How legit is win shares?

[Dasher]

It takes more than 3 games for Jeff Foster to play 48 minutes. Using a guy who has a limited sample size to disprove something that generally does a good job separating the wheat from the chaff is a bitch move. I expect more from you KBlaze.

[ballinhun8]

Crediting Defensive Win Shares to players is based on Dean Oliver's Defensive Rating. Defensive Rating is an estimate of the player's points allowed per 100 defensive possessions (please see Oliver's book for further details). Here is a description of the process (once again using LeBron James in 2008-09 as an example):

Calculate the Defensive Rating for each player. James's Defensive Rating in 2008-09 was 99.1.
Calculate marginal defense for each player. Marginal defense is equal to (player minutes played / team minutes played) * (team defensive possessions) * (1.08 * (league points per possession) - ((Defensive Rating) / 100)). For James this is (3054 / 19780) * 7341 * ((1.08 * 1.083) - (99.1 / 100)) = 202.5. Note that this formula may produce a negative result for some players.
Calculate marginal points per win. Marginal points per win reduces to 0.32 * (league points per game) * ((team pace) / (league pace)). For the 2008-09 Cavaliers this is 0.32 * 100.0 * (88.7 / 91.7) = 30.95.
Credit Defensive Win Shares to the players. Defensive Win Shares are credited using the following formula: (marginal defense) / (marginal points per win). James gets credit for 202.5 / 30.95 = 6.54 Defensive Win Shares.

I usually end up with a 6.7 defensive win share when I play ball outside with my boys.


I just give em that formula when I force my matchup into a bad shot.

[Kblaze8855]

It takes more than 3 games for Jeff Foster to play 48 minutes. Using a guy who has a limited sample size to disprove something that generally does a good job separating the wheat from the chaff is a bitch move. I expect more from you KBlaze.

Its a "bitch move" to point out dozens(well...hundreds no doubt if the lists were that long) of absurd results because the absurd ones shouldnt count when some factor throws it off?

Ever consider that...if a formula cant figure out how to not rank scrubs over hall of famers its probably not accurate to begin with?

You know how I know Jeff Foster isnt better than Chris Webber?

I have seen the game of basketball be played. by both of them.

But a formula tells me Foster is higher. He has worse numbers. Is less skilled. And has not really even won anything even as a role player.

The formula ranking him over Webber(and Cowens...and Worthy...and other hall of famers) doesnt invalidate it....it just means we shouldnt count those parts?

If you have to disregard hundreds of results because of the obvious absurdity...why look at the list at all?

If you need to modify the results to make sense....**** it.

The results making so little sense they cant even be pointed out without a disclaimer is reason enough not to give them to begin with.

[Dasher]

Just because The Eye-ball Test says that you are the most impactful player doesn't mean the numbers have to bear it out.

The small sample size of a playoff run also can skew numbers. The difference, which is rather slight, between Hakeem and Clyde's win shares seems to be that The Rockets performed much better on offense when Clyde was on the floor. The difference is 10 points. The team played better on defense with Hakeem on the floor, but the difference is only 4 points.

[Dasher]

In stats outliers, like Jeff Foster are thrown out all the time.

[Doranku]

Easily the worst "advanced metric" that is thrown around on this site.

[Kblaze8855]

Just because The Eye-ball Test says that you are the most impactful player doesn't mean the numbers have to bear it out.

Of course. There isnt a number in basketball im aware of that even paints a terribly accurate picture of the skill its supposed to represent. Some are closer than others(rebounds being closer than say...ppg). But none are al lthat accurate.

And all you get when you combine a gang of them is a cluster****. It always has good players at the top....which would be the case if you combine any positive numbers in any order that makes any hint of sense.

Add the positive numbers up and remove points for the negative....the results will be good players at the top.

But its just piling up numbers that dont mean that much on their own into one number that still manages to be less than the sum of its parts.

[Kblaze8855]

In stats outliers, like Jeff Foster are thrown out all the time.

Which is very well and good. But it has what to do with basketball?

And at what point are hundreds of crazy results closer to a standard result than an exception? Not like Foster is the only one. And if you just looked at them in individual matchups...there must be hundreds(really...thousands) of examples of it painting a picture that goes against common sense.

