Before the season started I predicted that the Boston Celtics would defeat the San Antonio Spurs in the NBA Finals. This prediction was motivated by two issues.
1. My forecast – which required that I guess how many minutes each player would play – indicated that Celtics were clearly the best team in the East and the Spurs were probably the best team in the West.
2. I thought it would be interesting if Bill Simmons – who argued last spring that Tim Duncan was clearly better than Kevin Garnett – got to root for KG and his Celtics against Duncan.
And of course, #2 was more important than #1. :)
Each Team at the 33% Mark
As I write, each team is about 1/3 of the way through the regular season. And at this point, each team is currently the best in their respective conference. As expected, the Celtics are clearly better than any team in the East. And the Spurs are only slightly better than any team in the West.
On the surface it looks like each team is performing as we would expect given what the players on each team did last year. But when we look at the current projection for each team this season – reported in Table One — we see that the Celtics look quite different from what we would have expected in October.
Table One: Projecting the Celtics and Spurs
Table One actually offers two projections of the Celtics and Spurs in 2007-08. The first looks at how many wins we could expect if each player plays as well as he did last year. This is the same as the pre-season projection, except now we have a better idea of how many minutes each player is going to play. The second estimates how many wins each player will produce if he keeps playing as well as he has this season.
The Spurs Story
For the Spurs, each projection is basically the same. Last year this team was led by Tim Duncan, Manu Ginobili, Tony Parker, and Brent Barry. This year, each of these players is again very productive.
The one surprise is Fabricio Oberto. Last year he posted a 0.117 WP48 [Wins Produced per 48 minutes]; a mark that’s slightly above average. This year his shooting efficiency and rebounds have increased while his turnovers have declined. Consequently his WP48 has risen to 0.242. Although his minutes are limited, this substantial increase in per-minute performance does improve the team. And that improvement would show itself in the standings if this team could manage to stay healthy. Unfortunately, each of the big four on this team has already missed a game. If injuries continue to be a problem, the Spurs might have trouble reaching the Finals.
The Celtics Story
Beyond the play of Oberto, the Spurs of this year – when healthy — look very much like the Spurs of last year. The same story cannot be told of the Celtics.
Last year the Celtics only won 24 games, the second worst mark in the Association. In the off-season, though, the team added Kevin Garnett, Ray Allen, and James Posey; three players who were above average last year. With this talent added to Paul Pierce and Rajon Rondo, it was expected this team would rise to the top of the conference.
In fact it was expected that this team would be competitive with the defending champions. The Spurs projected wins – based on last year’s performance – was about 63 wins. Given what Boston’s players did last year, we would expect this team to win about 64 games. Again, this mark would just about match the Spurs and clearly lead the Eastern Conference.
Although the Celtics are clearly on top of the conference, the team has gone beyond leading the East to crushing the league. Boston has already matched last season’s victory total, winning 24 of its first 27 games. Such a pace is consistent with a team winning 73 regular season games. And when we look at efficiency differential – offensive efficiency minus defensive efficiency — the team looks even better.
The Bulls in 1995-96 won 72 games and posted a differential of 13.0. This is the best mark in the league since 1973-74. The Celtics currently have a differential of 14.9. Yes, the current Celtics are posting a better mark than the team considered the best in NBA history. To give this result even more perspective, the Spurs differential this season is 6.9 (which is very good). Still the Spurs mark is only about half of what we see from Boston.
The Celtics efficiency differential, which is a mark outside of our previous observations of the NBA, gives us a wins projection that is – not surprisingly – also outside our previous observations. Currently the Celtics are projected to win about 78 games. Given that this team has already lost three times, this projection is obviously too optimistic.
Before I get to what this means, let’s quickly review why the Celtics improved. A big part of this improvement comes from the play of Kevin Garnett. If Garnett repeated what he did last year in Minnesota, he would be posting a WP48 of 0.330 and be on his way to producing 19.9 wins. Each of these marks is very good, yet short of what he is doing this year (0.405 WP48 and 24.4 projected wins). Although we might see this as KG improving, it’s perhaps better to think of this as Garnett returning to what he was before the disaster that was 2006-07. If you recall, the Timberwolves changed coaches in the midst of last season in a vain hope of returning to the playoffs with Garnett. As noted at the time, the talent around KG was simply not that good in Minnesota. So playoff contention was not realistic.
