Win Shares Tweak
Posted by Justin Kubatko on October 16, 2009
Thanks to the alert eyes of one of our readers ("Deepak"), I have corrected a minor flaw in the Win Shares system. For those who don't care, feel free to move on. All others, please continue reading after the jump.
To convert marginal offense/defense into wins I was using league marginal points per win, which is equal to 0.32 times league points per game. However, thanks to Deepak's e-mail, I noticed that teams that played at a below average pace tended to have positive errors (i.e., more wins than Win Shares) and teams that played at an above average pace tended to have negative errors (i.e., fewer wins than Win Shares). The relationship between a team's pace factor and their prediction error was not particularly strong, but it was strong enough that I realized it should not be ignored. The problem was using the unadjusted league average to compute marginal points per win. Teams that play at a faster pace need more marginal points to generate one win, while teams that play at a slower pace need fewer marginal points to generate one win. Marginal points per win are now calculated as follows:
To illustrate this let's look at two teams at the extremes last year. For the the Golden State Warriors, who played at the league's fastest pace, marginal points per win is equal to:
On the other hand, for the Portland Trail Blazers, who played at the league's slowest pace, marginal points per win is equal to:
That's a big difference, and something that should not be ignored. By using the league average rather than the pace-adjusted league average, I was overrating the Warriors players and underrating the Blazers players. In terms of the overall prediction error this did not have a large effect (the overall error did drop, but not by much), but at the player level there are going to be some significant changes. For example, using the old system Andris Biedrins led the Warriors with 5.8 WS, but using the new system he leads them with 5.4 WS. Meanwhile, Brandon Roy, the Blazers leader, goes from 12.8 WS to 13.5 WS.
October 16th, 2009 at 6:52 pm
Puts last seasons campaign by LeBron James in 12th place all time for a single season. wedged between one Michael Jordan.
October 16th, 2009 at 7:38 pm
Not only that, LeBron also takes the lead for playoff WS in 09 despite not playing in the Finals. That's pretty rare company -- the only other player that comes to mind that led the league in playoff WS in a season without getting to the Finals is MJ.
October 18th, 2009 at 6:25 am
Excellent find... it makes, for example, Isiah Thomas, a better player in the late 80's. Magic Johnson receives slightly less WS playing playing with Abdul-Jabbar, more after the center's retirement.
Unfortunately, my paper notebook full of stats now has to make a huge revision...
October 18th, 2009 at 5:14 pm
I disagree, LeBron should not have the lead in Win Shares for the post-season. In fact, his PER is also inflated. (Actually Kobe's quality of opponent was also higher but that is irrelevant right now).
He takes quite a bit of shots 16-24 seconds into the possession (47% of total shots in the regular season and 50% in the post-season). It is very clear to me that you are not measuring time usage correctly.
Posessions is simply not good enough. LeBron is one of the few All-Stars in this league that takes 45+% (usually stars hover around 30-35%) of his shots between 16-24 seconds left. 82games.com is the source.
While LeBron may have a good case for being the best individual player in the league, his style of dribbling the heck out of that ball has hurt against elite squads.
October 18th, 2009 at 5:19 pm
A Celtics Blogger worded this well for me:
"Lebron probably passed Kobe as an individual force last year, but I still think Kobe's skill set, which includes mental attributes, is more appropriate for a championship team. For all his brilliance as a one-on-one player, Lebron's style of going one-on-five against the best teams is always going to have a hard time beating a squad that can match Cleveland for talent and runs a balanced team offensive, like the Lakers, Celtics, Magic, and possibly the Spurs."
Getting 43 PER against the 09 Pistons and Hawks, especially with his dribbling per possession ignored in the equation, should not be rewarded to this degree. Great player, his style would not seamlessly combine with another group of talented players.
