… was the fourth best rookie in the preseason.
WE'RE NOT HERE TO COOPERATE
October 25, 2010
Manny Harris ... By One Set of Stats ...
October 24, 2010
WP48 Analysis of Cavs
Here’s a WP48-driven preview of the Cavaliers. The writer of the piece, Arturo Galleti (I thought Mesa reader Holland was supposed to be writing for Dave Berri’s Wages of Wins Journal), figures the Cavs to win 36 games.
We’ll see how close the projection turns out. I suspect it won’t be that close since WP48 doesn’t account for defense when utilized as a predictor of future success. The handy “team adjustment” doesn’t come into play until the games get going.
I think you can sense that I have a lot of reservations about WP48 as a statistical model for selecting players - although I do fundamentally agree with the attributes it places the most emphasis on (scoring efficiently, not turning the ball over, and rebounding). At the same time, the model drastically undervalues shot creation, which is something this season’s Cavaliers may be particularly challenged by.
And finally Arturo - who I’m betting has never been to Cleveland - should have his spreadsheet taken away for saying that Cleveland was “unsurprisingly” labeled the most miserable city in the US by Forbes.
July 16, 2010
Dirty Projections: The Do-Over
After a discussion with WP48 expert Holland, I realized that I semi-botched my projection of the Cavs’ team performance this upcoming season. I wanted to take some time tonight to correct the analysis.
For those of you actually interested in advanced statistics, the main error involves some of the nuances of WP48. The purest form of the metric involves position adjustments. For example, a point guard’s WP48 performance isn’t weighted in exactly the same way as a center’s. Without adjusting, the metric heavily favors big men because they’re so much more likely to, say, get rebounds and shoot a higher percentage, as well as much less likely to turn the ball over (all important components of the entire scheme).
My source for WP48 was Basketball Reference, which is still a fantastic resource for advanced stats. However, if anyone out there wants to use it, keep in mind that the WP48 figures they give appear to be unadjusted.
OK, with the explanation out of the way, here are the adjusted WP48 totals for the roster hold-overs from last season. I kept all of the other parameters the same as in the original post.
Varejao = .181 WP48 x 36mpg x 82 games = 11.1 wins
Mo = .116 WP48 x 36mpg x 82 games = 7.1 wins
Moon = .191 WP48 x 36mpg x 82 games = 11.7 wins
Hickson = .123 WP48 x 20 mpg x 82 games = 4.2 wins
Gibson = .042 WP48 x 9 mpg x 82 games = 0.6 wins
Jamison = .194 WP48 x 36 mpg x 82 games = 11.9 wins
Parker = .081 WP48 x 36 mpg x 82 games = 5.0 wins
Powe = .000 WP48 x 9 mpg x 82 games = 0 wins
Green = .074 WP48 x 9 mpg x 82 games = 1.1 wins
Jawad = -.069 WP48 x 9 mpg x 82 games = -1.1 wins
Telfair = .020 WP48 x 4 mpg x 82 games = 0.1 wins
TOTAL PROJECTED WINS PRODUCED BY 2010-11 CAVS: 52 wins
So after adjusting for position, WP48 actually projects the Cavs to win one additional game beyond what the unadjusted numbers projected. This is mostly due to dramatically increased ratings for Moon and Jamison. Their gains more than offset significantly lower ratings for players like Mo, Gibson, Parker, Powe, and Jawad.
The other error I made in Monday’s post had to do with the expected error involved in the calculation.
I said at the time that WP48 predicted win totals to about 80% accuracy. I undershot. It’s actually (allegedly) accurate to within 94%. The error range is about +/- 1.5 games. Essentially, what this means is that the 2010-11 Cavs have a 94% chance of winning somewhere between 50.5 and 53.5 games.
Let me repeat that: if no changes are made to the current roster, and the rotation looks something like what I’ve outlined above, the Cavs allegedly have a 94% chance to win at least 50 games in their first season without LeBron. That’s +20 wins higher than what even the most optimistic sports pundit (that I’ve seen, anyway) has predicted for the team next season.