Before the turmoil of last season, Garnett posted a 0.430 WP48 and produced 26.5 wins in 2005-06. If Garnett repeated that performance this season, and no one else improved, the Celtics would be on pace to win about 70 games this season.
Of course, about 70 is not the same as 78. When we look at Ray Allen, Paul Pierce, and Rajon Rondo we see that what we saw last year is pretty much what we are seeing this year. To see how the Celtics move from 70 to 78 projected wins, we have to turn to a player I didn’t think was going to help much, Eddie House.
Prior to this season, House had played seven NBA seasons for seven different teams. Across this career he had never posted an above average WP48 and had only offered 2.5 Wins Produced. This year, though, House has a much higher shooting efficiency, has grabbed more rebounds, and is getting more steals. As a result, he has already produced about two wins and is on pace to produce over six for the season.
What Does 78 Mean?
Okay, so we see that Garnett and House have led this team from a projection of 64 to 78 wins. What exactly does a forecast of 78 wins mean?
Jason Eshleman – a UC Davis anthropologist who often leaves comments in this forum (and authored a post on the Rockets a few days ago) – often reminds people in the comments that you cannot evaluate models by considering events outside what the model considers possible. In other words, it’s not important what Wins Produced tells us a team of Allen Iversons would do since we are never going to see an NBA team only employ a collection of speedy diminutive guards.
This lesson is important to remember when we look at the Celtics this year. The model predicts that a team with an efficiency differential of 14.9 would win almost every game the team plays. Clearly that’s not going to happen. The Celtics have already lost three times. Unfortunately, the data used to estimate the model did not include any team that was this much better than the Bulls of 95-96. So we should not be shocked that an observation outside the data the model was based on would give us such an odd result.
That being said, the projection of the Celtics may not be that far off. As noted, a 24-3 team is on pace to win 72 games. If the model says 78, and the team wins 72, such an error is not really that large.
Let me close by noting that I am not sure the Celtics are going to win even 72 games. The key players are still Garnett, Pierce, and Ray Allen. And these players entered the season old and are not getting any younger. An injury to any of these players would easily derail the Boston train. Of course, without such an injury, this train is headed for this team’s first title since 1986. Such a title would be sweet redemption for the much maligned Danny Ainge (general manager), Doc Rivers (head coach), and of course, Kevin Garnett.
– DJ
Our research on the NBA was summarized HERE.
The Technical Notes at wagesofwins.com provides substantially more information on the published research behind Wins Produced and Win Score
Wins Produced, Win Score, and PAWSmin are also discussed in the following posts:
Simple Models of Player Performance
What Wins Produced Says and What It Does Not Say
Introducing PAWSmin — and a Defense of Box Score Statistics
Finally, A Guide to Evaluating Models contains useful hints on how to interpret and evaluate statistical models.
Wages of Wins Journal 
Kent
December 29, 2007
Congratulations to commenter Jason Eshleman for receiving a shout out in this column!
Jason Eshleman … who often leaves comments in this forum … often reminds people in the comments that you cannot evaluate models by considering events outside what the model considers possible.
Ben Guest
December 29, 2007
In addition to House, the other revelation this year has been Rondo, who is putting together an incredibly efficient season.
SHC
December 29, 2007
I personally don’t think that Celts are this good at all.
I think that what we’re seeing is a result of motivation and weak schedule.
Basically, every significant player in the Celts is having a better season than last year. An interesting thing is that almost every one of them played for a pretty bad team last season.
Therefore, it’s not hard to imagine that they’re already exerting their maximum effort because they’re so sick of losing.
Another thing, they’ve only played 4 games against teams that could reasonably called ‘title contenders’ (Orlando twice, Lakers, and Detroit). Their record in these games is mere 2-2.
Garnett has historically been a ‘regular’ season player. If you look at his playoff numbers, they fall off quite a bit, especially his shooting percentage. He’s still a productive player, but this may tell you that he doesn’t have ‘another gear’ to elevate his game when the competition gets tougher.
