“All models are wrong, but some are useful,” George E. P. Box.
Metrics and statistical evaluation tools are never exact. Dean Oliver never claimed that using Offensive and Defensive rating would allow a MVP voter to instantly identify the precise best player in the league, much the opposite in fact. Metrics are meant to serve as one indicator of efficiency and production that can be applied against the complexities of a game with 10 men on the court at any given time employing various strategies against a shifting landscape of circumstances.
That said, metrics are useful as quick and rough single number evaluators of who the pool of players operating at the highest level may be, and journalists and fans are learning to fall back on these numbers more readily than in the past. So given that we know the metrics we rely on are not perfect, let’s take a quick look at who is left in these playoffs as an indicator of just how useful the derived numbers are.
| 1. | LeBron James-MIA | 31.6 |
| 2. | Kevin Durant-OKC | 28.3 |
| 3. | Chris Paul-LAC | 26.4 |
| 4. | Carmelo Anthony-NYK | 24.8 |
| 5. | Brook Lopez-BRK | 24.7 |
| 6. | Tim Duncan-SAS | 24.4 |
| 7. | Dwyane Wade-MIA | 24.0 |
| 8. | Russell Westbrook-OKC | 23.9 |
| 9. | Tony Parker-SAS | 23.0 |
| 10. | Kobe Bryant-LAL | 23.0 |
| 11. | James Harden-HOU | 22.9 |
| 12. | Blake Griffin-LAC | 22.4 |
| 13. | Andray Blatche-BRK | 21.9 |
| 14. | Anthony Davis-NOH | 21.7 |
| 15. | Kyrie Irving-CLE | 21.4 |
| 16. | Stephen Curry-GSW | 21.3 |
| 17. | Al Jefferson-UTA | 20.9 |
| 18. | John Wall-WAS | 20.8 |
| 19. | LaMarcus Aldridge-POR | 20.4 |
| 20. | Deron Williams-BRK | 20. |
The above table – courtesy of Basketball-Reference.com – shows the top 20 players by PER from the 2013 regular season. 7 of the top 10 (including Westbrook since he helped to win the first round series) are still in the playoffs. Of the other three in the top 10 no longer in the playoffs, Chris Paul and Brook Lopez were in their respective first round series through game 7, and Kobe Bryant missed the entire first round due to injury – though given that Tim Duncan and Tony Parker are represented in the top 10, clearly either Kobe or Tim and Tony would not be playing in the second round if Mamba had played.
Players 11-13 also played until game 7 of their series. 14 and 15 were talented young players on weak teams that missed the playoffs. 16 is Stephen Curry, the presumptive MVP of the playoffs so far, and Deron Williams rounds out the top 20 as yet another game 7 participant. Basically what we see here is that 15 of the top 20 players by PER advanced to the second round or pulled their round 1 series to 7 games, which I would suggest indicates that their teams were good enough to advance with a break or two in their favor. If you excuse the Lakers’ crummy showing due to Kobe’s injury, it’s 16 of 20. So either 75% or 80% of the top 20 players led teams good enough to win a playoff series.
Winning isn’t exactly what PER is trying to predict, but if we adhere to the old standard that winning in the NBA requires star players, we do see the players that PER considers to be stars well represented.
| 1. | LeBron James-MIA | .322 |
| 2. | Kevin Durant-OKC | .291 |
| 3. | Chris Paul-LAC | .287 |
| 4. | Tyson Chandler-NYK | .207 |
| 5. | Tony Parker-SAS | .206 |
| 6. | James Harden-HOU | .206 |
| 7. | Marc Gasol-MEM | .197 |
| 8. | Tiago Splitter-SAS | .197 |
| 9. | Blake Griffin-LAC | .196 |
| 10. | Russell Westbrook-OKC | .195 |
| 11. | Dwyane Wade-MIA | .192 |
| 12. | Tim Duncan-SAS | .191 |
| 13. | Brook Lopez-BRK | .191 |
| 14. | Carmelo Anthony-NYK | .184 |
| 15. | Deron Williams-BRK | .184 |
| 16. | Serge Ibaka-OKC | .181 |
| 17. | Stephen Curry-GSW | .180 |
| 18. | David West-IND | .179 |
| 19. | George Hill-IND | .177 |
| 20. | Chris Bosh-MIA | .175 |
Unlike PER, Win Share really does aim to predict winning, and it actually does an even better job by the standards we established in the above study. Every single player in the top 20 for Win Share per 48 minutes advanced to the second round of the playoffs or made it to game 7 (again including Russell Westbrook) of the first round.
