Monday, March 28, 2011

The Regend and the Ninja: A Response (part I)

New readers, please check out the basics before leaving any comments.


Yesterday, Tom Liston of Raptors Republic - TrueHoop's Raptors' affiliate - wrote a column about Reggie Evans and James Johnson. In his column, Tom mentions that he has some problems with Wins Produced and my analysis of Evans and Johnson in the past. Specifically, Tom's main issues seem to be:
  • Evans' WP48 is near the top of the league (currently at 0.356), but Reggie doesn't seem to belong with the other players (Kevin Love, Dwight Howard, Chris Paul, LeBron James, Dwayne Wade) at the top.
  • If Wins Produced doesn't overvalue rebounds, and the only thing Evans' does well is rebound, why is Evans' WP48 so high?
  • James Johnson is playing well in Toronto - "he’s pretty much exactly the same player" as he was  during his rookie season. So why did I say that he probably wouldn't be a good player when Toronto traded for him?
  • Wins Produced is one measure of player productivity among several, and one should "never rely on one metric to draw conclusions".

Let's start with Reggie.


Why is Reggie Evans so productive?

Some people seem to have this problem whenever I mention, use, or reference Win Produced; their eyes glaze over, they start frothing at the mouth, and they begin to speak some kind of incomprehensible language. Sure, that also happens when I'm not talking about Wins Produced, but still. Anyways, here's an experiment: of the seasons in the following table, which one looks the best to you? I will tell you that they are seasons from Reggie Evans' career, but not which ones specifically.


As you can see, all four seasons, A through D, are compared against the numbers from an average power forward. Which looks best? Clearly season A has the worst shooting (although the FT% is relatively good, which isn't saying much), but it also has the best rebounding, fouls, assists, and turnovers. Season B has the best adjusted field goal percentage, but unfortunately the poor free throw shooting ruins the overall shooting, as evidenced by the true shooting percentage. Notice that season B is also worse than season A in almost every non-shooting way. Season C has the best shooting, which is best measured by true shooting percentage. The free throw attempts, field goal attempts, and points are about halfway in between seasons A and B, but season C is worse than season B in every non-shooting area. Season D looks very similar to season C, but is better in almost every area.

So which do you choose? Which do you chooooooose?
There's a new Pokémon game out, after all (available in fine stores everywhere)
Despite the terrible shooting, I'd pick season A. Why? Clearly the shooting and scoring are terrible, but - and this is crucial - look at the FGA. They're incredibly low, aren't they? So even though the shooting is terrible, there isn't much terrible happening. All of these seasons are below average with respect to shooting, but season A minimizes the terrible by taking fewer shots! It's perfect: if you can't shoot, don't! And that's what Reggie did that season. The rest of the numbers from season A are impressive: tons of rebounds, good steals, below-average number of turnovers, and it has the most assists and least fouls of any of the seasons listed. The only problem is with blocks, and it's not really all that much different from any of the other seasons (although it's quite a bit below average). So season A is an excellent season, and the one I'd choose over all the others.

You will not be surprised to find out that seasons A-D are, respectively, the 2010-11 season, the 2009-10 season, the 2008-09 season, and Evans' career stats. Evans is playing a lot better this season than he has in the recent past, although he is no stranger to seasons of WP48s of 0.250 and higher (0.283 in 06-07, for instance).

I think that's straightforward enough - Evans is having a good year - his best - because he's doing one thing really well and minimizing the impact of the things he does poorly. But why is his WP48 up near the likes of Love, Howard, Paul, LeBron, and Wade?

Let's get this out of the way first: Reggie Evans is rebounding at a rate that only one player over the last 34 seasons has been able to match (I'll ruin the surprise for you: it's Dennis Rodman). Evans is really, really good at rebounding this season (and please remember: he played with Bargnani last year, too, and I don't want to get into another rehash of the endless "WP overvalues rebounds" argument, because it's been stomped to death). He's doing pretty well with regards to steals and turnovers, but the rebounding is what carries him. Those other players - Love, Howard, Paul, LeBron, and Wade - aren't one-trick ponies; they do a lot of things well. In fact, Kevin Love is pretty much a Reggie Evans who can shoot. So that's the main difference between Evans and the rest - Evans can do one thing amazingly well, and the others are more all-round players.

That doesn't fully answer the question though. Why does a one-trick pony - albeit a historically good one-trick pony - approach the WP48 of the best players in the game? It has to do with how Wins Produced is calculated. Wins Produced is based on the statistical correlation between box score stats and team point differentials. Rebounds, it turns out, are very important for success in the NBA (go figure). So we shouldn't be surprised that a player who rebounds at a historically good rate and who doesn't have many negative statistical areas in his stats is thought to be playing exceptionally well.