It’s not unreasonable that their numbers will all fall to what they were last year (or worse) in playoffs. (I think that you said yourself that regular season and playoff perforamnces are not very correlated.). Then you could easily see this team losing even before they reach the finals.
magicmerl
December 29, 2007
Is it worthwhile looking at PlayoffWinScore to see if that changes the picture?
I would imagine that buckets are harder to come by and the competition is better, so everyone’s WS would come down slightly.
Is that the case?
dberri
December 29, 2007
SHC,
Playoff performance is about as correlated with regular season performance as this year’s regular season is correlated with last years. This is detailed in the book.
Kent
December 29, 2007
dberri writes, “Playoff performance is about as correlated with regular season performance as this year’s regular season is correlated with last years.”
Is this true even if you adjust for sample size? i.e. Playoffs are smaller sample so the correlation between playoffs and regular season will be lower just because of the noise in the small playoff sample compared with looking at this year’s regular season vs. last year.
dberri
December 29, 2007
For playoff performance we only considered players who played at least 10 games in the post-season. Much to my surprise, the two correlations are pretty similar.
SHC
December 29, 2007
dberri,
Thanks for correcting me. But last season’s performance may have higher correlation with this season’s performance than performance from 2-3 seasons ago.
In your analysis, did you only compare the same year’s playoff and regular season performances? That would exclude a lot of players (including most of years for Garnett) since only a select few would play 10 or more playoff games in any select year.
Or are you saying that you compared a player’s career playoff numbers with his career regular season performances?
Thanks
Kent
December 29, 2007
dberri, thanks for your reply to my comment.
I wrote “Playoffs are smaller sample so the correlation between playoffs and regular season will be lower just because of the noise in the small playoff sample compared with looking at this year’s regular season vs. last year.”
After I wrote that I realized I made a mistake. The lower playoff sample size wouldn’t lead you to expect the regular season to playoff correlation to be lower. It would of course increase the standard error but not bias the results.
Anyway, thanks for your reply and clarification on how you did the calculation. And sorry for posting something that was in error.
SHC
December 29, 2007
Kent,
I’d expect that if there are a lot of noise (higher standard deviation), then the correlation coefficient wouldn’t be as strong.
I think that you won’t see as many abberations in NBA as in baseball. You don’t see a 20 ppg scorer scoring only 10 ppg in 4 straight games. (However you see a .300 hitter going 0 for 20 in playoffs in baseball).
However, I’d expect some slight dropoffs in playoffs for most players compared to regular season, because of escalated intensity and competition.
Kent
December 29, 2007
SHC, you’re right! So I was actually right initially. And then my intended correction was wrong.
givensna
December 29, 2007
Tim Duncan reg season: 21.8/11.9/3.2/2.5/.509/.680/37.5 minutes per game
Tim Duncan playoffs: 23.8/12.5/3.5/.507/.697 /40.1 minutes per game
Kevin Garnett reg season: 20.5/11.4/4.5/1.7/.491/.780/38.3 minutes per game
Kevin Garnett playoffs: 22.3/13.4/5.0/.46/.76/42.6 minutes per game
Conventional wisdom and the media seem to think that Tim Duncan does have that ‘extra gear’ that SHC speaks of and KG lacks. Duncan does appear to average more in all of the relevant categories while maintaining his shooting %’s. However, those upticks can almost entirely be contributed to the increase in minutes he has in the playoffs versus regular season. KG’s FG percentage has dipped slightly throughout his career in the playoffs, but the story is still very similar.
I’m not finding any evidence of lack of gears or KG or evidence of an extra gear for TD.
-Andrew
SHC
December 29, 2007
givensna,
I computed (and commented on another post) these playoff win scores per 48 mins for both players.
Duncan : 16.3511 over 138 games
Garnett : 14.9070 over 47 games
I don’t know how wins produced is related to this, but I’m assuming that they are linearly correlated. Considering that Garnett’s regular season career WP48 is slightly better than Duncan’s, I think that Garnett’s dropoff is noticeable.
And if you analyze their games qualitatively, Garnett is basically a 7′ jump shooter (and he ain’t Nowitzki in accuracy). This means that he’s not very reliable as a first option in your offense.