I’m both impressed and surprised by these results. I’m going to have to do a historical retrospective on this investigation and see if these two metrics pertaining to individual players have always been such accurate predictors of team playoff success.
For our purposes PER and WS/48 really do prove to be useful. Go nerds!
Tags: 2013 NBA Playoffs, Al Jefferson, Anthony Davis, Blake Griffin, blatche, Brook Lopez, Carmelo Anthony, Chris Bosh, Chris Paul, David West, dean oliver, Deron Williams, Dwyane Wade, george e p box, George Hill, James Harden, John Wall, Kevin Durant, kobe bryant, Kyrie Irving, lamarcus aldridge, lebron james, NBA, Russell Westbrook, serge ibaka, sports, Stephen Curry, tim duncan, Tony Parker, tyson chandler
Double Dribble 
May 10, 2013 at 4:17 pm |
I still prefer Win Shares because they are by game. Sometimes the best decisions in the last 4 minutes needed to win a game are not the most efficient players over 48 minutes. For example, lets say you can’t choose LeBron or Durant for the last 3 shots of the game. Wouldn’t you choose Carmelo or Kobe over Harden and Parker.
May 10, 2013 at 4:30 pm |
That’s a very good point! I actually used the per minute solely because it was pre-prepared in a tidy list @ BBR. Frankly these types of box score stats don’t make great situational evaluators for something like a late game possession user.
May 13, 2013 at 3:23 pm
It’s just a stat, far from any absolutes. Sure, probably the better players are towards the top, but you never know. Kobe is at least a top 3 player, and he’s nowhere to be found in your win shares list.
Also, Paul never made it to game 7 in the first round. And Hibbert and George are the pacers top 2 players, but West and Hill are ranked ahead of them. Splitter at #8?
May 13, 2013 at 3:51 pm
True re: CP3 only going 6 games. We’ll let them slide since they were missing Blake at the end, though it probably wouldn’t have made a difference.
WS has a tendency not to value Kobe’s production very highly. I think it underrates his defense historically and generally isn’t a big stat for high usage guards. He’s on the PER board.
For the Pacers guys again WS doesn’t always do a great job of assigning defensive credit unless there’s big defensive stats. So it probably broke the defense basically equally among the starters (and that’s where Paul George and Hibbert really shine) and then the offensive efficiency for West and Hill might be better. I could look it up.
May 15, 2013 at 2:03 pm
Probably true about WS. I don’t really know or really care that much about the intrincasies of these advanced stats. They are interesting to look at, and overall might be a good generalization of where players rank, but they’re obviously far from any truths. I appreciate your articles, and you seem to realize this about advanced stats. And I think PER and WS are probably 2 of the better advanced stats that rank players. Really, the most important advanced stats we should be analyzing isn’t rating player vs. player, but 5-man unit vs. 5-man unit.
Obviously, most of the time, the top players will play further into the playoffs. Somebody like Chandler is far from being a great player. His offensive game is extremely limited, but yet he’s rated extremely high.
May 15, 2013 at 2:31 pm
Yeah, Chandler is a great example of where these per possession or per minute metrics fail to fully capture the importance of context. Sure he’s very good at finishing lobs on the move, which is valuable, and he only takes those shots he almost can’t miss, which is smart, but the stats can’t acknowledge how limited his skills are in terms of generating offense.
Horace Grant was the same way. A couple seasons he outranked Pippen in Win Share because Pippen executed plays to furnish Grant with easy scoring opportunities while Grant simply finished those plays. Marion ranks too high for some of his Phoenix years as well.
June 11, 2013 at 12:41 pm |
Useful to what extent? I’ve critiqued why PER, in particular, is fundamentally flawed in response to a couple of your other blog entries, so I’ll save repeating my comments here. But PER suggests that Paul and Anthony are the third and fourth-most valuable players in the NBA, yet neither was playing in the conference finals. In fact, in eight NBA seasons, Paul has never played in the conference finals (despite PER’s absurd suggestion that he is a better point guard than Magic Johnson), while in ten NBA seasons, Anthony has reached that stage just once. Obviously, those failings cannot be completely attributed to the individual players, but where exactly is the precision here? You seem to be using the word “useful” very, very loosely. Likewise, Win Shares per 48 minutes (also flawed and unreliable, unless you actually believe that Amar’e Stoudemire was almost always more responsible for the Suns winning than Steve Nash) suggests that three of the four most valuable players in the NBA were Kevin Durant, Chris Paul, and Tyson Chandler, none of whom reached the conference finals this year. Granted, Durant obviously found himself unexpectedly stripped of his co-star, the injured Russell Wesbtrook, but if he constituted the second-most valuable player in the NBA, then presumably he might have led Oklahoma City to more than one win in the second round. (And actually, Durant may well have been the NBA’s second-most valuable player, but the point is that PER doesn’t offer much in the way of predictive value or absolute truth.)
