How useful are win shares?

[Precision9]

How accurate are win shares, how useful are they, and how do they work? I would like to know because I have never taken them seriously and I do not plan on it. That being said, of all the win shares to choose from defensive win shares seem to have the most credibility.

Top 5 in defensive win shares in the '10-'11 season
1. Dwight Howard
2. Kevin Garnett
3. Lebron James
4. Luol Deng
5. Andrew Bogut

That's a pretty decent list of great defenders, but if you just look at other than win shares, specifically offensive win shares there seems to be issues....

Win shares per 48 of the '10-'11 season
1. Lebron
2. Dwight
3. Pau
4. CP3
5. Chandler - ?????????

Plain old Win Shares
1. Lebron
2. Pau
3. Dwight
4. CP3
5. Rose

Offensive Win Shares
1. Lebron
2. Pau
3. CP3
4. Love
5. Durant

[Paydro70]

Well, because win shares are a means to apportion wins, I personally don't like to use them much. Two players can play exactly the same but end up with different win shares on the basis of their team's success, and to me that's a substantial problem when you're trying to compare players.

That said, they certainly aren't the worst metric... the numbers generally pass the laugh test (what about the OWS bothers you? Love?), and some people probably like the fact that they correlate to wins because they think better players will automatically generate more victories (which is, of course, the concept behind WS).

Of course, they *ARE* very good if you're looking to compare players on the same team.

[Chronz]

Well, because win shares are a means to apportion wins, I personally don't like to use them much. Two players can play exactly the same but end up with different win shares on the basis of their team's success, and to me that's a substantial problem when you're trying to compare players.

Thats actually not true for offensive winshares. The team component you speak of is prevalent in defensive winshares, which is why its important to distinguish between the 2 in every comparison. That doesnt mean Def.WS are worthless just that a players defensive impact can be over/understated. Off.WS does factor in Team Off.Rebounds and Assists in some way but based on what Ive read and the people Ive asked, it has a marginal influence on a players overall WS count. Which from my experience in comparing players seems right.

Of course, they *ARE* very good if you're looking to compare players on the same team.

When I consider the #'s, PER and WinShares are both polar opposites of the same argument. One stat favors efficiency (OWS) the other shot creation. On the same team or another, chances are if your exceeding your counterpart in both PER/WS and there is no other real evidence for him making anyone else more efficient, then your probably the better player. But intangibles are a tricky thing so there is no rule, just guidelines.

[Doctor MJ]

I actually think the team success component is a feature not a bug. If you want stats that try to measure players simply by box score success, there are plenty of them. It's good particularly on defense to see something try to use both box score & team success because the box score sucks on defense.

Does that make WS a full proof system? Definitely not, but it's better than nothing.

As far as Chandler appearing on the list, yeah, that's not good. I think the productive way to look at stats though is to try to understand why it screws up so that you can use it with an intelligent caution rather than just saying "Oh well this is useless then", because you could literally throw out any stat that way. In Chandler's case what's happening is that his efficient is at a far extreme which "breaks" the equation. An attempt at a perfect stat would respond to this by making the algorithm more complicated, and it wouldn't be hard to make manipulations that made Chandler's WS score look more reasonable.

There's a bit of dishonesty though in making such tweaks. The reality is the statmakers know not only that their stat isn't perfect, but that a perfect stat doesn't exist. Better to use a solid foundation to create a pretty informative stat whose formula can be readily understood than to muddy it all up for no reason other than to get bragging rights.

In general it's important to understand what types of players a stat is really built around, and when it comes to all-in-one stats like PER, WS, or even effectively APM, the focus is on judging high usage big minute players.

[parapooper]

I don't see the problem with Chandler being fifth in WS/48. The guy was top5 in trb% and had 70% TS.
70%TS is huge for winning. You could argue that someone scoring 30pts/game at a TS% at or below his team's TS% contributes less than Chandler scoring 10pts a game at insanely high TS%. Sure he profited from Dirk and Kidd getting him easy scoring opportunities, but he was extremely valuable to his team with what he was doing.
And he wasn't even in the top20 in WS due to his low minutes. So him being high in WS/48 is rather an argument for than against the usefulness of WS to me. What he does may not look too impressive but I think that WS/48 is correct in valuing it highly.

