APBRmetrics

Mountain · Joined: 13 Mar 2007 Posts: 800

1. Spurs finish with lowest win % since Popovich's rookie season as coach.

2. One of Blazers or 76ers don't win 45.

4. Golden State doesn't win 35.

5. Pacers make playoffs. Hawks don't.

davis21wylie2121 · Joined: 13 Oct 2005 Posts: 578 Location: Atlanta, GA

My preview is far from final, but here's my preliminary guess:

Mountain · Joined: 13 Mar 2007 Posts: 800

My 5 bullet points are very close to your predictions for all those teams.

I'd add I guess better from Raptors but it is a guess, lots of players with things to prove next season.

I am tempted to put Detroit a bit lower given second half of season slip. age especially for McDyess and uncertainty about Sheed, the new coach and of course the better east and probably league...

and the expected more minutes for the much hyped but very inconsistent Stuckey without the average stats to back it yet:

well below average TS% and incredibly low eFG%- but star level usage rate

-10 adjusted +/-, 11th worst amongst those rated across the league at countthebasket and 4th worst offensive impact of any rated PG.

below ,500 win % on court with either Billups or Hamilton... and MCDyess and Prince and just barely above .500 with Wallace.

A rookie just getting feet wet, but a rookie in a great spot...that hasn't proven much yet besides the ability to flash occasionally (but with average ratings below Flip Murray's from Detroit the season before and his performance in Indy this)

Mike G · Joined: 14 Jan 2005 Posts: 2061 Location: Delphi, Indiana

DW21, you are all over it.
You've got the East winning 12 games more than last year. That works out as an improvement from .425 vs the West, all the way up to .450 . Seems reasonable.

Ah, but -- 30 teams total 1229-1231 . Someone gets another W !
_________________
`
There's no I in analysys.

davis21wylie2121 · Joined: 13 Oct 2005 Posts: 578 Location: Atlanta, GA

Damn rounding errors! Very Happy

Um, give it to the Hawks, I think 33 looks a little low. And I'm surprised at how well Minny comes out looking. Pretty soon I'll go in-depth and tell you guys how this was generated (a combo of adj and stat +/-, last year's BoP method, and a goofy "scouting ratings" method that works surprisingly well), but I wanna get my preview written first...

supersub15 · Joined: 21 Sep 2006 Posts: 160

davis21wylie2121, Toronto were ranked 13th in DRtg last year. You'd think that the addition of Jermaine O'Neal would make them better defensively.

John Hollinger · Joined: 14 Feb 2005 Posts: 130

Can't argue with much in that list, but I guarantee Bucks won't be 29th in defense under Skiles.

Scott S · Joined: 10 Feb 2008 Posts: 27 Location: East Rutherford, NJ

How have apbrmetrician's projections compare to projections like ESPN's projections in the past? I noticed last year there were 4 projections posted on the forum and compared.

davis21wylie2121, I'm interested to see how your scouting ratings method works, sounds interesting.

For those of you who do these projections, how in depth do these they typically get? Are elements like age adjustments, team fit, credibility of past player data available, advanced boxscore statistics, projected strength of schedule and/or any other adjustments employed (such as cap room available for a mid season acquisition or coaching changes or injury probablity or something)? And is usage/minutes determined in a similar manner? Obviously this could get infinitely complex. It must be tough finding a cutoff point as to what is worth valuing.

Mountain · Joined: 13 Mar 2007 Posts: 800

DW I see you have Jazz with #1 offense next season. They were 14th last season on offensive efficiency . I could see 7-10 but don't expect #1. What contributed to that big rise in your projection?

Dallas seems to a bit low on media attention right now. Not sure that lasts the whole season but maybe they do better that way than in the past as an expected to be very near top.
Team numbers look a bit better than my impression of the individuals so the attention to construction, offsetting deficiencies and building up strengths incrementally is good.

Ben F. · Joined: 07 Mar 2005 Posts: 380 Location: MD

Actually, Mountain, Utah was the #1 offensive team last year. It looks like you were looking at Knickerblogger's Four Factors page which has been broken for awhile (I'm not sure what happened) but Justin at B-R.com was gracious enough to add a sortable Four Factors section to his season summary page (you can find it at the bottom of this page).

