2012 predictions

[Mike G]

This is a very preliminary first pass, using the last 4 years' eWins per minute to forecast players' rates for the coming season. Free agents and others not listed on the teams' roster pages are not included in the (guesstimated) minutes allotments, and the team total player-minutes short of a full season are listed.

Crap. I just realized this is not a full season.
Well, heck. Since we're all used to thinking in terms of wins per 82 games, I'm leaving that column, and also scaling it to 66 games.

Code: Select all

w82   tm    W   L    Min-     w82   tm    W   L   Min-
58   Por   47   19     0      62   Chi   50   16   200
56   Dal   45   21     0      58   Bos   46   20   200
54   Okl   43   23     0      58   Mia   46   20     0
53   SAS   43   23   400      51   Orl   41   25  2200
51   Mem   41   25     0      51   Ind   41   25     0
48   LAC   38   28     0      45   Phi   36   30     0
47   Den   37   29     0      41   Atl   33   33   800
40   LAL   32   34  4200      39   Mil   32   34     0
39   Sac   31   35   800      36   NYK   29   37  3400
38   Uta   30   36  1200      29   Det   23   43     0
37   Min   30   36     0      26   Tor   21   45   800
34   Hou   28   38  2000      23   Was   18   48  1600
31   GSW   25   41  4400      20   Cha   16   50  3600
29   Phx   23   43     0      16   NJN   13   53  3400
22   NOH   17   49  4400       9   Cle    7   59     0
42.4   West                   37.5   East         

If all the Min- shown are filled by replacement-level players, the W and L numbers stand.
Teams filling those minutes with better players should do better than shown.

Not included on any roster at this time: Humphries, Kirilenko, JR Smith, Dalembert, McRoberts, Jamal Crawford, Hamilton, Barea, KMart, Reggie Williams, Vujacic, Ilgauskas, Boykins, Willie Green, Reggie Evans, Shawne Williams, Stevenson, C Landry, Arenas, Weems, Forbes, Craig Smith ...

I've used quasi-optimistic estimates for player minutes: Not many are expected to go over 2600, but neither are many expected to miss a lot of games.
I haven't normalized to 50% wins and 50% losses, but I will once the rosters are complete. Right now, teams average less than .500

[EvanZ]

Mike, do you want this thread to be used for everyone's predictions?

[Mike G]

Heck yeah.
I'm also wondering if we should share minutes expected by players, and all use the same averaged minutes to do them.
Also realizing I didn't scale the minutes needed column by 66/82 of a season.

[EvanZ]

Yeah, I think it would be a good idea to use the same minutes. That way we're comparing our rating systems and not just how we predict minutes. I think most of us care about the former more, right?

[Mike G]

We could define it as a test of the methods we have that ascribe wins to players.
Would we also think that all predictions assume last year's rates should be applied to this year's minutes?

Reason I ask is that I've already done projections, as mentioned above, based on 4-year trends and averages.
For example, Andrew Bogut's 4-year eW/484 have been: 1.24, 1.39, 1.98, 1.43 .
A straight line fitted to those points predicts 1.80 as the next point.
His average over three years is 1.60 .

The avg of those 3 numbers is 1.61 -- and that's what I used in this year's predictions.

Of course this is still my method, but it's quite apart from eWins. So I may do a 'straight' version, along with a slanted version.

Example: For the Celtics, I expect everyone's rates to drop from last year's, except Rondo, Jermaine, and Green. Replacing projected rates with 2011 rates, the team predicted W rises from 57.7 (in 82) to 58.5 -- i.e., not much.

The ratio 66/82 is just over .80, so I'm trying to train myself to think like this:

Code: Select all

W/82  W/66 L/66
70     56   10
60     48   18
50     40   26
40     32   34
30     24   42
20     16   50
10      8   58

[J.E.]

Yeah, I think it would be a good idea to use the same minutes. That way we're comparing our rating systems and not just how we predict minutes. I think most of us care about the former more, right?

I think that would not be a good comparison of the rating systems because we're not using real minutes but projected. A rating system might be very bad at rating players, but at the same time might be able to do 'good' predictions when the minute projection is also far off. These can still be a lot of fun, but I like 'retrodiction' with real minutes a lot more for player metric comparison. It would be cool if everyone posts their player ratings as .csv again, preferably with basketballvalue player id's and in 'impact on points per possession'.