Of course, if Byron Scott chooses to start Jawad at 3 over Moon, this entire projection goes out the window. And I might, too. (Though I live on the first story, so the effect would probably be pretty muted.) But for now, things are definitely looking up.
Now, as I noted in my original post on this topic, I still think we’re running a significant risk in these projections if we only look at the numbers from last season. (Unfortunately, I don’t have adjusted career WP48 numbers.)
One of the points where I split with the WP48 philosophy has to do with synergy, for lack of a better term. The WP48 system assumes that player production is an inherent trait determined by the player’s skill level. In other words, he will produce roughly the same over time regardless of who the other 4 men on the court next to him are, not to mention who the 5 defending him are.
I, on the other hand, believe that a player’s teammates have a significant effect on what he’s capable of doing. For example, a spot-up 3-point shooter gets much better looks if he’s playing with someone on the front line who demands a double-team. Theoretically, he should be more open for his shots and should drain a higher percentage. I would argue that Mo Williams has been a great example of this. In his 2 seasons with the Cavs, he’s shot 43.6% 3P and 42.9% 3P, respectively. In the previous 4 seasons in which he’d played significant minutes with Milwaukee, his career high from beyond the arc was 38.5%. That’s a 4.5-5% uptick since starting next to LeBron.
This will be one of the elements to keep your eye on this coming season. Can the Cavs players find a way to do what they did with LeBron? As Mike pointed out, their current roster has a very low number of guys capable of creating shots for themselves. We’ll see how that affects things, along with the new up-tempo system Scott is already implementing in the Vegas Summer League. (Side note: WP48 disregards pace, so in theory this shouldn’t affect the projected win total. I remain skeptical on this point, but that’s a discussion for another day.)
That said, if things proceed as I expect, I still plan on checking into the over/under on Cleveland’s win total for 2010-11 if I’m in Vegas before the season starts.
All right, I think this is the last numbers-centric post I’m doing for a little while. Pseudo-philosophy, pseudo-psychology, and pseudo-humor back next week.
-T
July 12, 2010
The 2010-11 Cavs: Dirty Projections
Since LeBron announced his decision to “take [his] talents to South Beach” last Thursday night, I’ve seen a couple of different people pose questions to NBA pundits about the Cavs’ fate for next year. Specifically, how many games should we anticipate they’ll be able to win without James in the line-up?
The consensus answer from those pundits seems to be “not many.” Big surprise. Specifically, 30 wins seems to be about the best case scenario the pundits believe Cavs fans can hope for.
Of course, the pundits who were asked this question weren’t people who I would expect are using advanced statistics in their projections. I, on the other hand, anticipate that the team will be better than most of the rest of the people in the country would think, because I know how well many of the players on the roster rate in metrics that the average fan (and average pundit) aren’t considering. For instance, when the idea of Toronto signing-and-trading Chris Bosh to the Cavs was being heavily discussed, our buddy Chad Ford (one half of T.I.T.) made the point that the Cavs could potentially use the deal to unload a bad contract like…Anderson Varejao’s. Whereas we here at Mesa have regarded Andy as the Cavs’ second-best player since about 2006.
But since “better than these other guys think” still isn’t a good answer to “How good do you think the 2010-11 Cavs will be?”, I decided I would use Wins Produced Per 48 Minutes (WP48) to build a rough model that would spit out some actual numbers.
Here are the parameters I set:
- I can only use players who are actually under contract for next year
- I’m excluding Delonte because Chris Grant has all but said that he will be cut or traded prior to his contract becoming guaranteed early next month
- There is a total of 240 minutes per game to distribute among the roster, since 5 players have to be on the floor for 48 minutes (5 x 48 = 240)
- The starting 5 will be Mo at 1, Parker at 2, Moon at 3, Jamison at 4, Varejao at 5
- Each starter will average 36 minutes per game.
