And now for the small forwards…No surprises out East, LeBron and Pierce are the top two SF in the league, well ahead of Wilson chandler. In the West, Gay, Melo (if he’s still around), and Durant are the top three starters. Dorell Wright is well down on the list (#27), but there are few SF in the league that would be significantly better, and I’ve mentioned the bulk of them already. I’ve said this before on the GSOM forum, but I think Thaddeus Young might be an intriguing guy to look at. He’s vastly improved this season, and at 22, has probably not even reached his full potential. (Of course, Philly likely realizes this, too.)
Small Forward Rankings According to ezPM (1/12/11)
| RANK | NAME | TEAM | POSSESSIONS | ezPM100 | O100 | D100 | REB100 | OEFF | USG | ORR | DRR |
| 1 | LeBron James | MIA | 3009 | 5.70 | 3.51 | 2.27 | -0.07 | 11.30 | 31.02% | 19.62% | 78.60% |
| 2 | Paul Pierce | BOS | 2533 | 4.10 | 3.20 | 0.83 | 0.07 | 15.78 | 20.27% | 10.32% | 86.10% |
| 3 | Rudy Gay | MEM | 2933 | 3.07 | 0.64 | 1.89 | 0.53 | 2.87 | 22.43% | 29.15% | 76.47% |
| 4 | James Jones | MIA | 1702 | 2.61 | 1.85 | 1.41 | -0.64 | 18.39 | 10.04% | 13.46% | 80.43% |
| 5 | Matt Barnes | LAL | 1555 | 2.26 | -0.01 | 1.06 | 1.22 | -0.09 | 15.92% | 36.91% | 76.92% |
| 6 | Carmelo Anthony | DEN | 2223 | 2.24 | -0.07 | 0.59 | 1.72 | -0.23 | 30.15% | 34.48% | 84.91% |
| 7 | Kevin Durant | OKC | 2819 | 2.22 | 1.84 | 0.48 | -0.10 | 6.57 | 28.01% | 19.38% | 77.88% |
| 8 | Tracy McGrady | DET | 1375 | 1.97 | 0.15 | 1.41 | 0.41 | 0.82 | 18.35% | 15.56% | 86.24% |
| 9 | Thaddeus Young | PHI | 1838 | 1.86 | 0.80 | 1.24 | -0.17 | 4.25 | 18.82% | 34.03% | 62.94% |
| 10 | Wilson Chandler | NYK | 2735 | 1.37 | 1.46 | 0.18 | -0.27 | 8.14 | 18.00% | 21.50% | 79.05% |
| 11 | Richard Jefferson | SAS | 2455 | 1.00 | 1.15 | 0.25 | -0.41 | 7.87 | 14.64% | 17.39% | 76.00% |
| 12 | Marvin Williams | ATL | 1747 | 0.86 | 0.73 | 0.50 | -0.37 | 4.76 | 15.31% | 20.41% | 78.29% |
| 13 | Andrei Kirilenko | UTA | 2259 | 0.84 | -0.28 | 1.25 | -0.13 | -1.76 | 16.19% | 27.78% | 70.67% |
| 14 | Shane Battier | HOU | 2287 | 0.60 | -0.05 | 0.71 | -0.06 | -0.38 | 12.07% | 21.05% | 76.13% |
| 15 | Luol Deng | CHI | 2781 | 0.58 | -0.59 | 0.92 | 0.26 | -2.94 | 20.16% | 28.06% | 72.54% |
| 16 | Jared Dudley | PHX | 1645 | 0.46 | 0.82 | 0.34 | -0.71 | 5.47 | 15.07% | 25.71% | 65.49% |
| 17 | Grant Hill | PHX | 2250 | 0.30 | 1.71 | -1.02 | -0.39 | 9.32 | 18.37% | 30.20% | 65.56% |
| 18 | Tayshaun Prince | DET | 2286 | 0.27 | 1.14 | -0.92 | 0.06 | 5.56 | 20.52% | 23.98% | 74.67% |
| 19 | Quentin Richardson | ORL | 1144 | 0.11 | -1.10 | 0.28 | 0.93 | -7.92 | 13.87% | 24.42% | 81.72% |