Patrick
December 29, 2007
SHC,
“And if you analyze their games qualitatively, Garnett is basically a 7′ jump shooter (and he ain’t Nowitzki in accuracy). This means that he’s not very reliable as a first option in your offense.”
This ignores the fact that he led the league in rebounding a few years in a row, led all power forwards in assists in most of his seasons, and is a 1st team all-defensive player for most of his career. Calling him a poor man’s Dirk Nowitzki is a terrible comparison.
And, KG is actually quite good on the low block (throughout his career the criticism is that he is basically not a *fantastic* low-post player, like TD…yeah, those grow on trees…) — Dirk Nowitzki is actually a very poor low-block player.
TD’s contributions to a team are great defense, amazing low-post scoring. KG’s are great defense, great passing, and decent low-post production, and good mid-range shooting.
In addition, throughout the years, it has been far, far, far easier to double-team and shut down Kevin Garnett and not get punished by his teammates as has been the case with Tim Duncan, who got to play with David Robinson, Tony Parker, Manu Ginobli, and a collection of players who were famously good from behind the 3-point line.
You’re comparing apples and oranges.
Jason
December 29, 2007
Apples and oranges are both round fruit.
SHC
December 29, 2007
Patrick,
I was simply referring to the fact that he shoots a lot of jumpers from 15′ out (I read somewhere that 40% of his shots are from 15′ and out and 66 % from outside the paint area) and that his percentages from those areas are far worse than Nowitki’s.
He’s overall a far better player than Nowitzki no doubt.
magicmerl
December 29, 2007
Crossposted from the Bill Simmons link:
One factor that might not have been considered much by others is the different contracts.
Garnett too the fattest contract in history, and more power to him, because he was worth that much. But because of the salary cap that meant that the wolves *couldn’t* put talent around him to allow him to contend for a title.
Timmy on the other hand settled for a modest 16mil (he left about 6mil per year on the table when he signed his first post-rookie contract, money that essentially equals having Manu Ginobli on his team), and every time he resigns an extension he takes a pay cut back to 16 mil. (p.s. Bruce Bowen opted out of his contract to resign for a lesser amount a few years ago, also so that the spurs could resign Manu. now that’s a championship play).
Garnett is doing the same this time, which could bode well for his Celtics era. But he was a salary cap hog in minnesota, which is why he never had good teammates.
Does this make sense to anyone?
dberri
December 29, 2007
Although people often claim it is the salary cap and Garnett’s contract that was the problem, I remain skeptical. McHale had plenty of opportunities to pick better players in the past decade. He consistently made bad choices.
magicmerl
December 29, 2007
Well, you have to get the good players, but you need to keep them too. The spurs probably wouldn’t have kept Ginobli or Parker without Duncan / Bowen buying into taking less money.
Just saying.
steve
December 30, 2007
Now that Ray Allens shot is back to normal (38% shooting from the 6th game through when Doc finally sat him out for 3 games due to a minor lingering injury…… 50.2%shooting since he came back from that rest) and he is back to his normal self the Celtics are just flat out DEADLY and even better than they were from November 11th til december 16th.
That is huge, this team is getting better, MUCH better.
steve
December 30, 2007
SHC, apparently you havent watched Kevin Garnett play this year.
He is our primary interior option on offense, he plays in the low-post 75% of the time, jumpers are about a quarter of his game this year. We run the offense through him (Al Jeffersons old role from last season), he puts the pressure on the defense and scores every single time the opponent decides not to double team. When they do double down on him and trap he finds the open man as a great passer, whether its Allen or Pierce wide open behind the arc or cutting to the rim, Posey or House for the open 3, Rondo open for a midrange jumper or cutting to the rim, or Perkins on an interior pass for an open dunk.
I dont know as much about Garnett in the past and I have heard of that reputation a few times before, but as quoted at SI “NBA general managers are extremely surprised at how much more KG plays in the lowpost, his whole game is focused there compared to in Minnesota where only about half his game was in the low-post and the rest was midrange/elbows”…… that and the other most surprising thing to the GMs was the Celtics great defense.
