-3.29	2.23
19	-3.40	2.27
20	-3.51	2.31
21	-3.62	2.35
22	-3.72	2.39
23	-3.82	2.43
24	-3.92	2.46
25	-4.02	2.50
26	-4.12	2.54
27	-4.21	2.58
28	-4.31	2.62
29	-4.40	2.66
30	-4.49	2.70
31	-4.58	2.74
32	-4.67	2.78
33	-4.76	2.82
34	-4.85	2.86
35	-4.94	2.90
36	-5.02	2.94
37	-5.11	2.97
38	-5.19	3.01
39	-5.28	3.05
40	-5.36	3.09
41	-5.45	3.13
42	-5.53	3.17
43	-5.61	3.21
44	-5.69	3.25
45	-5.77	3.29
46	-5.85	3.33
47	-5.93	3.37
48	-6.01	3.41
49	-6.09	3.44
50	-6.17	3.48
51	-6.25	3.52
52	-6.33	3.56
53	-6.40	3.60
54	-6.48	3.64
55	-6.56	3.68
56	-6.63	3.72
57	-6.71	3.76
58	-6.79	3.80
59	-6.86	3.84
60	-6.94	3.88
Pick	Expected	Stdev
1	1.37	2.02
2	0.68	2.05
3	0.25	2.07
4	-0.07	2.09
5	-0.34	2.11
6	-0.57	2.13
7	-0.77	2.15
8	-0.96	2.17
9	-1.13	2.19
10	-1.29	2.22
11	-1.44	2.24
12	-1.58	2.26
13	-1.72	2.28
14	-1.85	2.30
15	-1.98	2.32
16	-2.10	2.34
17	-2.22	2.37
18	-2.34	2.39
19	-2.45	2.41
20	-2.56	2.43
21	-2.67	2.45
22	-2.78	2.47
23	-2.88	2.49
24	-2.99	2.52
25	-3.09	2.54
26	-3.19	2.56
27	-3.29	2.58
28	-3.39	2.60
29	-3.48	2.62
30	-3.58	2.64
31	-3.67	2.66
32	-3.77	2.69
33	-3.86	2.71
34	-3.95	2.73
35	-4.04	2.75
36	-4.13	2.77
37	-4.22	2.79
38	-4.31	2.81
39	-4.40	2.84
40	-4.49	2.86
41	-4.57	2.88
42	-4.66	2.90
43	-4.75	2.92
44	-4.83	2.94
45	-4.92	2.96
46	-5.00	2.99
47	-5.09	3.01
48	-5.17	3.03
49	-5.25	3.05
50	-5.34	3.07
51	-5.42	3.09
52	-5.50	3.11
53	-5.58	3.13
54	-5.66	3.16
55	-5.75	3.18
56	-5.83	3.20
57	-5.91	3.22
58	-5.99	3.24
59	-6.07	3.26
60	-6.15	3.28

This should beat using simply -2 for rookies by quite a bit!

[Mike G]

Any minutes to go along with these rates?

It looks like a high or low draft pick improves about 0.9 in APM from yr1 to yr2.
Why not just add that difference to their rookie rate?

I see Evan has assigned mpg to players, which are minutes per team game played (not mpg the player appears in), so a team sums to 242 mpg -- about the average, given overtimes.
That seems like a good convention.
( I did it by 100-minute units -- 16000 min/team -- but I can switch to integral mpg. )

Also, Evan: You don't list some players I've figured for major minutes: Krstic (Bos), Pachulia (Atl), Dante Cunningham and Reggie Williams (Cha), ... Are they not under contract?

I'm just looking at these pages: http://sports.yahoo.com/nba/teams/cha/roster
They seem to update daily.

[EvanZ]

Yeah, that just seemed to me like the way most people think about playing time, and it's easy enough to look up without having to do any other calculations.

[J.E.]

Here are my projections in a Google spreadsheet:

If you use RAPM numbers to make projections you need to multiply them with something between .5 and .8 first. (Normally, between .7 and .8 ist optimal, but because of the lockout something closer to .5 might be better). Taking the full values is definitely not optimal.
There will probably a small update of the latest '11 prior informed RAPM numbers tonight. Those new numbers*0.7 will be my "prediction" for the season.
I'll just use the code from the retrodiction contest to compare player metrics in their ability to "forecast" team differential given real minutes and projected ratings. I'll add some bits to compare the ability to predict "lineup differential", because to me, predicting team differential over the entire season feels like asking the weather man how many times it will rain next week, but not caring about whether it will rain Monday/Tuesday/etc.

[EvanZ]

Here are my projections in a Google spreadsheet:

If you use RAPM numbers to make projections you need to multiply them with something between .5 and .8 first. (Normally, between .7 and .8 ist optimal, but because of the lockout something closer to .5 might be better). Taking the full values is definitely not optimal.

Why is this?

[DSMok1]

Regression to the mean, I would presume.

[Kevin Pelton]

Here are my current minute projections:


(Haven't used Google Docs publicly, so let me know if that shares correctly.)

[DSMok1]

Here are my current minute projections:


(Haven't used Google Docs publicly, so let me know if that shares correctly.)

That works fine, Kevin, thanks! What sort of methodology do you use to generate those?

[EvanZ]

Interestingly, the regression of Jerry's RAPM onto my 2-yr ezPM gives the following:

ezPM = 0.68*RAPM -0.06 (R^2 = 0.45)

So, I'll probably use that 0.68 as my RAPM multiplier.

[Kevin Pelton]

Games played is based on games missed the previous two years, plus an average of five games missed per player during the shortened season. The minutes are just my guesses.

[EvanZ]

Here are my corrected blended predictions (using Jerry's updated RAPM):

Code: Select all

EAST			
MIA	76.5%	51	15
BOS	65.1%	43	23
CHI	64.9%	43	23
ORL	64.0%	42	24
ATL	54.5%	36	30
PHI	53.5%	35	31
NYK	53.2%	35	31
IND	48.8%	32	34
MIL	47.9%	32	34
NJN	37.6%	25	41
CHA	36.0%	24	42
DET	34.2%	23	43
CLE	33.0%	22	44
TOR	32.5%	21	45
WAS	28.6%	19	47
WEST			
OKC	64.2%	42	24
LAL	62.9%	42	24
DAL	62.7%	41	25
LAC	57.8%	38	28
SAS	56.7%	37	29
MEM	54.5%	36	30
POR	53.9%	36	30
DEN	53.7%	35	31
PHO	47.2%	31	35
HOU	45.2%	30	36
GSW	43.1%	28	38
UTA	39.2%	26	40
SAC	39.2%	26	40
NOH	38.7%	26	40
MIN	34.2%	23	43