APBRmetrics

[Mike G]

Is one wheel coming off?

Ray Allen's shooting dropoff, as the season progresses:

[Ben]

[Harold Almonte]

[Mike G]

Here are Sagarin's ratings for the best teams of the last 10 years.

[John Beattie]

Dang — Celts only meet Spurs twice during the regular season, and the first isn't until February 10 Sunday...

http://sports.espn.go.com/nba/teams/schedule?team=bos

...in Boston.

Then March 17 Monday in San Antonio.

[Mountain]

I developed 2 crude championship contender models previously. One said look at top 7 defensive efficiency teams and then pick the one with the best offense in that group as most likely contender and then the next closest are the main rivals. Another said sum offensive and defensive rank and set the order based on the average of the two.

By the first method Celitcs are the favorite and Pistons and Spurs are the biggest challengers. By the second Pistons actually come up 1 notch ahead and Spurs still third.

Magic, Nuggets, Hornets, Lakers and Suns (defense faded from good start back to just average) are in next group by either approach and closely bunched. Jazz, Mavs and Raptors further back and probably represent the final contenders. Rockets, based on performance to date and mostly the 22nd on offense (which could change of course) come up essentially same level as Philly and not on the short list right now.

But rank is not as good as actual efficiency. On net efficiency the contender rank order is Celtics, Pistons, Spurs (no change there), then Suns Mavs (they move up based on strong offenses) followed by Lakers Magic Nuggets, Jazz Hornets Raptors. So a couple teams move up but it is the same top 11.

[gabefarkas]

[Mountain]

The rationale for these 2 predictions models is discussed / can be seen in the original thread http://tinyurl.com/392l2o where it arose out of dialogue mostly with Mateo and Johnny Slick.

i talked about the pretty good performance of these models in last several playoffs somewhere this summer /early fall - I'd add the link if I find it. (I also previously noted that there was some variation that suggested a bit more flexibility would be needed to capture more of the top contenders / title winners. http://tinyurl.com/2zfjv8 )

Maybe I backed off the rank method too far. But I wanted to mention the alternative in recognition of past criticism or caution about the rank method. There isnt much difference in predictions but how well Suns and Mavs do could be somewhat useful in choosing between the rank approach and the actual value approach.

Use of z-scores would be another format.

[Mountain]

I've also wondered if performance (rank or actual) against top 16 or top 10 might be better than overall league performance (with a lot of games against non and low ranked playoff teams which may not be the best guide about performance in the upper playoff rounds).

I'd prefer using offensive and defensive efficency but hoopsstats.com has several splits easily available using NBA efficiency. Own and opponent efficency are available as is the differential.

http://www.hoopsstats.com/basketball/fantasy/nba/teamstats/08/1/diffeff/9-1

Using differential against current playoff level teams as quick first cut the top 3 is again Celtics, Pistons, a big dropoff then Spurs . Phoenix 4, Jazz 5. Lakers 6 closely bunched behind Spurs. After another dropoff Dallas 7 then a dropoff to Denver and another to Houston. Hornets and Magic fair very poorly, too far back to look like a contender by this measure.

But looking at performance against west top 8 for west and same for east Dallas' standing is much higher- in fact #1 right now followed by Spurs then Rockets. In east it is Detroit on top by a clear margin then Celtics and really no one else. Of course these are very small sample sizes right now. By end of season it might be more enlightening. The guidance given in the the past could be checked. However regular season is not playoffs even against top teams.

[Mountain]

Last season Spurs had 7th NBA own efficiency and best on efficiency allowed against playoff teams. They weren't in fact the top 7 defensive team with the best offense when the dataset is just playoff teams. Both Mavs and Jazz were just higher ranked on offense (Jazz by a very small amount) but lower ranked on defense. Spurs were however the best on combined strength rating, giving that method the edge on predictive correctness this season. 3 of the 4 conference finalists were top 4 by the combined strength rank against playoff teams in regular season method so it worked pretty well by that as well as identifying 15 of 16 playoff teams. Cavs werent anywhere close to the top contender by this method with the 5th worse offense against playoff teams. The top team identification performance (3 of the top 4) was equal for performance against the entire league. Whether performance against playoff teams is a better method than against league as a whole will take more checking.