| 20 | Ron Artest | LAL | 2304 | 0.02 | -1.01 | 1.35 | -0.32 | -7.69 | 13.13% | 29.71% | 58.26% |
| 21 | Gerald Wallace | CHA | 1968 | -0.13 | -2.18 | 1.09 | 0.97 | -10.22 | 21.38% | 27.52% | 81.87% |
| 22 | Danny Granger | IND | 2267 | -0.14 | -0.91 | 1.55 | -0.78 | -3.46 | 26.45% | 20.71% | 70.45% |
| 23 | Caron Butler | DAL | 1765 | -0.20 | -1.34 | 1.15 | -0.01 | -6.14 | 21.78% | 18.49% | 83.02% |
| 24 | Shawn Marion | DAL | 1766 | -0.23 | -0.94 | 0.71 | -0.00 | -4.99 | 18.82% | 32.28% | 67.76% |
| 25 | Omri Casspi | SAC | 1555 | -0.38 | -1.01 | -0.08 | 0.70 | -6.16 | 16.33% | 27.37% | 81.97% |
| 26 | Nicolas Batum | POR | 2136 | -0.57 | -0.63 | -0.03 | 0.09 | -3.81 | 16.49% | 32.65% | 68.61% |
| 27 | Dorell Wright | GSW | 2789 | -0.75 | -0.73 | -0.07 | 0.04 | -3.99 | 18.36% | 23.43% | 77.55% |
| 28 | Wes Matthews | POR | 2424 | -0.85 | 0.08 | -0.13 | -0.79 | 0.37 | 21.02% | 16.56% | 71.31% |
| 29 | Al-Farouq Aminu | LAC | 1152 | -0.97 | -3.68 | 1.13 | 1.58 | -18.52 | 19.86% | 36.56% | 80.18% |
| 30 | Kyle Korver | CHI | 1499 | -1.09 | -0.57 | 0.88 | -1.41 | -3.53 | 16.04% | 7.32% | 70.73% |
| 31 | Danilo Gallinari | NYK | 2411 | -1.11 | 2.45 | -0.97 | -2.59 | 15.72 | 15.57% | 15.57% | 56.57% |
| 32 | Chase Budinger | HOU | 1244 | -1.34 | -1.14 | -0.26 | 0.07 | -6.05 | 18.93% | 23.68% | 76.53% |
| 33 | Andres Nocioni | PHI | 1306 | -1.36 | -0.61 | -0.50 | -0.26 | -3.63 | 16.72% | 15.15% | 80.30% |
| 35 | Trevor Ariza | NOH | 2501 | -1.54 | -2.51 | 1.21 | -0.25 | -14.74 | 17.03% | 16.48% | 79.59% |
| 36 | Ryan Gomes | LAC | 2022 | -1.59 | -0.67 | -1.08 | 0.15 | -5.60 | 11.89% | 22.90% | 73.39% |
| 37 | DeMar Derozan | TOR | 2360 | -1.80 | 0.09 | -1.71 | -0.18 | 0.45 | 19.81% | 18.29% | 79.09% |
| 39 | Al Thornton | WAS | 1415 | -2.12 | -1.12 | -0.40 | -0.60 | -6.24 | 17.89% | 26.12% | 66.67% |
| 40 | Rasual Butler | LAC | 1163 | -2.53 | -2.80 | 0.37 | -0.09 | -18.89 | 14.84% | 11.76% | 81.43% |
| 41 | Travis Outlaw | NJN | 2165 | -2.91 | -2.05 | -0.54 | -0.32 | -12.23 | 16.74% | 14.38% | 81.33% |
| 42 | Hedo Turkoglu | PHX | 1373 | -2.93 | 0.78 | -1.78 | -1.93 | 5.16 | 15.10% | 15.65% | 61.95% |
| 43 | Linas Kleiza | TOR | 1843 | -3.43 | -2.90 | -1.54 | 1.00 | -14.50 | 19.98% | 29.36% | 81.29% |
The City 
How do explain James Jones’ 4th place finish? Doesn’t that challenge the accuracy of the model? Clearly the guy is benefitting a lot from his teammates’ ability to draw defensive attention…
It doesn’t “challenge” the model. The model is doing what it is supposed to do. James Jones is incredibly efficient, for what he does. But he does it at extremely low usage, and with very low shot creation ability.