October 18th, 2009 at 5:32 pm
Sorry forgot to post the source I got that from.
http://forums.celticsblog.com/index.php?topic=32294.0
Didn't mean to triple post, my bad. :]
October 18th, 2009 at 9:41 pm
@ Hk
Not sure what shot-clock management has to do with anything, but what I DO know is that players become less efficient when they have to take on more offensive possessions (which is what the WS model is based upon). LeBron has had to carry on a larger burden of the offense throughout the reg season and the playoffs and was STILL more efficient than Kobe. The problem with the Cavs not getting to the Finals in '09? Subpar play from the rest of LeBron's teammates -- just check out Neil's SPM scores from the '09 ECF, LeBron is the only player on the Cavs (outside of Terance Kinsey, who played four minutes in the series) who even had a POSITIVE +/- score!And LeBron's running mates were also pretty subpar using WS as well. I don't know about you, but when the guys around you aren't making shots, I'd be handling the ball on offense more often too :)
Kobe is a brilliant player and no one is taking away from his play in '09. But he also had a more consistent supporting cast throughout the reg season and the playoffs. And you win titles as a team.
October 19th, 2009 at 12:06 am
Let me ask, why does taking so many shots late in the shot clock automatically imply overdribbling? Couldn't it just be the natural result of playing on a team without other shot creators, and being the last resort as the clock runs down? Aside from Mo Williams, who else did Cleveland have that could remotely create a shot of their own off the dribble?
Also, this idea of James going one-on-five and having less of a team-oriented mentality than Bryant simply doesn't hold up -- by any assist metric (Hollinger's, Pomeroy's, AST/36, whatever), LeBron has always been a better passer and more willing setup man than Kobe. And as far as James' style "not seamlessly combining with another group of talented players," look at these numbers from the Beijing Olympics:
So much for the idea that James would play worse when surrounded by talented teammates...
October 19th, 2009 at 8:25 am
Haven't we had this discussion before? Sheesh...
October 19th, 2009 at 12:06 pm
You know, it does sound familiar... :)
October 19th, 2009 at 1:44 pm
I think the other error we make when attributing playstyle and clock management to players is the impact of coaches. It's a lot easier for Kobe to get quick hit scores not only because Odom and Gasol demand attention, but also because Phil Jackson runs a precise offense of cuts, post ups, and passes that allows for catch and shoot and quick one on one opportunities.
If LeBron operated in the triangle, I imagine he would be ridiculously good at it and less likely to pound the ball all day. Kobe would be likely to get more stats (and I'm not 100% sold on the notion that increased usage leads to lower efficiency as regards great players. makes sense for the more skill-limited unwashed masses sure), but even at his most aggressive with his weakest team, he just wasn't on the same level of production as James.
October 19th, 2009 at 2:43 pm
I love how every blog post on this site turns into a Kobe-LeBron debate. lol
October 20th, 2009 at 2:51 am
Actually Neil, during the Olympics Kobe had two terrible games at the start. But he rebounded quite well after that and was (maybe he and Wade) the best player by the end of the tournament. Certainly in those playoff rounds.
That isn't really a fair comparison of their production, as far as the Olympics is concerned. And he plays one manner for the Cavs, and in another for the USA Redeem Team. So I find that to be irrelevant. Michael Jordan didn't put up the best numbers on the Olympic Squad, so I never wanted to bring that up.
I was speaking about 2009 Playoff production though to be precise. By any measure (Hollinger, the guy across the street, or anyone), the Hawks and Pistons are not exactly the best teams in the league. Getting 43 PER in that manner is not extremely impressive. Writers like Dwyer wrote articles specifically praising LeBron, while Kobe had 30 PER in the other Conference Final.
And the Rockets are a "little bit" ;), better defensively, than what LeBron faced in the first two rounds.
October 20th, 2009 at 3:00 am
"Not sure what shot-clock management has to do with anything, but what I DO know is that players become less efficient when they have to take on more offensive possessions (which is what the WS model is based upon). LeBron has had to carry on a larger burden of the offense throughout the reg season and the playoffs and was STILL more efficient than Kobe. The problem with the Cavs not getting to the Finals in '09? Subpar play from the rest of LeBron's teammates -- just check out Neil's SPM scores from the '09 ECF, LeBron is the only player on the Cavs (outside of Terance Kinsey, who played four minutes in the series) who even had a POSITIVE +/- score!And LeBron's running mates were also pretty subpar using WS as well. I don't know about you, but when the guys around you aren't making shots, I'd be handling the ball on offense more often too :)"
Players on the Warriors and Suns have inflated numbers because (while they may have a nice True Shooting percentage or whatever), they needed more time to score those points. I never said LeBron wasn't a great player, but his PER is indeed inflated because it is not completely penalizing him for time.