[Doctor MJ]

I don't see the problem with Chandler being fifth in WS/48. The guy was top5 in trb% and had 70% TS.
70%TS is huge for winning. You could argue that someone scoring 30pts/game at a TS% at or below his team's TS% contributes less than Chandler scoring 10pts a game at insanely high TS%. Sure he profited from Dirk and Kidd getting him easy scoring opportunities, but he was extremely valuable to his team with what he was doing.
And he wasn't even in the top20 in WS due to his low minutes. So him being high in WS/48 is rather an argument for than against the usefulness of WS to me. What he does may not look too impressive but I think that WS/48 is correct in valuing it highly.

Well here's the thing: When you see a player score sky high in a per minute metric without playing sky high minutes, it begs the question of why the player isn't playing more minutes. When said player has been in the league a decade without ever racking up per minute numbers like this before, and without ever developing a reputation as a superstar, it makes it awfully hard to swallow that said player isn't playing more because the coach getting him to play this well is an idiot who doesn't know what he has.

I do consider Chandler the 2nd most important player on those championship Mavs and someone who should get consideration for all-star status if this continues, but no, I don't think he's a secret superstar. He's overrated by this stat because the way he's shooting the ball (essentially not shooting hard shots at all) is so different from the norm.

Is it possible that the way Chandler's doing things is just the way everyone should play? Nah. Efficiency is a very important part of the game, but a player scoring 10 PPG simply isn't someone that his team can rely on when they need a bucket.

[parapooper]

I don't see the problem with Chandler being fifth in WS/48. The guy was top5 in trb% and had 70% TS.
70%TS is huge for winning. You could argue that someone scoring 30pts/game at a TS% at or below his team's TS% contributes less than Chandler scoring 10pts a game at insanely high TS%. Sure he profited from Dirk and Kidd getting him easy scoring opportunities, but he was extremely valuable to his team with what he was doing.
And he wasn't even in the top20 in WS due to his low minutes. So him being high in WS/48 is rather an argument for than against the usefulness of WS to me. What he does may not look too impressive but I think that WS/48 is correct in valuing it highly.

Well here's the thing: When you see a player score sky high in a per minute metric without playing sky high minutes, it begs the question of why the player isn't playing more minutes. When said player has been in the league a decade without ever racking up per minute numbers like this before, and without ever developing a reputation as a superstar, it makes it awfully hard to swallow that said player isn't playing more because the coach getting him to play this well is an idiot who doesn't know what he has.

I do consider Chandler the 2nd most important player on those championship Mavs and someone who should get consideration for all-star status if this continues, but no, I don't think he's a secret superstar. He's overrated by this stat because the way he's shooting the ball (essentially not shooting hard shots at all) is so different from the norm.

Is it possible that the way Chandler's doing things is just the way everyone should play? Nah. Efficiency is a very important part of the game, but a player scoring 10 PPG simply isn't someone that his team can rely on when they need a bucket.

I certainly see what you are saying. But I would say that his high WS/48 from last season doesn't say "chandler is one of the best players in the league" as much as it says something like "IF you have Dirk spacing the floor AND you have Kidd distributing the ball having Chandler next to the bucket is one of the best things you can have in the league". So it's more situational than inherent value. But there is no way for advanced stats to differentiate these unless stats over long timeframes of the player in different environments are available. For Chandler these indicate he is not as good as last year's stats indicate. So I would still say win shares gives a good value for the player in a certain setting as well as in general. So to me Chandler's WS numbers don't seem that weird.
Of course I do not know the actual formula and if you say that it does not handle outlier values like Chandlers TS% well then that is certainly sth to consider.

Why Chandler didn't get more minutes last season I don't know. Maybe it's health issues. He did get 28min during the season, 32min during the playoffs and almost 37 minutes in the finals, which seems to indicate that his coach gives him low minutes to save him for the important games.