So, numerically, he comes up fourth, but I would heavily discount that. If he went to another team, I would predict his value would drop quite a bit.
I’m thinking about ways to take into account the quality of teammates, but it’s not a simple issue, obviously.
First off let me say I love your blog and I love that you’re trying to apply statistical models to this game. It’s a very interesting thing to do and you do it very well! So don’t take my criticism the wrong way.
Regarding James Jones’ 4th place finish in SF rankings, I suppose I shouldn’t say that it challenges the model – of course the model can be seen as doing exactly what it is supposed to do (tautologically). I would say, however, that it challenges the VALUE of the model in evaluating basketball players, in that it inflates James Jones’ importance. James Jones is doing a good job being efficient in his role, but I don’t think anyone would argue if I say his role is “easier” than say, Kevin Durant’s.
Aren’t we trying to make value judgments about these players based on the numbers? I would argue that we do. If no, then why do we do it?
Isn’t it a concern that the model ranks Jones higher than Durant? Doesn’t that indicate that there may be other problems as well?
Eric, right now the model says his ezPM is higher than Durant. I think the question you want to ask me is whether I believe he is better than Durant. My answer would be emphatically no. There are several reasons for this. First, Jones is not a starter. Some of his value is inflated because he will come in against second units. The model is not currently accounting for this, but I think there are definitely ways to address this issue, for example, adjusting the values used for PPP. Second, Jones has incredibly low usage compared to Durant. The model does not currently penalize players for low usage or reward players, such as Durant, for high usage. Once again, this is something I am trying to think about how to incorporate into the model. Third, look at the defensive component. I have little doubt that Durant is a better defender overall, but Jones is boosted in that component by playing in Miami. To begin to address this issue, next on my list of features to add to the model is counterpart defense.
You might have noticed that ezPM is split into three components: offense, defense, and rebounding. This is so people can look at each individual component and consider the the player in the context of those areas. When looking at the data, don’t just look at the ezPM. Look at all the numbers, they will tell you a lot more, I think.
Lastly, and this is along the lines of what Crow has said before, you should probably never rely solely on one boxscore-based metric, such as ezPM or WP, without also looking at other metrics, especially adjusted +/-. For example, if you look at two-year APM, Durant is much higher than Jones. You can use both a statistical +/- together with a box score based +/-, such as mine, to give a more balanced valuation.
As they say in the stock game, DYODD.
BTW, for reference, James Jones’ ezPM100 last season was -1.25 (in just over 800 total possessions).
Just to go a little deeper, consider the following:
Let’s assume that players surrounded by good teammates have “inflated” numbers. So, James Jones looks better than he is. However, LeBron and Wade are also teammates. Which one of them have “inflated” stats? In other words, how do we distinguish between players who are actually doing what we think they are doing, versus players who are feeding off of others.
Unfortunately, the stat sheet or play-by-play doesn’t tell us, “So, yeah, on this play James Jones was wide open in the corner, because LeBron was getting a double team and kicked it out…”
What about trying to incorporate a player’s performance against his historical norm? Or something of that ilk. It’s obviously a tough choice.
Or maybe the past 5 years, since rookies do tend to improve over the next few seasons. And maybe only incorporate this component when a player has played at least 4 seasons or something.
Just a thought.
I just realized ezPM100= O100 +D100 and +REB100.
There are advantages and disadvantages to a 3 way split. Advantage- skill split. Disadvantage- harder to compare apples to apples with some other metric splits. But it is a manageable challenge.
Yeah, that’s how I’ve been doing it lately. It seems the logical way to break it down to me. I may go even more atomic and break out all the team components (rebounds and tov).
Pingback: Rating the Centers | The City