Mike, why is Miami so low? They only lost Bibby and signed Battier. Portland had a +2 SRS last year, I don't think they can win 47 games. Do your numbers include the McRoberts signing? I don't see the Lakers below 0.500 and what's up with Indiana at 41W? They were a -1.5 team last year and their roster didn't exactly change much. Also, I don't think Minnesota can win 30 games, they have last years' -6 SRS roster plus a bunch of rookies. Further, I didn't like NYK moves either, but Bibby/Fields/Carmelo/Amare/Chandler should get you at least to 0.500 even with a horrible bench

[EvanZ]

Jerry, if we're talking about predictions for this season, does it make more sense for everyone to make their own projections concerning minutes or use the same set of projections? Given a choice between those two, I'd say the latter makes more sense. Retrodiction is great, but it's not going to be useful until next summer.

[DSMok1]

I'd just post my player per-minute numbers and the minutes estimate along with total numbers.

[J.E.]

Jerry, if we're talking about predictions for this season, does it make more sense for everyone to make their own projections concerning minutes or use the same set of projections? Given a choice between those two, I'd say the latter makes more sense.

I don't think it matters and everyone should be able to use the minute projections he prefers. You will never be able to tell which is the better metric either way, because you never know if the predictions are good/bad because of the player ratings or the minute projections.

Obviously it would be cool if someone came up with minute projections and shared it with everyone.

Retrodiction is great, but it's not going to be useful until next summer

If we have all the player ratings for the upcoming season we can compare
(expected team point differential given actual minutes and projected ratings) with (actual team point differential)
starting from the very first game, can we not?

[EvanZ]

Retrodiction is great, but it's not going to be useful until next summer

If we have all the player ratings for the upcoming season we can compare
(expected team point differential given actual minutes and projected ratings) with (actual team point differential)
starting from the very first game, can we not?

Oh, I guess. I never thought about doing it in real-time like that, but it makes sense to me. I'll still probably do full-season projections, as well, because that's what most people tend to be more interested outside of apbr.

[Mike G]

A rating system might be very bad at rating players, but at the same time might be able to do 'good' predictions when the minute projection is also far off.

That supposes that compound errors are likely to cancel, but in fact they generally don't. If we average our minutes expectations, that might benefit all of us here, vs outside predictors: Vegas and such.

why is Miami so low? They only lost Bibby and signed Battier.

Last year, LeBron, Wade, and Bosh missed just 14 games total. That's pretty good, and I gave them all about 200 minutes less than last year.
I expect LeBron and Bosh to bounce back a bit in their per-minute rates, Wade to maintain. But their only other above-average player is Haslem (.79), and I give him 2400 min. Ilgauskas is gone? That's 2 to 4 wins.
Figuring Miller for 1600 min., Battier 1800, Jones 1200, Chalmers 2200 -- they need some other quality minute-takers.
Oops. That's in 82 games. Subtract 20%.

Portland had a +2 SRS last year, I don't think they can win 47 games.

For Portland, I have Oden at 800 minutes, worth 2 wins in 82. Roy at zero. Batum improves, Felton and Matthews are good for many minutes, and G Wallace is very good. Aldridge is allLeague. Patrick Mills is coming on, Camby holds the fort, and Kurt Thomas is still reliable. Very deep team here.

[Mike G]

.. what's up with Indiana at 41W? They were a -1.5 team last year and their roster didn't exactly change much.

Improvement and more minutes from Hibbert, Collison, George, and Hansbrough. David West may mean 8 wins, a significant upgrade over McRoberts. George Hill is also added. Rush, Foster, Price, D Jones, then Posey, Stephenson pad the bench. Not many replacement-level minutes here.
Assuming Granger and West don't cancel one another.

[EvanZ]

Losing Dunleavy helps in eWins?

[Mike G]

J.E. Sorry to the max. I thought I was quoting your latest post, but I was editing it. After the butchery, I just deleted it.

[J.E.]

Err, em, well basically
- I think Battier is above average
- I think Oden won't play this year (something like that was said in the press)
- Patty Mills is playing in China right now, not NBA
- Camby has not played many games per season on average, is now 37 and I don't think this years' back to back to backs will be very good for his body
- I completely forgot David West (but still think 41W for Indiana is too high)
- Posey was amnestied by Indiana