APBRmetrics

[Harold Almonte]

I agree this team won't have the def. eff. enough to be in the 55/60 wins range. But, given they will be one of the best off. team of the East (maybe the best), they only need to run a little more than the past year Boston's pace, something that probably they won't do. But they will go far inside the playoffs, to the point where their defensive level carries them. They won't be an under 105 deff. eff. team.

[jeffpotts77]

[Mountain]

There have been 3 teams and 4 occurrences of an eastern team winning more than 53 games in last 5 years. Detroit twice for their championship year and the season after, Miami's championship season and Indy in 03-04.

Can Boston be close to 61 win Indy? Indy was 9th on offensive, 3rd on defensive efficiency. Not only that but Indy was top 16 on all 8 parts of the 4 factors.

07 Spurs top 20 in all but one. Miami in championship year was top 17 on all but one factor. 05 Spurs top 12 in all but one. 04 Pistons top 20 in all. 03 Spurs top 17 in all but one. Champs limited to one major flaw? All but one of the recent champion flaws, by this standard, were on offensive side of play with Miami opponent turnovers the exception.

Boston, at this point, seems likely to have several 4 Factor flaws and likely in defensive side.

I haven't done a full accounting of this more general question but from related searches I get impression that most of the recent 55+ or 60+ win teams are those with superiority on both offense and defense.

50 plus or minus 3-5 is a little vague but the chemistry unknown creates that range.

[DLew]

Brian M,

Look at the way in which adjusted plus-minus is derived. Roughly speaking, team point differential is divvied up among players on the team. I say roughly because quality of opponent is taken into account and regression tries to minimize the squared residuals, not the the residuals.

This will pretty much never be true for plus-minus done on actual data due to player movement, and will also fail if there is any crunch-garbage time weighting or statistical approximation like Dan did in his article. So, I would say that Schtevie's method is not inherently incorrect, but is far from the best method for predicting future team performance.

[bchaikin]

using mike's scenario with min/g:

min/g - player - simulation stats
28 k.perkins (28 C) - 5 pts/g, 6 reb/g
40 k.garnett (24 PF 16 C) - 19 pts/g, 11/12 reb/g, 3/4 ast/g
40 p.pierce (40 SF) - 24 pts/g, 6 reb/g, 4 ast/g
40 r.allen (24 SG 16 PG) - 23 pts/g, 4 reb/g, 3/4 ast/g
32 r.rondo (32 PG) - 8 pts/g, 4/5 reb/g, 4/5 ast/g
16 l.powe (12 PF 4 C)
12 b.scalabrine (12 PF)
32 t.allen (24 SG 8 SF)
240

and using only 06-07 stats for the above players and their opponents, simulation shows the above sub pattern averaging 54-55 wins per average 82 game season when playing 8200 games against the celtics actual 06-07 schedule...

however the above assumes garnett, pierce, and r.allen all playing over 3200 total minutes. pierce has never done that (but has played over 3000 minutes five times), garnett hasn't done that in 4 years (but has played over 3000 minutes 7 times), and r.allen hasn't played 3200 minutes in a season in a decade (but has played 3000 minutes five times). in the past 30 seasons of the league only 8 times has a team had 3 players play even 38 min/g and over 70 games in a season (79-80 lakers, 90-91 warriors, 90-91 t-wolves, 96-97 hawks, 99-00 grizzlies, 00-01 hornets, 03-04 rockets, and the 05-06 76ers), and all three are 30+ years of age...

i think a more likely best case scenario would be each of the big three playing about 36 min/g, and assuming glen davis playing something like sean may has in the nba (their college stats were similar) and gabe pruitt something like telfair, a sub pattern like the one below might be a more realistic best case scenario:

min/g - player - simulation stats
24 k.perkins (28 C) - 4 pts/g, 5 reb/g
36 k.garnett (20 C 16 PF) - 17 pts/g, 10/11 reb/g, 3 ast/g
36 p.pierce (36 SF) - 21 pts/g, 5 reb/g, 3/4 ast/g
36 r.allen (36 SG) - 21 pts/g, 4 reb/g, 3 ast/g
32 r.rondo (32 PG) - 8 pts/g, 5 reb/g, 4/5 ast/g
16 s.may (16 PF)
16 l.powe (16 PF)
08 b.scalabrine (4 PF 4 SF)
20 t.allen (12 SG 8 SF)
16 s.telfair (16 PG)
240

simulation shows this sub pattern averaging about 49-50 wins per average 82 game season, and that's with garnett, pierce, and r.allen all playing over 2900 total minutes each. the bench is weak, so unless glen davis or brandon wallace (or someone else) plays better than expected, the big three could still play well and the team not win 50 games...

KG seems to be nearly indestructible. Three years of purgatory were probably wearing on him. I expect a major resurgence and an MVP run.

resurgence? garnett has been a dominant player, and a legitimate mvp candidate, each of the past 3 seasons - just not a dominant scorer. he's scored 22 pts/g playing 39 min/g the last 3 years. now you're going to team him up with not just one but two players who each averaged over 24 pts/g the last 3 seasons. its hard to see garnett being a better rebounder, passer, or defender (all-D 1st or 2nd team each of the last 3 seasons) in 07-08 than he has been the last 3 years. yet i just don't think he'll get the high touches on offense with the celtics like he did as a t-wolve, and will thus score less (allen and pierce had high touches/min for an SG and SF in 06-07). for garnett to win the mvp he'll need to score big like in 03-04, and unless allen or pierce miss time due to injury that most likely won't happen...