[Doctor MJ]

I certainly see what you are saying. But I would say that his high WS/48 from last season doesn't say "chandler is one of the best players in the league" as much as it says something like "IF you have Dirk spacing the floor AND you have Kidd distributing the ball having Chandler next to the bucket is one of the best things you can have in the league". So it's more situational than inherent value. But there is no way for advanced stats to differentiate these unless stats over long timeframes of the player in different environments are available. For Chandler these indicate he is not as good as last year's stats indicate. So I would still say win shares gives a good value for the player in a certain setting as well as in general. So to me Chandler's WS numbers don't seem that weird.
Of course I do not know the actual formula and if you say that it does not handle outlier values like Chandlers TS% well then that is certainly sth to consider.

Why Chandler didn't get more minutes last season I don't know. Maybe it's health issues. He did get 28min during the season, 32min during the playoffs and almost 37 minutes in the finals, which seems to indicate that his coach gives him low minutes to save him for the important games.

Ah, an excellent observation about what Chandler's WS truly mean. I don't disagree, but from a "WS as a valid stat" perspective, you want to have more balance between the superbly helpful role player and the core players that foster the niche the role player inhabits.

[tclg]

I like to look at them. Usually I look at all the advanced stats and try to draw a conclusion based on the whole. And of course a dash of the eye test

[jambalaya]

If one uses Offensive WinShares, probably best to check offensive Adjusted +/- as well. If the first is high but the second isn't, then the individual performance of the player may be coming at the expense of the effectiveness of his teammates.
Example: Warrick.

Neither Defensive Adjusted +/- or Defensive WinShares are perfect or in my mind sufficient alone. Defensive RAPM would generally be my preference between them (since Defensive WinShares gives everyone on the team the same scores for shot defense) but 1/4 to 1/3 of the estimates may be off fairly significantly. A careful check of Defensive Win Shares and counterpart defensive stats along with defensive RAPM can help with making overall defensive judgments, especially when the metrics are not in close agreement. Using just one is more likely to produce misimpressions.

[tsherkin]

Well here's the thing: When you see a player score sky high in a per minute metric without playing sky high minutes, it begs the question of why the player isn't playing more minutes. When said player has been in the league a decade without ever racking up per minute numbers like this before, and without ever developing a reputation as a superstar, it makes it awfully hard to swallow that said player isn't playing more because the coach getting him to play this well is an idiot who doesn't know what he has.

Yeah, but they also had the expensive Brendan Haywood to use as well, and Chandler's still a foul-prone player, as much so as he's ever been. His per-minute rebounding and shot-blocking were similar to what he's always brought as a healthy player, and even his FG% wasn't out of line with what he's done a few times. So no, it's not true that he hasn't had the same kind of per-minute numbers. Yes, it's true that his per-minute win shares haven't been this high before, but his actual production has been similar. His turnovers were down a lot and his TS% spiked but it wasn't so high as to be a complete outlier compared to what he's managed in other seasons (he does have two other seasons of 63%+ TS, one of them the previous season).

In 07-08, he had a similar split of OWS and DWS but a lower overall WS/48 because he played 5 more games and about 7 more minutes per game, but again, this is his third season with 8+ WS. He's cracked 120+ ORTG before (122), so while I wouldn't expect him to maintain this full level of ricockulous offensive efficiency, given the spacing on this team, the passing they have and his limited role, sticking near this isn't totally out of the question.

Remember, just because you're really efficient doesn't mean you can necessarily maintain that efficiency if your sphere of responsibility is expanded. Chandler's been improving his FT shooting and his face-up jumper for a while now, so his utility to the team is there. He doesn't have a broad arsenal of post moves and his foul trouble limits his overall time on the floor, so there are obvious reasons he doesn't play more, but that doesn't actually cast doubt on his utility to the team, it just means in a direct comparison with a player who posts the same rate on larger minutes, that player is necessarily going to be considered more valuable.

I don't think he's a secret superstar. He's overrated by this stat because the way he's shooting the ball (essentially not shooting hard shots at all) is so different from the norm.

Yup.

but a player scoring 10 PPG simply isn't someone that his team can rely on when they need a bucket.

Yeppers, that's the one.