March 07, 2011
Ervin Santana's Inside Pitch
By Dave Allen

A couple weeks ago Mike Fast posted two graphs showing the average plate location of each pitcher from 2007-2010 to LHBs and RHBs. I thought it was a very cool idea as it offers some insights into how pitchers attack the zone differently. It is particularly interesting to see the outlier pitchers: Derek Lowe has the lowest (ha) pitches, and Livan Hernandez pitches the farthest away to both left-handed and right-handed batters. And without being presented the data I would never have guessed that Ervin Santana throws the farthest inside to left-handed batters of any pitcher. With most pitchers throwing extremely away to lefties I wanted to see what was going on with Santana.

*As a quick side note, I think that these are the raw values straight from gameday and not Mike's corrected values (Mike can correct me if I am wrong). But recently both Mike and Max Marchi described park correction systems to the plate locations they developed. