THE BOOK--Playing The Percentages In Baseball

Here’s another comparing Finley and Lofton in the Outfield.  I was lucky enough to have access to the BIS data in 2004, and the first thing I did was to make plots of graphs just like that.  I started doing it for the league overall, and let me tell you, it’s tons of fun to do so.

One thing that is very interesting to do is to split up the graphs against LHH and RHH, or if you have enough data, first assign each hitter what kind of hitter he is in terms of his “sprayability”.  Does he like to hit it to LF, CF, or RF?  Group those hitters, and then, track how fielders do against those hitters (including handedness).  This is a great way to see how well fielders are positioned against hitters.

I looked at MGL’s UZR for the same time period.  There is a 45 run gap between the two, in Lofton’s favor, which compares favorably to the 40 run gap according to ProTrade.

All these systems, Protrade, MGL, Pinto, Dewan, etc, are all based on the same premise: what is the chance of this play being made.  The inputs are fairly straightforward: handedness of batter, location of ball hit, distance of ball hit, hardness and trajectory of ball (though flight time would be loads easier), park, and whatever other variables makes each of these systems unique.  You may need a smoothing function so that you get a consistent pattern, so that two adjacent zones aren’t wildly different.

The key is the data.  You either have to luck into the data, or pay alot for it.  The other key is time.  Data + time equals great fielding system.  It’s pretty straightforward. 

Every chance I get, I go on a rampage on giving a stopwatch to the scorers.  And the scorers always tell me the same thing: they are overworked, so how can I make such a suggestion?  They are right.  The NHL has something like 5 to 7 scorers recording tons of data.  MLB I believe has one official scorer, and then a PBP guy.  I think that’s the extent of it.  The NHL revenue stream is less than half of the MLB revenue stream.  If they NHL can afford 5 scorer, MLB can afford to have 10 per game.  I’m *not* suggesting they have 10, for logisitcal reasons.  But, they can certainly afford it.  If logisitically the NHL can handle 5 to 7, then so can MLB.  But, all I’m saying is to get 1 more, and give the guy a stopwatch.  And maybe he can record the fielding location of all the fielders before each pitch.

Clements’ Protrade ERA is 3.62??  I don’t believe it.  His FIP is 5.42.  That would imply that just about every ball in play he has allowed would be fieldable!

Okay, I’m overstating the case, but not by very much.

So protrade is saying that a guy with a 43:44 K:(BB+HBP) ratio is a 3.62 ERA pitcher?

Anyone know the last pitcher to have an ERA anywhere near that low with more free passes than strikeouts?

Yeah, I’m a big-time Red Sox fan and I can tell you guys that he doesn’t deserve a 3.62 ERA.  I would like to see the full methodology behind their system because there is no way in hell Clement is pitching at that kind of ERA.  If anyone on the Red Sox has been “unlucky,” it’s probably Beckett, whose had a few games where he’s given up like 7 runs in 2 innings, which bloated his ERA.

Tangotiger:

Can you expand a bit on some of the math and theory behind szERA?

Thanks.

- Kent

I took all the pitchers since 1994, with a minimum of (I don’t remember how many BFP… 500? 1000?), and summed their K, BB, ER, BFP, IP and simply ran a regression of K/BFP, BB/BFP against ER/IP*9.  What I ended up with, to my great delight, was something like:

5.40 - 12*(K-BB)/BFP = ERA

The correlation was something like .9x, IIRC.  In essence, a pitcher’s ERA can be almost completely explained by a pitcher’s K, BB differential.