Actual net efficiency differential picked 2 of top 4 and 14 of playoff 16. Only slightly worse. The rank method called Utah better while actual differential expected Dallas. This time rank got it right but seeding obviously played a role and you can't say which was more right about these team's realtive strength in an absolute sense from the playoff results. Using offense as the deciding criteria among top 7 defensive teams also gave the edge to Utah.


By combined strength method against playoff teams Pistons and Celtics this season are currently tied for 1st with Spurs third and same next 4 as found when looking at performance against all teams. By the look at top 7 defensive teams and then use offense to decide method Pistons have the slightest edge with Spurs again 3rd. But after you find the top 4 it comes down to the particular matchup rather than who scores higher by any of these methods.

By actual net NBA efficiency differential Suns are 4th best and Lakers are 6th right now. By rank Lakers are 6th and Suns in a 4 way tie for 9th. 2 more test cases for the method comparison. By defense first then offense neither is on the radar for top contention.


In 05-06 actual differential ranked the top 4 Dallas Detroit LA Clippers Phoenix. If you go by conference it picked Miami as 2nd best there. Combined strength did and Top 7 by defense then offense called those 4 by conference too.

You'd have to looked at much longer dataset to judge prediction based on performance against playoff teams vs all teams as the most recent 2 years didnt show better results from looking at performance against playoff teams. I don't think I will dig deeper right now. I was mainly interested in the early check on the 2008 race. The original thread cited above took the long historical view and demonstrated the relative power of the basic approaches.

[Mountain]

In the east it appears to be Boston vs Detroit and little reason for me right now to expect others to get in the way.

In the west if top 7 on defense (or near it) is quite important then Spurs look like they have pretty good separation from the rest, with Nuggets and Hornets perhaps being leading contenders but maybe not enough offensive efficiency to go all the way. (Their weak performance so far against top 10 teams by record also mitigates.)

If both sides of play matter equally then Suns and Lakers are much closer than if this is not the case (Lakers 8-5 against top 10 is one of the best records in such games so far).

Mavs defensive efficiency at 20th pulls them to 6th best in west on each of the methods but as some teams have shown in past if you can crank up the intensity in April you can still do great things. 5-4 against top 10 makes them one of the 6 teams to be above .500 against that level of competition. Hanging around.

Of course the bracket and specific team to team matchups could help some have a better chance to go further than expected and trip up a few.

[schtevie]

Not sure if periodic updates of the Bulls-Celtics comparison is especially interesting or illuminating, and maybe I will resolve to cease and desist. For now though, here's the year-ender.

29 games in, the Celtics - surprise, surprise - show no signs of not being better than the team the '95-'96 Bulls were. They surpass them in productivity by about 4 points per 100 possessions, a margin which is almost unimaginable.

The consistent caveat is that the Celtics' schedule to date has been weaker than what the Bulls faced. Last year's record of Celtics opponents so far averaged 41.2 wins, whereas the Bulls played in a stronger East and faced teams having averaged 43.3 victories. The point of this is that any intertemporal comparison must take this two game difference into account. At the same time however, given the relative weakness of the East, what this means is that the Celtics' schedule isn't going to get any harder on average (with possibly the opposite being true) so my current belief is what you see is what you are going to get.

Then, what remains interesting is a detailed explanation of how what has happened has happened. In particular, the improvement has been (just under) two thirds on the defensive end, and I don't think anyone imagined that this would be the expected ratio. Perhaps this failure of imagination is largely explained by the familiar arguments of defense being a statistical black box, but maybe there is more to it that that.

[Mike G]

The Celts have had just the 26th-hardest schedule so far. It will probably get harder. 3 of the 4 weaker schedules are in the West; the East has won 48% of inter-conference games : more parity than we've seen in a few years.
_________________
`
36% of all statistics are wrong

[Ben]

[schtevie]

What I see is that the Cs have played every team in the East at least once, and the average margin of victory has been 13. And they've played 6 of the 15 teams out West (including those finishing 4 through 8 last year) beating these by an average margin of almost 16 points.

What remains of interest, of course, are the results against the top four clubs of last year's West which must be the basis of the "26th hardest schedule" determination. But given who the Cs have played, I cannot see how there would be much difference between the 26th easiest and the median schedule.

As for the Sagarin "3.5 points better than last years Spurs" determination, if this metric is represents net points per 100 possessions, this suggests the Cs would finish better than the Bulls, if raw net points, merely the same. Whatever.

Again, to my mind, the interesting question now is how and why the dramatic change occurred.