THE BOOK--Playing The Percentages In Baseball

Heres’s a quick look at pitches from all RHP this season in the questionable zones, with a similar buffer zone for height

RHP vs. RHB
Inside-860 pitches
21% balls
21% called strikes
8% swinging strikes
26% BIP
24% foul

Outside-1535 pitches
24% balls
30% called strikes
11% swinging
18% BIP
18% foul

RHP vs. LHB
Inside-620 pitches
36% balls
18% called strikes
8% swinging strikes
16% BIP
22% fouls

Outside-1542 pitches
16% balls
30% called strikes
8% swinging
26% BIP
20% foul

A few notes.  Both RHB and LHB swung at 45% of pitches to the umpire’s right hand side (inside to a RHB), compared with 55% on the ump’s left.  There were a higher percentage of called strikes on the outside part of the plate for each group of hitters, as umps called 20% of inside pitches strikes and 30% of outside pitches.  For LHB, only 16% of outside pitches were called balls, compared with 36% of inside pitches.

Many things could be causing these differences between inside/outside pitches.  A couple of possibilities are...the strike zone might not be centered at 0, a calibration error at MLB could be occurring, or umpires might not have the same strike-zone as Gameday.

Very fascinating.

When the umpire made a call:

v RHH: inside, 50% ball
v RHH: outside, 44% ball

v LHH: inside, 67% ball
v LHH: outside, 35% ball

When the batter took a swing:

as RHH: inside, 14% swing-and-miss
as RHH: outside, 23% swing-and-miss

as LHH: inside, 17% swing-and-miss
as LHH: outside, 15% swing-and-miss

When the batter made contact (in play or foul):

as RHH: inside, 52% in play
as RHH: outside, 50% in play

as LHH: inside, 42% in play
as LHH: outside, 57% in play

***

The umpires, in this sample, called tons of strikes on outside pitches, against LHH, while calling tons of balls on inside pitches.

And, these batters, on outside pitches put alot of balls in play, once contacted.

It’s possible that we have a calibration issue with LHH.

***

The umpire is also set up over to one side of the catcher, and that may be a parameter to consider.  I don’t pay attention how umps sets up with the handedness of the batter and pitcher. Are all umps the same in this regard?

I ran the overall numbers that now include almost 41,000 pitches and the “agreement” is 81% given a strike zone that is +/- .96 inside to outside and +/- .121 top to bottom. It also appears that the agreement has improved slightly when you look at it by day. This makes sense since I think when a new ballpark comes online the calibration is included in the file posted at the end of the game. I worked the 5/10 game in Colorado when the system was being tested and there is data in my database for 91 pitches from that game.

It’s interesting that the QuesTec data from 2002 had an agreement rate of 84.8% and when scoring the umpires they used a 2 inch buffer in which the calls were deemed acceptable. In QuesTec, the operators also threw out about 3% of the pitches as “bad tracks”. I’m not sure what’s happening with that in EGameDay although 15.4% of the pitches in games in which EGameDay was used are not tracked. Some of that is because of the tester games but I haven’t calculated how much.

Interesting stuff.

Dan, can you clarify your first statement?  Did you mean 1.21, rather than .121?  And, the “agreement” is strike calls in strikezones and ball calls in outside strikezones?

Can you generate it in terms of zones?  After all, at the inside/outside and up/down +/- .50 feet, there should be virtually 100% agreement.  They are in effect “noise” to the data.

And the same for the really outside zones, say anything outside the +/-1.50 feet area.  That is, any pitch outside the 3ftx3ft area should have 100% agreement.

The “questionable” area is the 0.50 to 1.50 area (though you can certainly come up with better numbers by noting at what point it no longer sticks to the 99%+ agreement level).

Much thanks to all you guys who are digging in here.  I really ought to jump in on this too.

The total width of the strike zone should be 17 inches plus an additional 2.9 inches, which is 1/2 ball width on each side. That works out to +/- 0.83 feet, not 0.96 feet. I think you guys were counting a full ball width on each side. So, Maddux is getting an even bigger break than we thought!

As for the vertical strike zone, since that is changing with every batter, it seems difficult to me to draw hard conclusions about high/low pitches.

Great point.  The measurement is based on the, presumably, mid-point of the ball.  Therefore, we should be adding 0.12 feet to each side of the strike zone.  So, the strike zone is indeeed +/-0.83 feet.  However, *some* part of the ball has to hit the strike zone (though not necessarily just in the two-dimensional plane) at the front of the plate, but anywhere over the plate.  Someone like Jeff Reardon could look like he has a bigger strike zone if GameDay had tracked him.

The up/down is a little tougher of course, since it is batter-dependent.  I don’t know why though.  Tiny Billy Wagner doesn’t have a different strike zone, and Shaq’s 3-point line isn’t farther away than the regular-sized players.  Not sure why Rickey Henderson should have a different strike zone, other than the ease of reference for the umpire.

re #4 - I meant .121 feet which is 1.45 inches. I was giving that much on the top and bottom of the zone to account for the radius of the baseball (the diameter being 2.9 inches) assuming that the x and y coordinates mark the center of the ball (which is not always the case I’m sure) and allowing for any part of the ball to touch the zone.

Yes, “agreement” meant strikes called strikes and balls called balls and yes, I can break it out further but I’m wondering if I should discount the first n games in a park to account for calibration of the system?

re #5 - I forget now why I went with .96 but I think I gave an inch and a half on either side to account for “the black” and then the radius of the ball. But still, I think we’d have to account for at least an inch on either side since Cory mentioned that the accuracy is supposed to be within an inch.

I’m not sure I understand the point about high and low pitches and the changing strike zones. Each pitch has a top and bottom number for the strike zone and so theoretically every pitch is custom from that perspective. What I’ve noticed is that for the duration of each plate appearance the zone remains the same but that it varies just a little between plate appearances for the same hitter. A quick calculation showed the Brandon Phillips had the smallest average strike zone and Carlos Lee the largest thus far. I also noticed that several pitchers were higher in the list probably both because they’re on average taller and because they may stand straighter at the plate.

Very interesting.  I didn’t realize that the upper/lower part of the strike zone was in the XML file for each batter.

I understand how this makes sense in your Sammy Sosa presentation, but now it looks confusing with Greg Maddux.  After all, since each hitter has his own upper/lower, presenting the pitcher with a fixed box is a bit confusing.

I agree. The box was simply an average of the all the strike zones of all the hitters. I added it mostly to give some perspective and of course inside/outside is the same.

But even with Sammy it’s still an average. The standard deviation on Sammy’s at bats for the top of the zone was 2.6 inches and on the bottom it was 1.9 which seems a little high to me. When you look at it by park, in Chicago and Toronto the top of the zone is on average quite a bit lower:

Park  #   Top   Bot
ANA  39  3.61  1.76  
CHA  40  3.16  1.49
SEA  41  3.43  1.66
TEX 167  3.51  1.57
TOR  63  3.18  1.56