- JJ will be the 6th man, and will average 20 minutes per game
- Sebastian Telfair will be the “most likely to DNP” and will only average 4 minutes per game
- All of the other subs on the roster will average 9 minutes per game
- Each player will play all 82 games
If use each player’s WP48 from the 2009-10 season and the minutes-per-game assumptions above, we get the following projection.
Varejao = .179 WP48 x 36mpg x 82 games = 11 wins
Mo = .136 WP48 x 36mpg x 82 games = 8.4 wins
Moon = .131 WP48 x 36mpg x 82 games = 8.1 wins
Hickson = .123 WP48 x 20 mpg x 82 games = 4.2 wins
Gibson = .117 WP48 x 9 mpg x 82 games = 1.8 wins
Jamison = .115 WP48 x 36 mpg x 82 games = 7.1 wins
Parker = .103 WP48 x 36 mpg x 82 games = 6.3 wins
Powe = .097 WP48 x 9 mpg x 82 games = 1.5 wins
Green = .083 WP48 x 9 mpg x 82 games = 1.3 wins
Jawad = .069 WP48 x 9 mpg x 82 games = 1.1 wins
Telfair = .020 WP48 x 4 mpg x 82 games = 0.1 wins
TOTAL PROJECTED WINS PRODUCED BY 2010-11 CAVS: 51 wins
Now, supposedly the WP48 statistic only accounts for 80% of wins. If we assume that to be true, then we should adjust this 51-win total by dropping it 20%.
TOTAL ADJUSTED PROJECTED WINS PRODUCED BY 2010-11 CAVS: 41 wins
The problem here is that it’s flawed to assume that these players are going to be able to play the same way without LeBron that they did with LeBron. In that sense, it’s misleading to use their 2009-10 WP48 stats in a projection. It’s likely more accurate to use their career WP48 stats.
Admittedly, this revised model would still be a very rough one, since there are some players—Andy, Jawad, Boobie, Hickson, and Danny Green—who have played with LeBron for their entire NBA careers. But it’s the best we can do right now.
When we plug career WP48 stats into the model, we get these numbers:
Varejao = .156 WP48 x 36mpg x 82 = 9.6 wins
Mo = .095 WP48 x 36mpg x 82 = 5.8 wins
Moon = .122 WP48 x 36mpg x 82 = 7.5 wins
Hickson = .116 WP48 x 20 mpg x 82 = 4.0 wins
Gibson = .107 WP48 x 9 mpg x 82 = 1.6 wins
Jamison = .118 WP48 x 36 mpg x 82 = 7.3 wins
Parker = .100 WP48 x 36 mpg x 82 = 6.2 wins
Powe = .183 WP48 x 9 mpg x 82 = 2.8 wins
Green = .083 WP48 x 9 mpg x 82 = 1.7 wins
Jawad = .068 WP48 x 9 mpg x 82 = 1.0 wins
Telfair = .013 WP48 x 4 mpg x 82 = 0.1 wins
TOTAL PROJECTED WINS PRODUCED BY 2010-11 CAVS: 48 wins
Again, if we drop this “best case” number by 20%, we get something likely much closer to reality.
TOTAL ADJUSTED PROJECTED WINS PRODUCED BY 2010-11 CAVS: 38 wins
In short, WP48 suggests that the pundits are, as usual, low-balling the Cavs. It’s possible that the team could still be competing for the final playoff spot in the East next season, even if they made absolutely no adjustments to their roster. Granted, this would take a remarkable level of health and consistency from all players involved, but it’s not out of the question.
However, we’re undoubtedly going to see changes to the roster in the coming weeks. My personal hope is that those changes are geared toward making the team better in the long run rather than the short. But as it stands now, we should feel comfortable that the Cavs are not going to be a 15-win team. Cold comfort if you consider where we thought the team was only a few months ago, but the first step towards improvement is realism. Hopefully the front office agrees.
-T
July 8, 2010
A Look at NWO: The Wolfpack Through WP48
Interesting look at a Wade-James-Bosh trio from the WP48 guys, via the Wall Street Journal.
The RSS for this blog is http://www.josemesaisdead.com/rss
1 of 2
next page
