Here's a graph of WAR (on the x axis) and WSAB (on the y) for all position players from 1900 to 2008:
As you can see, the two stats not only track very closely, but they almost intersect at zero (I divided WSAB by three to make them equivalent to the WAR scale). If you're a math nut, the R squared between the two is .96. That's the good news.
However, as I showed the other day, the difference between WAR and WSAB can matter a lot for individual players. Also, the variance between the two is greater for individual seasons, as you would expect. That's the more discerning news.
I've got no other point. Just thought you'd like to know.
Dave was called a "national treasure" by Rob Neyer. Seriously. Comments about this article can be sent to him through the miracle of e-mail.
Comments
Rally said...
It would be very interesting to see, once the measures are put on the same scale, the players they disagree on the most.
Posted 08/14 at 12:36 PM
jinaz said...
Nice to see they match up so well. Corresponds to Jeff’s post a while back at BtB.
So, let’s see, what are the likely causes of any differences…
1. WAR and WSAB use slightly different baselines, right? Bench vs. Replacement? I’d think that this would be a cause of the differences you mentioned, simply because WAR will give more credit for playing time.
2. Fielding. WAR has the more detailed fielding metric (TotalZone with Rally, UZR with FanGraphs). Might make a difference for some players, at least in the retrosheet era.
3. Position adjustments. I have no idea how win shares does this, but I’m sure it’s not based on fielding talent disparities as is the case in WAR.
Others? Maybe issues of team wins vs. hypothetical wins?
-j
Posted 08/14 at 01:48 PM
Dave Studeman said...
I think the baselines are extremely similar. As proof, they both cross very close to zero.
The hypothetical vs. real wins are definitely an issue, but the biggest one is the difference in fielding between the two systems. Obviously, there are lots of smaller issues, too.
I have a feeling that the difference for pitchers is much larger. That’s the next step.
IIRC, Jeff compared Win Shares vs. WAR, not WSAB.
Posted 08/14 at 02:07 PM
Cyril Morong said...
Dave
Really interesting. Which player had the biggest differences between the two metrics? How many players were in the study and what was the PA or playing time minimum? Thanks.
Cy
Posted 08/14 at 04:15 PM
Dave Studeman said...
Hey Cy, there were 6,800 player careers in the study with about 40,000 player seasons—no minimum playing time, except that I excluded player seasons with expected Win Shares of zero, which means they probably had ten at bats or less, something like that.
The players with the biggest variances toward WAR were the best fielders in premium position (Win Shares doesn’t have a position adjustment for anything except fielding). Ozzie Smith, Mark Belanger, etc.
Conversely, those with the biggest slant toward Win Shares were sluggers: Gary Sheffield, Frank Howard, etc.
I’ve got an article planned for that.
Posted 08/14 at 04:32 PM
Cyril Morong said...
Dave
Thanks. I am not surprised about the guys at premium positions. When I looked Win Shares per 648 PAs a few years ago, I found that SS and 2B had a lower average than OFers. OFers had 21.13 while SS had 18.44, Seems like a pretty big difference.
It would be very interesting to see, once the measures are put on the same scale, the players they disagree on the most.