It seems, sadly enough, I have to conclude that we're not looking at one of the NBA's elite teams here.

could be if someone like glen davis or brandon wallace pans out, and pruitt becomes a capable backup, or you pickup someone else by november. otherwise the bench will be really poor...

[Mountain]

Mike's and Bob's fine work adds to the convergence of expectations. Esimates could also be done using PER, WinShares or NBA efficiency. Any other approach I am forgetting?

It might be possible to try to use backward mapping Team Factor Wins approach for this forward looking purpose later.

Bob does your simulation find the all pieces necessary to assemble a standard 4 Factor rate chart for this team? Same question goes out to Mike and "DavisWylie".

[94by50]

I haven't seen anyone account for the Eddie House signing yet. Did I miss something, or are we assuming that he's just a warm body at this point?

[davis21wylie2121]

[Mountain]

If past data changes essentially "one cycle" then adjusted offensive and defensive ratings ballpark things pretty well. But you consider the whole system interactive, and usage and eFG% levels change, and turnovers change and maybe turnover rates too, and available offense rebounds and that rate too, etc. and you go thru several cycles of interactivity (for lack of better description) and end up with a 4 Factor chart that shows more differences than the first movements do alone. The interative process I am describing is something Bob's simulation is built to capture but I didnt know if it showed offensive rebounds and free throws made and all the things needed to see the 4 Factor chart for the averages of the simulations, so I asked to see if it is available.

[Mountain]

In Minny Garnett was a one man defense according to revised DCS ratings. Boston does have 4 young guys near +10 or well above that. That gives Garnett a bit more help if Doc can stitch it together. Pierce a very modest positive on DCS while Allen a mild negative.

[jeffpotts77]

[schtevie]

Various comments.

Regarding historical analogies to how teams acquiring Big Seconds and Thirds fared, two teams were noted, the '99 Rockets and the '69 Lakers. Regarding the Rockets, they won a fraction that would be equivalent to 50 wins in a full season. My only comment would be that their Big Three was significantly older on average (14 seasons and almost 35 years of age vs. 11 2/3 seasons and 31 years of age). Then the Lakers won 55 games with aging superstars of a more similar cohort (10 and 32).

Regarding the "crudity" of the approximation using minute-weighted Adjusted +/-, I would characterize it rather as a back-of-the-envelope calculation using a very sophisticated measure. And were I to have two additional pieces of information (the 2006-2007 numbers as well as some sense of trend to estimate the effects of aging) I would feel very comfortable with this approach. David Lewin implies that he has these numbers and that they suggest a less impressive performance. Fair enough.

As for David's rejoinder that "the problem with your calculations is that adjusted plus-minus, like almost any statistic, exhibits diminishing marginal returns" this merits some discussion. First, though I made reference to this potential in previous remarks, I am not sure in the particular instance that it is reasonable to expect significant diminishing marginal returns. Were I to have made "unrealistic" assumptions, such as stocking a hypothetical team with high Adjusted +/- Statistics, paying no mind to the role that they would be playing on the court, there would be a problem. This is not the case at hand. The three players play distinct positions and given their usage rates "should" not be getting in each other's way on offense. (At least that is my reading of the lovely data of 82games...thank you Roland.) And again, against this is the fact that they will be facing diluted defenses compared to their average career experience. And then on the defensive end, I am not sure what the diminishing marginal returns might be.

Taking another back of the envelope approach to things, here is a maximum estimate of diminished offensive productivity due to decreasing offensive marginal returns. Between the beginning and end of the shot clock (as defined in 82games) there is an approximate 20% reduction in field goal percentage (and let's assume this for every individual player as well). Suppose that within that mix Paul Pierce had been getting all the good early shots and not taking any of the bad ones (this is not true, but suppose). Along comes Ray Allen and Kevin Garnett who had been getting the same mix in with their clubs (also not true). Suppose, the two interlopers then agree that it really is Paul Pierce's team and they agree to let Paul have his and share the leftovers. This would imply no reduction in PP's efficiency (never mind the hypothetical gains realized by diminished defensive opposition) and a 20% reduction in productivity on the part of RA and KG.

This, I think, is wildly high (as it doesn't accord with the actual shot selection of any of the players in question) but doing this wouldn't dramatically change the picture (from an Adjusted +/- approach) of the Big Three being hugely effective (especially when such an adjustment would only diminish the effectiveness on the offensive and not defensive ends (of particular importance in the case of KG).

Finally, as for the intangibles, though I don't typically tend to put much weight on these, I think they strongly cut Boston's way. I expect that everyone associated with the franchise, players and coaches alike, know that this is the year where opportunity knocks. And given that all of the Big Three players are apparently very competitive (and unselfish) guys, I think we will see unexpectedly positive outcomes....barring catastrophic injury.

Anyway, as I said, this year will be very interesting. If John Hollinger were offering the 46 games, I would take the over and be very happy. And I expect MikeZ would too.

[DLew]

Schtevie,

I didn't mean to imply that there was anything wrong with your calculations, just that one reason why the wins estimate may be extremely high is because diminishing marginal returns are not accounted for. This may or may not come into play, I was just offering it as one possible explanation if we do in fact believe the estimate provided by those numbers to be too high.

[Mike G]

Using just last year's numbers to predict next year's performance might be questionable. All of the Big3 had worse team records than the year before, by an avg 5 W. Coincidentally, all had worse TS% than the year before, by an average of .030 . All 3 had Ast/40 rates last year lower than their career highs (avg -1.2).

I think all of these guys could improve their Ast rates and TS% with one another in the lineup. All 3 have peaked at 5.3-6.0 Ast/40, so they may not need a classic PG.
_________________
40% of all statistics are wrong.

[Statman]