The Win Shares Baseline

Oh, I totally understand the relationship between 0.52 and 1.52.  I totally understand that.  And I find the “Why .52?” part of James’ book funny.  He has absolutely no good reason, other than it looks good.  That and his strange abhorrence for negative offense.  I haven’t seen teams with negative offense yet, but if you’re okay with players having negative offense, then you should be okay with making that number higher.  My presumption is that a pitcher should be as valuable as a non-pitcher.  To me this means, that we want a number such that PitWS = BatWS+FldWS.  Strangely enough, the golden ratio seems to be that number!

The other number I’m talking about changing is the 0.52 to determine marginal runs for BatWS.  That number seems way to high for me.  And I think that James’ whole discussion of replacement levels misses the point.

Well, just a couple of last comments.

I actually think that James is right on in his Replacement Level section.  It’s an extremely complex subject (even more than park factors) and you have to be careful about it.  Changing the 52% to approximate a replacement level is a mistake, IMO, because we should talk about replacement players, not replacement levels of skill.  This is an issue James refers to.

Also, the PitWS=BatWS+FldWS doesn’t make sense to me.  Batters account for nearly 100% of runs scored; pitchers and fielders split responsibility for runs allowed.  I actually think that major league salaries are probably close to right: 40% of salaries are paid to pitchers.

I’ve played around with win shares a bit myself and believe that “game shares” are, in fact, the key to making the concept work, esp. as regards pitchers.  IOW, rather than trying to find ways to get pitchers more win shares, we should be reporting the game shares, and comparing, for example, Pedro picking up 25 WS in 27 games, to A-Rod picking up 33 WS in 40 games (just picking numbers out of the air).

The main problem I’ve run into, however, is that comparisons between the NL and AL don’t work because of pitchers hitting.  James does a good job of explaining why he DOESN’T adjust for pitchers hitting in win shares, but once you introduce game shares you wind up with NL pitchers picking up a bunch of “offensive game shares” that their AL counterparts don’t.  And since very few pitchers pick up any offensive win shares, this has a very real impact.

From studes’ charts, it appears that NL pitchers were, collectively 185.6 WS below average last year, whereas AL pitchers were only 7.1 WS below average (which is well within the range of either a rounding error or the actual difference between the AL & NL in interleague competition last year).  And those extra WSAA have to be redistributed somewhere, which leads to NL OF being 146.5 WS above average (all AL positions fall within 20 WS of average).  So using WSAA is actually less accurate than using straight WS when comparing players across DH & non-DH leagues.

You’re absolutely right, Paul.  Comparisons across leagues don’t work with this WSAA approach.  I think this is the main reason James included negative batting claim points in the pitching win shares calculations themselves.

My suggestion for negative batting shares sort of blows that apart.

Charlie has suggested that there be a second WSAA for pitchers only, which takes out the batting portion.  In that way, you could compare pitchers across leagues and years.

But you also may be pointing out a flaw in the system that I should tackle.  Perhaps’ pitchers’ batting win shares should be pulled out of the overall batting Win Shares averages, and applied just to pitchers.  That would increase the expected average Win Shares for everyday players, and decrease them for pitchers.  It would also make the total WSAA for pitchers equal zero, even in the National League.

This makes sense to me, and I’ll play with it over the next few days.  Thanks!

What I figured was best for WSAA (or XWS, as I call it, since I like just presenting the two numbers side-by-side and then everyone picks his own baseline) is that pitchers are considered to be expected to contribute 0 Win Shares as hitters. You need to make a small adjustment for those few players who both pitch and play somewhere else, but basically, for pitchers, you zero these out. I had some other things in there, but I didn’t use Negative Win Shares and had some ways I would use to correct for it, which are unneeded with Negative Win Shares.

One other thing I think would work, and I don’t know if studes did this or not, would be to adjust hitting XWS for the DH rule. It’s a little work, but it does lower the XWS for everyone in the AL.

I’ve just finished reconfiguring the approach to National League pitchers, and the results are very interesting.  Although no pitcher jumps ahead of the top five in WSAA, they now occupy the sixth through eighth spots, and also account for nine of the first seventeen players in WSAA.  I’m going to rewrite the article to explain what I did.

One caution, by the way: WSAA is still not the correct way to go.  We will eventually have to take the next step to replacement level for this methodology to make sense, and that might change pitchers’ relative ranking again.

Charlie, I didn’t follow your DH methodology, because I think it isn’t necessary, and total average Win Shares wouldn’t foot with the league totals.  If you think I’m missing something, let me know.

Yes. Total Average Win Shares would not foot with league totals. I see no reason why this is important, incidentally, if you zero out expected hitting Win Shares for pitchers.