Completely ignoring any stat, and just looking at him play I don't know why anyone is surprised. Yes his teammates are spot up shooters and he needs to dribble the ball and waste more shot clock than anyone to run that kind of offense. It works to a degree (we'll see about the Post-season), he's a great player, but 82games adds another time component simply not accounted for on this great website. Regardless of my disagreements with Blog Writer X, I still like having such discussions and reading anyone's thoughts on these subjects. I am not saying he's some crazy hog (he's been given the green light to do this within his offense), but other nuances are not being taken into account.
October 20th, 2009 at 3:07 am
"I think the other error we make when attributing playstyle and clock management to players is the impact of coaches. It's a lot easier for Kobe to get quick hit scores not only because Odom and Gasol demand attention, but also because Phil Jackson runs a precise offense of cuts, post ups, and passes that allows for catch and shoot and quick one on one opportunities."
Kobe has played even better without Gasol and Odom, I don't know what that really has to do with anything.
"If LeBron operated in the triangle, I imagine he would be ridiculously good at it and less likely to pound the ball all day. Kobe would be likely to get more stats (and I'm not 100% sold on the notion that increased usage leads to lower efficiency as regards great players. makes sense for the more skill-limited unwashed masses sure), but even at his most aggressive with his weakest team, he just wasn't on the same level of production as James."
At his most aggressive, he doesn't need nearly as much time as LeBron to score. That's exactly the point. I wouldn't be poo-pooing 28 PER for 41 minutes a night. LeBron's teammates are also better defensively.
LeBron does not have a mid range jumpshot or post moves, he would be more miserable in the triangle. Nor would he handle the ball nearly as much since the Triangle requires constant ball movement.
October 20th, 2009 at 12:40 pm
Wow it's annoying to write a post that is hugely complementary of Kobe, even going so far as to contest the notion that increased production diminishes his efficiency - hence suggesting that were he put in a position of carrying a team we could expect better stats all-around - only to have a Kobe-dite come back and try to fight every word with rehashed arguments that Neil has already dissected like a dozen times. So basically, despite LeBron's superior total production, per-minute production, per possession production, FG%, TS% and eFG%, his lack of a reliable midrange j makes him a lesser player? Does that mean Patrick Ewing was better than Shaq? Cause it sounds like that's what we're saying here.
Oh, and I've been watching Phil Jackson coach the triangle offense since 1990. I know whereof I speak. LeBron might not play the same role as MJ or Kobe exactly, but he would be monstrously effective in Tex Winter's system. Scottie Pippen had a suspect jumper too when he didn't use glass, and his post game, while adequate, wasn't anything spectacular either (it got better in their second three peat, but he hardly got to use it because MJ was option 1, 2, and 3). But he was amazing inside the triangle because of his point forward skills. LeBron brings those same assets to bear - except he's better. Like Pip, he's a point forward, but he's a much better playmaker and a much better passer overall. Like Luke Walton, Bron's court vision would be much more of a blessing in an offense designed to create open shots through ball and player movement than it is in a heavy-paced, drive and kick system. Bron is also an excellent post-feeder, a great skill to have in the triple post offense. And LeBron excelled without the ball in the Olympics playing with Paul and Kidd. Moving without the ball is not a concern.
Why am I arguing a hypothetical Bron / PJax pairing, you may ask? Because I'm on my lunch break, and nobody here wants to talk ball!
October 20th, 2009 at 7:12 pm
@ Hk #14
It doesn't matter who you are, if your teammates either aren't producing or creating shots you're gonna be handling the ball on offense more often than those who play on loaded teams. If you followed Kobe during the '06 season (and he's just one of about a gazillion examples out there that demonstrate this fact) you would know exactly what I mean.
The beauty of WS is not only does it take one's role in the offense into account (i.e., superstars vs. all-stars vs. role players), it is also pace-adjusted. So there's no bias against teams that play at a slower tempo than others -- which, by the way, LeBron's Cavs has played at pretty much his entire career.
Also, the Cavs has played against the Hawks in the playoffs, who were virtually identical to the Jazz defensively, and the Magic who was the best defensive team in the league. LeBron has played like an MVP against both squads.
