Now that baseball is back (for now, at least…) I thought it would be fun to do another look at some stats! Specifically, from a given ball-strike count, what are the odds that the plate appearance ends in a walk or a strikeout?
Like last time, I extended the parsing code of my baseball win expectancy finder to gather these statistics. (it’s available on GitHub, see
parseretrosheet.py) Without further adieu, here are the numbers!
Here are some interesting observations:
- From any given count, getting a ball doubles your chances getting a walk. This isn’t exact, but it does almost hold at every count, which is interesting. At first I wondered if you could model this as there being a constant chance of getting a ball or a strike from any count – this would also explain why from a 0-0 count a strikeout is twice as likely as a walk. (since you only need 3 strikes for a strikeout but 4 balls for a walk) But there doesn’t seem to be a similar relationship for strikes.
- The 1-0 and 2-1 counts have basically the same probability for a walk as a strikeout, which seems reasonable since it takes the same number of balls to get a walk and strikes to make a strikeout. But for a 3-2 count there are significantly more walks than strikeouts. I’m not really sure what to make of that.
- 3-0 is a weird count. There is evidence (not from this data) that the strike zone gets wider at this count, so the same pitch that might have been a ball some other time will be a strike here. But you still have a 2/3 chance of ending up getting a walk, which is very high! It’s rare to get a strikeout after getting to this count (5.0%), but not as rare as getting a walk after getting to 0-2 (2.8%)
- You’re significantly more likely to get a walk from 3-0 than a strikeout from 0-2. This kind of lines up with the 3-2 weirdness – maybe balls are just a bit more likely than strikes, all else being equal?
This is a sample of just under 6 million plate appearances, although the numbers for counts other than 0-0 are fewer. The least common count was 3-0 – there are just under 300,000 of these. This is all pulled from Retrosheet data that have pitches recorded which range from 1957-2019, but it’s heavily weighted from 1988-2019 (where all games have pitch data, I believe); before that seasons had 1,000-10,000 plate appearances with pitch data, after that it was more like 160,000.