Comments on: 2009-10 Win Share Rates, Updated

By: Jay

Jay — Sat, 31 Oct 2009 01:48:39 +0000

One metric that looks like it might capture player contribution to a team is comparing the following: (Player Offensive Rating - Player Defensive Rating) - (Team Offensive Rating - Team Defensive Rating). It's possible that this would be better expressed as a ratio. I haven't tried to look at this rigorously, but it looks at what's happening while a player is on the court, as well as controlling for a team's offensive and defensive strength. It appears to reward players who do the little things (e.g. Shane Battier) while punishing players who may get high PERs merely because they're ball hogs (e.g Allen Iverson).

By: Kevin

Kevin — Mon, 26 Oct 2009 00:41:36 +0000

Tsunami,

Kobe's defense is (and has been for some time now) quite overrated.

By: Tsunami

Tsunami — Sun, 25 Oct 2009 06:48:55 +0000

Something's not right here. When I sort by Projected Defensive Win Shares per Minute I have to scroll for a coons age before I see Kobe's name. You know, the guy whose defense GMs fawned over in this years preseason poll. What gives?

By: Anon

Anon — Fri, 23 Oct 2009 21:47:56 +0000

Don't mean to pick on your Celtics here Neil, but I think a great example of a player who might have frustrated people with his gap between his "talent" and "actual production" is this dude:

http://www.youtube.com/watch?v=F7gOPVFvif8

Get ya "shimmy" on.

By: Neil Paine

Neil Paine — Fri, 23 Oct 2009 17:24:46 +0000

That gulf between value and ability is touched on here. Basically, we'll never measure talent -- we can only go based on what players actually did. But then again, even the scouts have to do that. Only if we had players play a million games at a million positions in a million roles (and keep them from aging in between those games) would we ever get the definitive answer on ability. But we can say "Player X did this, and it was worth Y marginal points, resulting in X Win Shares". It's backwards-looking, to be sure, but luckily some of that value is retained in the future, allowing for projections like these.

By: Jason J

Jason J — Fri, 23 Oct 2009 16:52:18 +0000

i have one other minor issue with WS that i think is inherent in any boxscore-based metric. because we need to break the game down into possessions to account for pace, players on good team or in systems that require decoy action or set up plays, tend to look a little worse than they probably should. Players on some of the stacked teams of the 80s come to mind like Worthy, Bird, and about half of the Pistons come to mind. Larry Bird for instance might have, ironically, put up better numbers WS on a worse team. WS loves free throws and despises TOs (which no one can disagree with), but because he played on such a stacked front line, Bird's, a great post scorer, got limited post touches. He was more of a playmaker / shooter in the offense. Fewer opportunities for free throws and more opportunities for TOs results. It's not a big deal (particularly for Bird whose numbers were still impeccable), but it's one of those instances where we're measuring what a player did and not what a player could do, and then there's the tendency to, even unconsciously, use the result as an as absolute measurement for both.

By: Neil Paine

Neil Paine — Fri, 23 Oct 2009 07:06:32 +0000

Me neither, but thanks for sticking up for us, guys... Anyway, as far as Win Shares go, I think their biggest strength is also their biggest limitation, and that's the fact that they're tied strictly to team W-L and team efficiency (especially on defense). They're a great value metric because they do allocate a team's wins to its individual players based on Dean Oliver's individual offensive and defensive efficiency ratings. So if a player played on a bad defensive team and there isn't compelling evidence that the D was bad in spite of his efforts (as opposed to because of them), his offensive production is not going to be considered as valuable as equivalent production on a better defensive team. I think this is a major strength over something like PER, which almost considers player performance to have occurred in a vacuum (I remember Ike Diogu's PER was off the charts a few years back because no defense is really factored into the equation, other than BLK/STL/DRB). It's a double-edged sword, though, because we can't really measure individual defensive ability with a great deal of accuracy using just team defensive performance and box score stats. So you end up with wild year-to-year fluctuations in DWS when somebody like Ray Allen leaves a terrible Sonics defense for an all-time great Celtics D -- did Ray really get remarkably better at defense between 2007 and 2008, at age 32? Seems unlikely, but his DWS/mp increased by 285%! When you take a multi-year weighted average like we did above, it reduces this effect somewhat, but you're still never going to completely get away from that contextual problem (which is the story of all basketball stats, really). That said, Justin and I have found that WS do "retrodict" past seasons with a pretty decent amount of accuracy, so our final W-L projections should be as good as anybody's. They're just not going to be able to compare to the predictive accuracy of a similar system in, say, baseball.

By: Caleb

Caleb — Fri, 23 Oct 2009 05:49:35 +0000

Yeah I'm not sure what Sublicon is talking about.

By: izzy

izzy — Fri, 23 Oct 2009 02:47:40 +0000

What's with the hate? I have two responses to this...a) it's the off-season (unlike baseball and football), so there isn't as much subject matter for them to post b) it's not like the last time they posted was a month ago...this blog is updated very regularly. On another subject: Neil--what would you argue are the strengths and limitations of Win Shares as a metric?

By: sublicon

sublicon — Thu, 22 Oct 2009 18:34:51 +0000

"Following up on the work we did here about a month ago..." Is that the last time you guys actually did anything on this site? Baseball and Football Reference are updated often. You guys, on the other hand, aren't. Please do more make this a site worth visiting. I love all the reference sites and wish I could love this one as much as I love the others.