[ShaqAttack3234]

Crediting Defensive Win Shares to players is based on Dean Oliver's Defensive Rating. Defensive Rating is an estimate of the player's points allowed per 100 defensive possessions (please see Oliver's book for further details). Here is a description of the process (once again using LeBron James in 2008-09 as an example):

Calculate the Defensive Rating for each player. James's Defensive Rating in 2008-09 was 99.1.
Calculate marginal defense for each player. Marginal defense is equal to (player minutes played / team minutes played) * (team defensive possessions) * (1.08 * (league points per possession) - ((Defensive Rating) / 100)). For James this is (3054 / 19780) * 7341 * ((1.08 * 1.083) - (99.1 / 100)) = 202.5. Note that this formula may produce a negative result for some players.
Calculate marginal points per win. Marginal points per win reduces to 0.32 * (league points per game) * ((team pace) / (league pace)). For the 2008-09 Cavaliers this is 0.32 * 100.0 * (88.7 / 91.7) = 30.95.
Credit Defensive Win Shares to the players. Defensive Win Shares are credited using the following formula: (marginal defense) / (marginal points per win). James gets credit for 202.5 / 30.95 = 6.54 Defensive Win Shares.

How can you possibly make any sort of estimate of how many points an individual player gives up without +/- numbers available? This is especially true for players in the 60's.

An estimate of how many points that player allows....yeah, I can tell you where that estimate came from.

Defensive ratings for teams actually makes sense. Points per 100 possessions. Very easy to explain and a logical way to measure a team's defense if you're going to try to look to stats for that.

But for players? Again, box scores don't tell you who that player was guarding and for how many possessions, or what that did player did when the other player/team was off the court.

Just because The Eye-ball Test says that you are the most impactful player doesn't mean the numbers have to bear it out.

The small sample size of a playoff run also can skew numbers. The difference, which is rather slight, between Hakeem and Clyde's win shares seems to be that The Rockets performed much better on offense when Clyde was on the floor. The difference is 10 points. The team played better on defense with Hakeem on the floor, but the difference is only 4 points.

And again, how could they state even statistically that the Rockets performed better in those situations without complete +/- numbers available? So that doesn't explain the playoffs either.

[KevinNYC]

Its a fine stat. People don't understand what it is or what it does so they whine about it. All is does is breaks down the wins a team had and who was statistically responsible for those wins. If you miss 10 games you get a zero for those games.

The Olajuwon line that was brought up came from a year when they were the 6 seed and he missed a lot of time. Schremf played on a 62 win team. Barros had a ridiculously efficient line.

http://www.basketball-reference.com/about/ws.html

So it sounds like you can use win shares to compare seasons, but you may also need something like winshares/game to account for an absence.

Also does that mean the very best player on the worst team gets less winshares than the worst starter on the best team?

[Dasher]

The offensive rating of The 1995 Houston Rockets with Clyde on the floor is 120. Hakeems Offensive Rating was 110. Per 100 offensive possessions the team performed better with Clyde on the floor.

The team predictably fared better defensively with Hakeem on the floor. 108 for Hakeem to 112 for Clyde.

I still think arguing that Hakeem's 1995 invalidates Advanced statistics like Win Shares is shaky because the 1995 Rockets are arguably the flukiest champion aside from the 1974-75 Warriors.

[Kblaze8855]

Also does that mean the very best player on the worst team gets less winshares than the worst starter on the best team?

No but it seems to be a factor. For example...Tmac in 04 had 8.4 win shares. Jeff foster(just by chance its him again...his team had the best record) was 3rd in win shares on the pacers with 8.3. Third on his team in win shares had him almost the same as a top 5 player in the NBA. He was probably the 5th or 6th best player on the Pacers(Jermaine, Artest, Reggie, and Harrington at least have a case to be over him). Reggie only did 10ppg on 44% shooting and had 8.9 win shares since he was in almost all the games and they won a lot. Prime Tmac on a bad team having missed some games...8.4. Last legs Reggie doing 10ppg on 44% shooting on a good team in most of the games....8.9.

[ShaqAttack3234]

The offensive rating of The 1995 Houston Rockets with Clyde on the floor is 120. Hakeems Offensive Rating was 110. Per 100 offensive possessions the team performed better with Clyde on the floor.

But how could they know this without having play by play stats available for the game, or without going through each game and tracking them....which would be too time consuming and it wouldn't explain how they'd come up with individual's offensive/defensive ratings for before such a thing would be possible.

[Kblaze8855]

The people who come up with many of these systems wouldnt want to even attempt to make a play by play. They arent basketball people in some cases. Isnt win shares a concept taken from baseball and modified? I feel like I read that at some point.

[HurricaneKid]

The people who come up with many of these systems wouldnt want to even attempt to make a play by play. They arent basketball people in some cases. Isnt win shares a concept taken from baseball and modified? I feel like I read that at some point.

Yes. It certainly is. You probably read my post on that a week or so ago. Baseball analytics are so far beyond all other sports because of the huge data sets and the limited number of variables (always a pitcher vs a hitter) so they have an amazing level of accuracy. Attempts to use similar statistical analytics in FB are borderline inaccurate due to the enormous number of variables, the limited sample sizes, etc. Basketball falls in the middle. It is fairly accurate but a few weird data points give enough credence to those who would disparage its fundamentals.