THE BOOK--Playing The Percentages In Baseball

Did Josh make a huge mistake in the numbers?

A strike is worth about -.08 runs (and a ball around +.08 runs).  Something like that.

So, a pitch down the middle, that is not in play, would, necessarily be either a strike, or a “let” (2-strike foul).

I don’t see how it’s possible that it could average out to -.03 runs.

What am I missing?

He’s summing vertically, so you get balls high and low there.

Maybe some of them are “down the middle” horizontally, but either high or low and, hence, called balls?

"So, a pitch down the middle, that is not in play, would, necessarily be either a strike, or a “let” (2-strike foul).”

I didn’t eliminate pitches that were above and below the zone, so what’s in the quote is not true. The pitch could be down the middle, but below, or above the zone, and be a ball.

I think it would make more sense to restrict the study to pitches that are vertically within the strike zone.  I usually use pitches between 2.1 and 3.1 feet high for things like this (based on this: http://www.baseballprospectus.com/u/images/strike_percentage_by_height.jpg).

There are a number of complicating factors with an analysis like this, but that would at least remove one of the biggest.

Don’t think I can post the picture here without going through a third-party, but if I restrict the pitches to those with a height between 2.1 and 3.1 feet, the run value of a pitch not put into play is about -.06 when down the middle.

Excellent stuff, Josh!

Echoing Mike Fast’s point about the vertical, it’s a crucial piece - I was hoping to see a graph like the second one in your article, but for balls in play (and also including home runs)…

Have you got any further plans to slice the data in different ways, e.g. pitch type, count, etc.?

Ah, excellent guys, thanks.

To post image, just do

As several of the comments on the site pointed out, sometimes the results from aggregating data yield a misleading or at least a not very useful conclusion.

For example, let’s say that there are only 2 times that a pitcher throws a pitch down the middle: One, on a 3-0 count and the batter never swings, and two, at an 0-2 count and it is a fastball and a mistake.

The overall value of the pitch might be positive (for the pitcher) or neutral, but that is not a useful piece of information and tells us nothing about what might be correct on either an 0-2 or 3-0 pitch.

This can happen quite often with baseball research - where it is imperative to break up the data.

Again, not saying anything that Josh doesn’t know or that the commenters on the site have not pointed out…

The goal of the study was not to quantify the exact value of throwing down the middle, or when it’s best for the pitcher to throw the pitch down the middle.

My intention was to show that the general perception that “throwing down the middle is always really bad” is silly. And I do think a lot of people believe that.

Were I more focused on the “true” run value of a pitch down the middle, then yes, I would be much more concerned about breaking down the data a lot more. Finding the “true” value would be something really complex. 

And to answer Greg/#7:
I’m not sure how much more time I’ll spend on command. Right now I’m thinking more about pitch sequencing which there isn’t a lot of research about. But I could probably generate a run value map for every combination of count and pitch type and batter handedness without that much work, but I’m not really sure what that would tell us.

Well apparently I can’t resist making baseball graphs. In the link is a pdf with run value graphs for every single combination of:

pitcher handedness, batter handedness, pitch type, and count.

Uses MLBAM classifications, and I only made the graphs for subsets of data with at least 2000 pitches. Also didn’t adjust the strikezone box for batter handedness, so keep in mind that for lefties it should be a little more to the left.

The graphs are a little ugly, but this type is faster than the method used to create the graphs that I used in the article. 

https://docs.google.com/open?id=0B0i1tMtbkcC9YzA1MTk1NDktOTkzYi00NTQwLTlmNTQtMzdlYTQyOGUzZWIx

Not really sure if I used the best smoothing parameter. I can post the code if someone wants to see what’s going on.

My intention was to show that the general perception that “throwing down the middle is always really bad” is silly. And I do think a lot of people believe that.

Again, no one thinks that throwing the ball down the middle on a 3-0 count is always or even usually bad so I’m not sure how looking at the run value without separating the data can lead you to a useful conclusion.

Josh, can you change your layout to a 3 x 4 (by pitch type), like here:

http://www.tangotiger.net/halejon/allcounts.html