But make no mistake: good production does not necessarily mean "good player". Personifying Wins Produced, WP looks at Reggie's stats and says "here is a productive player." It does not say why this player productive, nor does it say that this player is "better" than all the players with lower WP48s; it simply says "here is a more productive player." It does not tell us who would win a game of one-on-one, nor does it tell us who would win a game of horse, 21, bump, a three-point shooting contest, or a beauty pageant. It simply says that Reggie has been very productive this season, almost as productive as some of the best all-round players in the league. With that established, we can then perform more in-depth analysis, like we did above:
  • Why is he productive? (rebounding)
  • Would I want him on my team? (to get rebounds, yes)
  • Should Reggie take more shots to increase his scoring average? (no; while his scoring average will increase with more shots, he is so bad at shooting that he will increase the negative aspects of his game and thus hurt his own production, as well as his team's)
  • Down by two with three seconds left, should he be taking a turn-around, fade-away three pointer if he's double-teamed and there are other teammates open behind the three point line? (in the name of all that is good in the world, NO!)

So ends part one. In part two, I will discuss L'affaire Johnson and examine whether or not Wins Produced is just one metric among many that we should consider before we evaluate NBA players (hint: I'm still a writer for the Wages of Wins Network).

 - Devin

11 comments:

  1. Hey Devin, could you explain how Wins Produced measures the degree to which Evans not shooting hurts the Production of others on his team, who get more attention from the opposing defense as a result? The so-called "having to play 4 on 5 on offense" effect? Would love to know. If he does hurt the production of others on the floor with him, does that have any bearing on him being a productive player or not?

    ReplyDelete
  2. @Quick: Although not specific splits of with and without him on the floor, they seem to have little difference in eFG% when he plays in the game. They lose 1.5% in FG% but go up 1.5% in 3pt%, which is actually a higher eFG% than when he doesn't play. I think this shows that at a high level it doesn't matter much.

    To be more accurate though, we need stats when he is in the game and out of the game, and would likely have to compare the stats of the individual players and not the net team.

    ReplyDelete
  3. Devin,
    Brilliant observation! That's a funny one "What's he doing better?" "He's not shooting!" I think people forget things like these. You can improve by getting fewer turnovers, taking fewer [bad] shots, fouling less but those don't show up as readily as "He's scoring 5 more points a game. . . .on 8 more shots a game. . . . in 5 more minutes a game" :)

    ReplyDelete
  4. Those links you posted allegedly stomping the argument that wins produced overvalues rebounds do nothing of the kind.

    I am curious: if a higher WP48 number doesn't mean a player is better, but only more productive... what does the distinction between better and more productive amount to? I mean, if I were to take Wins Produced seriously, I should presumably want Reggie on my team over Pau Gasol, no, given he's a more productive player? If not, why not?

    ReplyDelete
  5. Actually, the comments by "Guy" on the Arturo Gallati post linked in the article http://arturogalletti.wordpress.com/2010/11/23/the-rebounding-myth/

    exactly cover why we shouldn't believe Reggie really contributes all that much to his team's wins.

    ReplyDelete
  6. Quirk:

    Nice leading question. Almost every team that has ever existed has at least one important player who you don't have to cover out on the perimeter. Remember Shaq, Rodman, and Ben Wallace? The "four-on-five" effect isn't exclusive to Evans, nor does it prevent a team from being successful. However, I think that pairing up two "non-shooters" to make it a "three-on-five" effect might lead to some problems.

    Eugene:

    There are many, many more links (and don't forget the hundreds of comments) on rebounding on the WoW; I don't have the patience to go pick them all out for you. There are some in the Required Readings on this site, too. If you are truly interested, go look around. If not, I'm not going to babysit you.

    If you look at the one number - WP48 - for this season, then yes, you'd choose Evans over Gasol. But that would be stupid. This is Reggie's best year, and he's getting older. Pau is getting older too, but he always plays more minutes, produces more total wins, and is generally better on a per-minute basis. Then there's the fact that, as I mentioned, you need to understand why a player is productive. Reggie does one thing well; Pau does multiple things well. Given roughly equal levels of production, I'd always side with the better all-round player. You also have to consider the context - if your team is absolutely terrible at rebounding, you might consider Evans over Gasol. But then there's the fact that Evans is injured all the time...but also the fact that Evans is much cheaper...needless to say, there's a lot to consider. Wins Produced is a starting point.

    Really, I'm not getting into the rebound thing again. I've already considered every objection to the way that Wins Produced treats rebounding, and I find those arguments to be lacking. At this point, all I can do is agree to disagree

    ReplyDelete
  7. It's certainly fair to prefer Reggie to Pau based on age, health issues, etc. But ask a different question: take Reggie from last week (before his foot started hurting) and Pau: we're playing a game where we each choose 8 man teams and whoever's team wins gets a cool million bucks. For one of your player spots, you have the option of having either Pau or Reggie as one of your forwards. Which one do you pick? Aren't you pretty much committed to going with Reggie at this point?

