THE BOOK--Playing The Percentages In Baseball

Hmm...you hear first pitch strike statistics cited all the time.  The chart seems to suggest getting a strike on at least one of the first two pitches might be just as important a measure.

I think this image is one of the best presentations I’ve ever seen.  Look at the information he’s conveyed:

1. The size of the circle tells you the frequency
2. The pie slices tells you the proportion.
3. The state-to-state transition is shown by the ball or strike
4. He tells you how often a PA ends at that count
5. Each pass-through count shows the wOBA

Simply a wonderful and perfect illustration of linear weights by count.

Here is the latest iteration of the graph:
http://img220.imageshack.us/i/splitcount20101214.png/

Based on a comment by Patrick in the Fangraphs comment, I added the ball/strike split on AB-ending counts (3-0, 3-1, 3-2, 1-2, 0-2), and based on an earlier request from Tango, I added lines for each .25 of wOBA (instead of the .50 I had previously).

Brilliant.  Conveys a wealth of information elegantly. 

This should be used in presentations on the basics of Saber theory.

That graph pretty much represents the essence of baseball in a way that numbers on their own just could not get across.

Look what more it tells you: in hitter’s counts, you are more likely to transition to another state via a strike than a ball.  In a pitcher’s count, you are going to transition more often on a ball than a strike.

I asked Josh to complete the chart to transition into the walk/strikeout states as well.  It’ll show the impact of what happens when you get to the .000 or .720 wOBA levels.

This chart also makes me want to look at variation in 2-strike effectiveness between batters.  I suppose I knew this in some form, but I’m surprised by how effective batters are at avoiding the strikeout once they get to 2 strikes.

Based on more comments from Tango, here is the result:

http://img20.imageshack.us/img20/6922/splitcount20101214bb.png

Excellent!

Your 3-2 count is wrong for the pass-through wOBA.  Way too high.

Maybe how you handle 2-strike fouls is incorrect?

One minor thing: The color shown in the legend for “At-Bat Ends” doesn’t seem to match what’s in the graphic. Regardless, excellent work (again), Joshua.

tango, you’re spot on. I was using the wrong line for 3-2 “through” count (using all counts with 3 balls accidentally instead). Here is the updated version:

http://img97.imageshack.us/img97/6922/splitcount20101214bb.png

NaOH: You’re quite right. AB ends should be changed to Batted Ball. Before I had left out walks/strikeouts as a pitch count, and just included them in the grey (hence the label). I should change that.

I have updated the main image.  Thanks to Josh for a fantastic presentation.

I agree, great stuff.

After 10 minutes of staring at the chart, I was still seeing new things.

Thank you for the kind words.

Very interesting.  Great job.

One nitpick: there is a path-dependency that is not captured here.  For example, a 1-1 count that comes via first-pitch ball is different than a 1-1 count that comes via first-pitch strike.

I looked at this a few years ago at THT and never really pinned down the whole framework. I bet Josh could do it, though.

Very interesting.  Great job.

One nitpick: there is a path-dependency that is not captured here.  For example, a 1-1 count that comes via first-pitch ball is different than a 1-1 count that comes via first-pitch strike.

I looked at this a few years ago at THT and never really pinned down the whole framework. I bet Josh could do it, though.

salb, I tried, but it doesn’t really seem to be doable with the (little) data I have. I’d have to get the data before doing something with it. If you know of somewhere I can get it (or if you have it), I’d be more than happy to give it a shot.

Josh, I culled the data from Retrosheet.  One of the database mavens in the saberworld could probably get you the data from gameday.

I think the best way to do it would be through an interactive graph where a mouseover one of the counts would give you the graph for all pitches that went through that count. It would be all but impossible to put in a single graphic.

Ways to get to each count:
0-0: 1
0-1: 1
0-2: 1
1-0: 1
2-0: 1
3-0: 1
1-1: 2
2-1: 3
1-2: 3
3-1: 4
2-2: 6
3-2: 10

(at least on a quick look at it)

I remember reading your article on THT about that, I will look it up and take a few pointers, then try to track down the data so I can see what it shows.