As promised, here are this year's team stats for every team in the NBA (full spreadsheet here):
And here is the sortable chart, still with the double scrolling issues (full spreadsheet here):
Here are my thoughts and observations:
- The Raps were not a good-shooting team. They were okay at shooting twos (10th in 2PT%), but the inability to take (29th in 3PT tendency) and make (30th in 3PT%) threes was very damaging. This lead to an adjusted field goal percentage that placed 21st in the league.
- Despite this, they were pretty good at jacking it up: 10th in FGA/48.
- In terms of free-throws, the team was middle of the pack in attempts, at 16th, and 17th with respect to free-throw tendency. But the team was 22nd in FT%.
- Put it all together, and the team was 21st in true shooting percentage.
- I can imagine casual fans saying: "true shooting percentage? That made up stat probably means nothing!" So which teams were tops in the league in TS%? In order: DEN, MIA, SAS, DAL, BOS, NYK, OKC, PHO, ORL, and HOU. Only two of those teams missed the playoffs - and the Rockets and Suns would've been the 6th and 8th seeds, respectively, if they were in the East.
- After a promising start to the season the Raptors finished 22nd in rebounds per 48 mins. The team was 8th in offensive rebounds per 48 minutes (and 8th in offensive rebounding percentage), so the culprit was poor defensive rebounding (27th in defensive rebounds per 48 and 25th in defensive rebounding percentage).
- Toronto placed 8th in turnovers (not a good thing) and 21st in steals.
- All told, the Raptors were 25th in net possessions (REB+STL-TO). Teams that do well in this area are generally successful. The top ten teams were: LAL, CHI, HOU, PHI, OKC, MEM, SAS, IND, DEN, and MIL. Of these teams, only the Rockets and the Bucks missed the playoffs; the Rockets I've already mentioned, and the Bucks were hampered by being the worst-shooting team in the league (and had a little bad luck).
- As far as "helpers" go, the team was 14th in assists/48 minutes, 26th in blocks/48 minutes, and 5th in fouls/48 minutes (not good).
- To top it all off, the Dinos were also the 9th most unlucky team, as measured by the difference between the number of games the team actually won and the number of games we would expect them to win (based on the team's point differential).
- Devin
I am only recently acquainted with Wins Produced but am mostly convinced of its utility. I would agree that luck has much to do with the discrepancy between Wins Produced and Actual Wins that you have listed here. However, some of this discrepancy must arise from flaws in the model. I couldn't help but notice that this season's Spurs was especially lucky while the Magic was particularly unlucky. I recently read the summary of John Huinziga's "The Value of a Blocked Shot" in SCORECASTING, which argues that a Tim Duncan blocked shot in past seasons has been around twice as likely to create a turnover than a Dwight Howard blocked shot. Thus, the marginal value of a Duncan block vs. a Howard block could actually differ by a factor of two. I wonder how much of the relative "luckiness" of the Spurs compared to the Orlando could be explained by the difference in the values of blocks of these two individual players. My back-of-the-napkin calculations seem to indicate that this could difference could explain away 1.5 of the Spurs' lucky wins and 1.5 of the Magic's unlucky losses.
ReplyDeleteAnonymous:
ReplyDeleteThat's a great comment - I wish you had used a pseudonym so I would know when you reply in the future.
There are many topics to tackle in your comment. The first is expected wins as compared to actual wins. Wins Produced is not the only metric that makes the distinction between expected and actual wins; in fact, if you go to Basketball-Reference and look at this year's Spurs team, you'll see something called "Expected W-L", which uses the same principles that Wins Produced uses. The general idea is that point differential is a better indication of the quality of a team than its actual record, and for whatever reason, some teams miss their expected records.
The question is, though, what is that reason? You have a hypothesis: flaws in the model. I'd say that the fact that both Wins Produced and the Pythagorean method come up with very similar results means that the flaw is not unique to Wins Produced. No, I'd attribute it to something else: randomness and blowouts. Sometimes crazy things happen in sports, and those crazy things can influence the outcome of a game. A few lucky bounces here and there, some unfortunate calls, and bam: a good team can lose to a team that it should have beaten.
There's also the problem of blowouts. Blowouts can severely affect a team's statistics, but whether a team wins by just one point or 50, a win's a win. Say a team wins a game by 51 points, then loses their next game by one point. If you sum up the differential for both games, on average the team won by 25 points! But of course they actually had a blowout win and a very close loss. In this case, the difference between expected wins and actual wins would be 1 win (50% off!) My hunch is that this is the major problem with expected records.
I do like your thinking on the blocks - clearly blocks where your team retains possession of the ball are more valuable than blocks that are sent sailing out of bounds (I always get mad at players who smack the ball into the stands when they could simply suck the ball up and start a fast break). But changing the value of blocked shots would not affect the Pythagorean method of expected record. As for Wins Produced, the story is a bit more complicated - and I'd have to verify this with some people who understand the math better than I do - but I suspect that differentiating the value of blocked shots would offer very little change (if any change at all).
Devin, would you be at all interested in posting a downloadable version of this Google Doc? I'd like to take it with me so I can tease out some of the Detroit-specific stuff. (You could always email me if you don't want to publish it!)
ReplyDelete