THE BOOK--Playing The Percentages In Baseball

I wonder - if Wins are never going to go away, has anyone ever come up with something like Expected Wins (xW)?

What I mean is this - can we use something like Win Expectancy to see, for instance, what percentage of games where the starter is charged with 3 runs in 5.2 innings end with his team winning? And then, depending on which side of 50% the result lands on, we can assign an xW or xL?

Sure, it’s still problematic for all the usual reasons, but at least it would effectively neutralize their offenses in order to see whether the Ws and Ls are deserved or lucky, right?

Michael Wolverton did that very thing back in the mid-90s with his Support Neutral stats.

There are a few stats like that already.  I don’t know that they are more than a bit better than what we have.  The problem is really with the whole “binary” part of the pitcher winning or losing.  You have Weaver v Verlander, and one of them gets a W and the other an L, even if 1 run is scored in the whole game.  How does this help us?

Well, presumably, some sort of xW stat would award them both a W under the presumption that their team *should* win that game, right? Which is not to say that it’s really a great stat, but the traditionalists might be more easily swayed because it *looks* like the traditional stat. And if W-L isn’t going anywhere any time soon…

If you are going to allow both pitcher’s to “win” or get partial wins such that the total is over 1.00, then Game Score is what you want.

Or about a dozen other metrics out there.  This is not going to appease the traditionalistas, not that we really care to.

All I care about it removing pollution.  I don’t care about necessarily selling them on a better form of energy.

If you are going to allow both pitcher’s to “win” or get partial wins such that the total is over 1.00, then Game Score is what you want.

Not if you want to measure run prevention and don’t want to assume that DIPS applies at the game level.  And if you want to measure in fractions of a win.  Michael Wolverton’s stats are exactly what’s called for in that circumstance, not Game Score.

I thought Wolverton’s stats were well-known among saber folks who’ve been around for a while.  IIRC, they were around in the Usenet days of baseball analysis before they made it to Baseball Prospectus.

When I say “Game Score”, I’m not limiting myself to the James version.  I have 4 different versions, each of which applies to different factions, from a pure FIP to a pure runs allowed, etc.

Each of those is on a scale of 0 to 100, and would be on a win scale.

As far as xW mentioned above, isn’t that basically what QS does? Rewards a pitcher’s performance that *should* be enough for “a win.”

I think the world would immediately become a better place if we just used Team W-L when a pitcher starts instead of pitcher W-L. Should a reliever really be allowed to get “a win?” Really? You came in, picked off a guy, we scored 2 runs… and you get the win?

Sure, Team W-L has flaws, but I think it has fewer than people acknowledge. If a team scores runs in the 9th to “bail out” a starter on the hook for a loss, how is that actually different from the team that scores the same runs in the 1st to put a pitcher on pace for a win?

8/

I’ve always wondered why team W-L in games started is ignored.  I’d be very surprised if that wasn’t better correlated to pitcher skill than wins as currently defined.  Plus, if “wins are what matters,” then whether the team won or lost the game is the result, the team doesn’t get to take a no-decision. 

I’d love to know why the current wins stat gained favor over team W-L in games started.

SittingCurveball/8: QS isn’t a terrible idea, but I think that the parameters stink. If a pitcher throws 5.2 scoreless innings, he doesn’t get a QS; if he throws 8 and allows 4, again, no QS. And if a QS is like a W, what’s akin to an L?

We don’t like the binary decision to lead to another binary result.  So, the QS doesn’t help, because there’s always going to be a problem on the boundary calls, and the end result is always either “1 W or 0 W”, which is a huge gap.

Hence, Bill James created Game Score, which gives you a continuous scale.  You no longer have to decide if a “72” is a QS or a “63” is a QS.  It’s simply a 72 or 63.

I then enhances that by creating 4 different Game Scores, because some people live and die by runs allowed being the only thing, while others swear to just FIP in being used.

There’s a solution that will appease everyone.  What we won’t get is coming up with a single solution.

Even WAR only appeals to less than half the people, because a substantial number of people still prefer a higher comparison level (average) and others an even lower baseline level (closer to zero).

All you can do is present different options, so that each person can pick and choose which one he will go for.  Heineken, Sam Adams, whiskey, or OJ.  There’s no need to have tension.