And as for PER, yes, of course Brook Lopez is not a better player than Kobe Bryant. I mean, Bryant is overrated historically, but he’s a heck of a lot better than Brook Lopez, a center who hasn’t averaged 7.0 rebounds per game in any of the last three seasons and one whose assists-to-turnover ratio has been underwater throughout his career. PER possesses two big biases (de facto biases, at least): one in favor of players, like Paul, who play on teams with slow pace factors, and the second in favor of front-court players, probably because PER overvalues rebounds relative to assists. In truth, a slow pace can be as beneficial and possibly more beneficial to a star’s numbers than a fast pace (because a slow pace better enables a star to control the ball and milk the possessions), and assists are more valuable than rebounds because they are more scarce and they lead directly to points.
But back to the question of usefulness, according to Win Shares per 48 minutes, the Clippers, Knicks, Nets, and Thunder (even without Westbrook) all featured two of the top-sixteen players in the NBA, yet none of those clubs reached the conference finals, while the Clippers and Nets failed to even make their way out of the first round. If anything, these metrics may contradict your argument and their supposed value. Indeed, I sense that you’re seeing in them what you want to see and looking for ways to uphold their value, whereas close scrutiny reveals a ton of problems. Of course teams generally need stars to win, because stars tend to be the best players and teams generally need better players to win. If that’s all that PER and Win Shares tell us, then they’re not telling us much at all. PER and Win Shares are supposed to provide greater precision, but they’re not doing so, or else the Clippers, Knicks, Nets, and Thunder would not have all failed to reach the conference finals. And, yes, I realize that Miami and San Antonio also possess multiple players in the top-sixteen, so someone had to be left out, but under a microscope, there is no clear or consistent correlation between these rankings and playoff success.
Moreover, Tiago Splitter, while probably an underrated role player, is obviously an example of a player whose value becomes inflated based on his system and star teammates. You couldn’t say, “Splitter’s value is apparently equivalent to that of Marc Gasol’s, so let’s build a team around him,” yet such is the suggestion offered by Win Shares per 48 minutes, another mark of the metric’s unreliability.
The bottom line is that in today’s technologically crazed age, people are trying to sum everything up with a singular value, and life (or basketball) just isn’t that simple. You’re never going to understand anything based upon one number derived from a contrived formula. Rather, you have to account for a variety of data pieces without false reductions, and that’s where the human mind is supposed to come into play, blending science and art. Instead, people are basically seeking to surrender the mind to pseudo-science that creates false reductions through metrics. When you affirmatively write that “journalists and fans are learning to fall back on these numbers more readily than in the past,” you’re actually (if inadvertently) speaking to the problem. Journalists and fans are lazily suspending critical thinking in the search for fashionable convenience, and harming intellectual discourse in the process. Ultimately, metrics tend to be more progressive-regressive, rather than progressive.
October 15, 2013 at 1:56 am |
That was an intolerably long post just to say that critical thinking and the eye and common sense test should always be the determining factor in rating any players.
PER and WS ranks Dirk over Bird in their careers. LOL. That’s all I need to know about those not so “advanced” stats. I bet 98% of the people who trot that crap out don’t even know what the exact formulas are, or have even played 2 secs of organized basketball before.
Just a buncha unathletic fat slobby bald-headed nerds infiltrating the sacred game we love, not to improve them, but just to egotistically fawn over endless pages of text and numbers just to live vicariously through them b/c in real life they can’t physically involved themselves in sports.
In baseball, it’s been FAIRLY helpful, b/c it’s not really a team game. It’s a game played by a bunch of individuals that doesn’t really rely on teamwork exactly.
Basketball, however, DOES. So synergy and chemistry matters BIGTIME.
Another reason why WS/48 is BULLSHIT? Tyson Chandler is ALWAYS in the top 5. I’m a Knicks fan, but that’s hilarious.
Ranking players is stupid anyways. Everyone knows Lebrons’ the best right now, and that Durant is basically #2. I don’t need fancy freakin stats to tell me that. Nor do I need them to tell me that Chris Paul is a top 6 player, or that Parker was a top 6 player last year, or that Melo was, etc.