P.S. I love "Is LeBron the best player in the league?" arguments too. Never gets old lol
October 21st, 2009 at 12:16 am
Anon and Jason, read the original post in this thread. I don't think you noticed the tweak made or forgot about it along the way.
The Hawks with an injured Joe Johnson and Al Horford? Yeah they got even more awful. Kobe had 30 PER in the Conference Finals and locked up Melo after Ariza couldn't guard him, also had 27+ PER against the Rockets who have better perimeter defense than the Magic. Then 27+ PER against the Magic in the Finals, while deferring the ball much more than Bron-Bron had the series before.
Justin Kubatko tweaked Win Shares to account for time. Simple as that, we're not discussing how sweet his stats look, but an objective analysis of the truth which is he handles the ball (good or bad), more than any other Superstar in the league.
Let me quote Justin:
"I was overrating the Warriors players and underrating the Blazers players. In terms of the overall prediction error this did not have a large effect (the overall error did drop, but not by much),"
There is no reason to complain about these issues to me, my comments are completely on-topic. This isn't a love/hate situation, Win Shares has overrated pace-inflated players before according to Justin himself.
October 21st, 2009 at 12:48 am
I don't think you understand how Win Shares are calculated, and/or what this change means on a mathematical level. He didn't tweak the formula "to account for time", he tweaked it because he realized the number of marginal points required for a high-pace team to win a ballgame is greater than the number of marginal points required for a slow team to win. The conversion between marginal points and wins used to assume a constant mpts/win rate across all teams, which obviously adds up at the league level (pace across all teams evens out to the league average) but was ignoring the fact that every point scored in a high-pace game is less valuable than one in a slow-pace game, because they are less scarce -- they're easier to come by. 2 points in a 85-80 grind-it-out game are more important than 2 points in one of those 80s-era 165-160 shootouts. So if anything, the old formula used to understate the value of taking shots late in the shot clock by undervaluing guys on slower-paced teams.
October 21st, 2009 at 11:44 am
"I disagree, LeBron should not have the lead in Win Shares for the post-season. ...
He takes quite a bit of shots 16-24 seconds into the possession (47% of total shots in the regular season and 50% in the post-season). It is very clear to me that you are not measuring time usage correctly.
Posessions is simply not good enough. LeBron is one of the few All-Stars in this league that takes 45+% (usually stars hover around 30-35%) of his shots between 16-24 seconds left. 82games.com is the source."
Are you suggesting that a player contributes less to wins the more often takes shots late in the shot clock? How exactly should Win Shares take into account shot clock usage?
October 21st, 2009 at 12:04 pm
It's the same thinking that causes people to believe that walk-it-up teams aren't as good offensively as fast-break squads because they score fewer PPG...
October 23rd, 2009 at 8:20 pm
"I don't think you understand how Win Shares are calculated, and/or what this change means on a mathematical level. He didn't tweak the formula "to account for time", he tweaked it because he realized the number of marginal points required for a high-pace team to win a ballgame is greater than the number of marginal points required for a slow team to win.The conversion between marginal points and wins used to assume a constant mpts/win rate across all teams, which obviously adds up at the league level (pace across all teams evens out to the league average) but was ignoring the fact that every point scored in a high-pace game is less valuable than one in a slow-pace game, because they are less scarce -- they're easier to come by. 2 points in a 85-80 grind-it-out game are more important than 2 points in one of those 80s-era 165-160 shootouts. So if anything, the old formula used to understate the value of taking shots late in the shot clock by undervaluing guys on slower-paced teams."
Dude, I know who "Deepak" is (I believe he is also known as "durvasa" IIRC), I am quite aware of what was tweaked.
Your site is interesting but different from 82games. I simply pointed out a nuance not accounted for. Pace is part of "time", you did indeed tweak for "time".
October 26th, 2009 at 7:32 pm
"Are you suggesting that a player contributes less to wins the more often takes shots late in the shot clock?"
It doesn't properly adjust for "time" needed to score. Minutes and pace are not enough.
"How exactly should Win Shares take into account shot clock usage?"
Maybe it can't yet Deepak (durvasa? ;])? Oh well.