So you are wrong, watch the Celtics play and you see that 75% of what you see from Garnett is in the Pivot. He still offers up his versatility and has that wired into his game, but the pivot is where the majority of his play is. That is his job offensively on this Celtic team. And he does his job very very well.
steve
December 30, 2007
The KG i’ve seen all year has been completely unstoppable in the lowpost. Then the other team has to double down and trap, and KG is a great passer (Best passing big man I’ve watched since Larry Bird) and finds our deadly open shooters (House, Posey, Ray Allen, Pierce) for the 3-pointer or Rondo or Tony Allen for the midrange open jumper or Perkisn for a wide open dunk/layup, or a cutter slashing to the rim (Usually Pierce, Rondo, Ray, or Posey)….. it is a win/win every single time.
Not to mention it gives Pierce and Allen single coverage to work against, to play 1 on 1 where they are absolutely deadly.
SHC
December 30, 2007
steve,
It’d be nice to support your theory with shot charts or something. But I haven’t been able to find any cumulative shot charts. You can go to NBA.com for his shot charts from the past 3 seasons or so, however.
If you look at past 3 games:
http://sportsline.com/nba/gamecenter/shotchart/NBA_20071226_BOS@SAC
http://sportsline.com/nba/gamecenter/shotchart/NBA_20071227_BOS@SEA
http://sportsline.com/nba/gamecenter/shotchart/NBA_20071229_BOS@UTA
You’ll see that the majority of his shots came from pretty far for a post player.
SHC
December 30, 2007
Actually, my bad. There’s a cumulative shot chart from NBA.com for this season.
http://www.nba.com/hotspots/
It seems that KG has only taken about 118 shots out of 371 in the deep post. He took 135 shots from 15′ and out and the rest of them in between (I think that majority of them would be outside the paint).
So your theory that KG spends more time in the post than he’s done in the past seems to be wrong.
SHC
December 30, 2007
Compare that to the other premier big men like Tim Duncan, Dwight Howard, and Carlos Boozer – they literally live in the deep post.
I bet that’s the reason KG’s shooting percentage falls in the playoffs.
Animal
December 30, 2007
What I don’t get is why somebody would start posting after as “Animal” after I switched my name to Animal. I decided to stop posting, and than some other animal kept on posting. This is my last post, just letting people know that there had been two animals posting
Evan
December 30, 2007
You have to stop referring to the NBA as ‘the Association.’ Seriously.
dberri
December 30, 2007
Evan,
Why?
Sam Cohen
December 30, 2007
(mostly off-topic, but not entirely):
Dave- your aside about the team of iversons made me curious, have you applied WP to college basketball at all? It appears to me (anecdotally) that college has much greater variation in the strategies employed, so we’d be much more likely to see a team of “five iversons” (or something moderately close) and we could get a sense of if the WP model does consider a team of 5 iversons possible (and maybe something in general about the importance of position adjustments).
Does that make sense?
Kent
December 30, 2007
Evan, it is very common to refer to the NBA as “the association.” For example, see http://www.nba.com/aroundtheassociation , an NBA.com page entitled “Around the Association.” I also hear it on ESPN all the time. It’s not confusing at all. What “association” do you think Dave is refering to when he says it? Just assume he’s always alluding to the NBA.
dberri
December 31, 2007
Hi Sam,
First of all, I liked your recent analysis of the Iverson trade. For those that missed this, here is the link.
http://sixerpride.blogspot.com/2007/12/365-days-later-or-so.html
As for your question, Erich Doerr — who wrote a guest column — has been looking at college performane. And I have also. One issue is figuring out positions played. You are correct that positions are less defined in college. I would suspect that we would still see clear differences between big men (C, PF) and guards. But the differences, as you suggest, may not be as great as we see in the NBA. So the position adjustment used in the NBA would not apply.
givensna
December 31, 2007
Please continue to say “the Association” or “the L.”
Happy New Year
-Andrew
William
January 5, 2008
Data from 82games.com suggest that in “clutch time”, KG takes jumpshot most of the time (80%) , while Duncan takes “close shot” most of the time.
If you look at the “clutch play stats” of last year, you will find that Duncan is indeed very clutch while KG… CHOKES!