from Amazon
Friday, December 04, 2009
Max on run value of pitch, by count
By the way, I call them “plate count” to distinguish them from “pitch count” (like, he threw 121 pitches). If anyone’s been similarly bothered, you can see if my alternative helps you a bit.
I’m trying to understand Max here. First, I like his graph here, which makes it clear:
You want your slider to go through the plate high, outside, or middle-and-in. And if you are going to miss the strike zone, make it down-and-away. Basically, avoid down the middle (like for any pitch) and avoid down and in. I get that. Very nice pic.
On a 1-0 pitch, it’s the same kind of rule, but make sure it’s in the strike zone (and down the middle is not so costly):
Basically, this means that batters have shrunk their strike zone on the outside, and interestingly, have shrunk their “down the middle” strike zone as well. That is, a pitcher can’t get away from throwing a pitch on the edge of the strike zone, but can get away (a bit more than usual) if it’s down the middle. I thought that was interesting.
On an 0-1 however, things get clearer:
You must avoid throwing low-and-in at all costs. The batter has severely shrunk the strike zone. Rather than it being a pitcher’s count, it is now a hitter’s count!
HOWEVER, Max’s chart shows us that the batter actually swings at more pitches at the 0-1 count (as you’d expect). Basically, the charts don’t match.
If I have to guess, I’d say that batters are killing the 0-1 sliders, and they are probably letting the 1-0 sliders be called for strikes more than they should.
I dunno. I think. I’ve had his article up since this morning, and only now have I been able to try to figure it out. I’m not sure I have.
That said, I love the article.
Tango - If you’re still unclear as to the point of the article, it was a response to a few concerns MGL had with some recent Pitch f/x articles using run value. You can read them here:
http://www.hardballtimes.com/main/article/all-about-the-high-inside-fastball/
http://www.hardballtimes.com/main/article/medium-and-short-time-memory/
His concern was that run values didn’t properly account for “plate counts”. I disagreed, as the way run values are calculated are by taking the linear weights value of each pitch and subtracting that by the linear weights baseline for that count. That should do a good enough job of “count adjusting” run values. However, as Max shows here, you also have to adjust for pitch type and location, which makes it a lot more complicated.
On another note, after reading Max’s article I checked my database to see the average run value (per 100 pitches) by count:
That’s a lot more variance than expected. I suppose selective sampling could play a role (good pitchers will be in more pitchers counts than bad pitchers, and vice versa), however, for the 3 ball x strike counts, we see a far larger effect than you would get from selective sampling.
Does anyone know what’s up?
Does anyone know what’s up?