    Now, I certainly can't prove that my Pau-full team will be hugely better than the Reggie team. I can add my voice to the chorus explaining why I distrust Wins Produced in judging individual production, especially for people on the tail end of the rebounding distribution. But I don't want to harass you on the topic, if you're sick of it. Me, I find it interesting, partly because to me it illustrates some interesting confusions people have with the distinction between causation and correlation. I'll just say that my issues with WP48 don't stem from lack of familiarity with the topic.

    ReplyDelete
  8. As I was hinting at previously, I'd actually prefer Pau over Reggie. So no, I'm not committed to going with Reggie.

    I'd want to choose proven, better all-round players for a fantasy game. Dwight Howard would be my first choice, and LeBron James, Kevin Love, and Chris Paul would be up next. Because it's only one game, I might gamble with someone like Andrew Bynum (injury-prone), Steve Nash (old), or Kevin Garnett (oldish and injury prone) once all the clear cut best players are gone. I'd pick up Reggie as a gamble towards the very end of the picking, and I'd make sure that if I played him during the game, I'd pair him with players who complement his abilities.

    We know that rebounders are not the only players responsible for causing rebounds. Rebounds can be caused by good team defence, for instance. But if we split up credit for rebounds, we should also do the same for other stats. For instance, is a shooter the only one responsible for making a shot? What about the ball movement and screens that led to the open shot? So we have no other choice but to credit the players who actually grab the rebounds.

    You could start splitting every stat up into ever-smaller pieces, but as Berri notes:

    "In my view, you have to have a reason to do this. For example, if splitting up credit improves something about the model, then maybe this is justified. But as we see with Win Shared, the splitting process
    - makes the model more complicated
    - does not improve explanatory power
    - and reduces consistency over time

    So there are costs with splitting but no apparent benefits."

    ReplyDelete
  9. It is quite true that splitting credit for stats won't improve WP's explanatory power -- in terms of the correlation between team WP and team wins. But if our concern is how WP allocates credit to individual players, considering how to divide credit for box-score stats may address the concerns despite not being relevant to WP's explanatory power in that narrow sense.

    Reggie is such an interesting test case for box-score metrics precisely because his output is so skewed. Speaking for myself, if I was picking for a one-off game, there are quite a few PF's with substantially lower WP48 I would pick before I would take Reggie. That could, of course, suggest that I'm giving my subjective impression of 'contribution' too much weight. But it also could well be that the case of Reggie illustrates a skew in how WP48 assigns credit to individual players. Certainly if there was such a skew in its weighting, valuing a perrenial bench player with one elite skill as if he was a top 15 talent would be how we might expect it to come out.

    ReplyDelete
  10. @Evan, Isn't the eFG percentages you are comparing for different players? Seem like it would be difficult, if not impossible, to isolate Reggie's impact considering it's not the same players playing (on either team) when he is on, vs off the court.

    @DDignam, while your personal feelings, guesses and allusions to history may be interesting, but it's not a very rigorous argument you are making here, is it.

    Should I take it on faith that it's ok to put your teammates into a 4-on-5 situation because Michael Jordan managed? Not to mention that while Rodman took very few shots, his career avg eFG% was .529 compared to Reggie's.466 (this year .375).

    Also, I would like you explain why you would choose Pau over Reggie, when you believe that WP clearly shows Reggie to be the more productive player? Please explain why you would choose a player your analytical convictions tell you is less productive?

    ReplyDelete
  11. Quirk:

    Wow, my personal feelings, guesses and allusions? You've got to be kidding! I can go through every team:

    CHI - Noah
    BOS - Shaq/Perkins
    MIA - Dampier/Anthony
    ORL - Howard
    ATL - Pachulia
    PHI - Young
    NYK - Turiaf
    IND - Hibbert/Foster
    CHA - Kwame
    MIL - everyone/Bogut
    DET - Monroe/Wallace
    NYN - Humphries
    TOR - Reggie/Dorsey
    WAS - McGee
    CLE - Hickson/Hollins

    That's just the East. Every team has one or more significant players who can't do anything from outside. Generally it's a centre, but not always.

    I believe I've already explained why I'd choose Pau over Reggie. Really, you should go read it. Short answer: Same age, Pau plays more minutes, doesn't get injured, and is more well-rounded. If you take away Reggie's rebounding (somehow), he's got nothing. Take away Pau's and he's still able to contribute. Pau's also bigger and can play centre (properly).

    ReplyDelete