Mountain · Joined: 13 Mar 2007 Posts: 800

Thanks Ben for the correction. I was using kb. I have stumbled upon the Utah dataset being based on just 7 games before, even reported it but forgot.

davis21wylie2121 · Joined: 13 Oct 2005 Posts: 578 Location: Atlanta, GA

Yeah, at 113.8, Utah actually did have the best ORtg in the league, which was surprising to me given Dallas and then Phoenix's stranglehold on the #1 slot most of the decade.

"Scouting" is actually kind of an overly fancy way of describing my method (introduced here)... I basically look at each player, with no stats at hand, and give them a 1-10 rating (although practically every player is 4 or above; only Jerome James and Allan Houston's theoretical comeback got 3s). Then I split each team by depth chart: starters, bench, and "other"... In 2008 each team split their minutes 60-30-10 between the 3 groups, so I weight each 5-man unit by those percentages, and come up with a power rating which I then force to look like win totals using a z-score transformation. Bootleg? Hell yes. But it actually doesn't work half badly.

For the other methods, I basically did the same thing as last year... Predict G, MPG, ORtg, %Poss, and DRtg for each player, use the depth charts for each team's starting and 2nd fives, and change each individual's ORtg using Eli's 1.25 tradeoff based on whether their team was over/under their number of possessions from last season.

The APM method used the same G and MPG projections as above, but also involved a projected APM using pure and stat +/-. Projected point differential is achieved through a weighted average of individual +/-'s, where the league avg is forced to 0. What you see above is the average of win estimates from the "scouting" & APM methods, and 2 BoP-based methods (one normalizes ORtg and DRtg to recent league avg/stdev before using pyth, one just normalizes pyth).

As JH mentions, offensive and (especially) defensive ratings are significantly affected by coaching styles, which just isn't picked up when every team is simply considered the sum of its parts (players). Besides, as we saw with the C's last year, individual DRtg is really team-dependent, so when players change teams it can significantly skew the results.

So that's what happening here. You'll get all of the gory data when I finish my preview and rosters get (relatively) finalized next month. And we can compare how each of the component methods (APM, BoP, off-the-top-of-the-head player ratings, etc.) perform against each other as well.

Mountain · Joined: 13 Mar 2007 Posts: 800

Adjusted +/- best fits the data of total activity team & opponent when a player is on the court for the league but I wonder how close a blind to those ratings / from scratch subjective ranking of +/- impact would get to the performance of adjusted +/- for players on teams by an informed observer and the differences between the two would be useful fodder for further consideration. Or as you did, a meta-metric, objective/subjective blend. Can we get closer to a specific team's performance - and player ratings- that way?

Kevin Pelton · Posted: Fri Sep 12, 2008 3:09 pm Post subject:

Mountain · Joined: 13 Mar 2007 Posts: 800

Misc notes:

Lakers tallest (unweighted) team in league last year. How will Bynum-Gasol work out? A twin towers that achieves glory? Probably with the rest of cast and coaching, though haven't seen together yet.

Denver was next to smallest last season and may feel more of the effects of that without Camby especially if they use the two small PGs. Nene, Steven Hunter and Chris Anderson at 6-10 and over isn't a lot of depth. Someone more must be coming? I think they will need more to have a successful season.

Hickson for second Rookie team. If Cavs stay at 50 or advance I'll take a flyer that it may be more about him than Mo Williams. If he plays D. And if LeBron likes him and decides to help / use him actively.

Rose starts slow. Alexander was taken too high.

Portland's defense will be pivotal. Fernandez and Bayless may make it an even more foul-prone rookie season for Oden (beyond his own transition learning process) and the D might not move much from average. DW shows a decline.

Magic regress on DW's list. Wouldn't be surprised but I am not going to predict that. If they do slip it will be because of guard play, especially at PG. Nelson -6.4 on adjusted +/-. Arroyo's - 8.4 is gone but bringing in Anthony Johnson at -10.9 last season doesn't impress or give a good alternative when Nelson isn't on which is often. But maybe it was an unlucky measurement: it is far far worse than prior 3 years but he is getting older. Play Mike Wilks? +3.8 in 2006-7. Turnovers own and forced may be the key Factor.