I like the site, but it would be nice if you presented Win Shares above replacement(the 50% of average).
Posted by .(JavaScript must be enabled to view this email address) on 05/31 at 12:04 PM
Win Shares Winning Percentage
May 31, 2004
Over at The Hardball Times, I’ve included something in the Win Shares stats called Win Shares Winning Percentage (or WSWP). Now, winning percentages and loss shares don’t really work in the Win Shares system, as we’ve discussed before. But I thought it would be enlightening to include some sort of Win Shares rate stat, so I gave it a shot.
The formula I used is simple: WS/(WS plus ExpWS). The beauty of this formula is that it expresses a Win Shares rate as a winning percentage (from .000 to 1.000) except for players with negative Win Shares (whom I zeroed out). The bad thing about this formula is that it’s not quite correct, for a couple of reasons I can think of:
- A true winning percentage formula would be WS/(2 times ExpWS). So as a player moves from his expected Win Shares total, the formula turns increasingly wrong.
- On the other hand, extreme players like Barry Bonds are not given enough credit, because their “negative Loss Shares” are not given full credit.
In a way, I’m trying to create a formula in which two wrongs make a right. So this is a formula that “satisfices.” It’s not quite correct for a lot of players, but it’s not bad overall, particularly for player comparisons.
The other idea I had is to just list the ratio of Win Shares to Expected Win Shares. This is probably a more honest metric, but it’s not as intuitive to many readers.
What do you folks think? Should I go with one or the other, or just drop the idea altogether?
Thanks, Hi. I was going to do that, but it’s not at all clear that 50% is the right replacement level. It sort of depends on what you’re trying to measure. I like to use 50% for salary analysis because I think it’s representative of “Freely Available Talent.” For straight player analysis, however, a higher number is probably appropriate.
So I wimped out and let the reader choose. Maybe that’s a mistake—any other thoughts?
Posted by studes on 05/31 at 01:10 PM
Didn’t like WSWP, cause it’s not quite right. Instead, I’ve settled on this formula: WS/(2*expWS). It’s good in that it is close to a winning percent formula. It’s bad in that players can go above 1.000 and below .000. But that’s Win Shares for you.
What I do like about this new formula (which I’ve called WSP) is that it equals the team’s winning percentage, if you take the WSP of all players, weighted by their proportion of the team’s expWS.
Posted by studes on 06/05 at 08:07 PM
Page 1 of 1 pages
Commenting is not available in this channel entry.
| <<Previous Article: 2004 Win Shares Have Arrived | Next Article: Orioles' Win Shares>> |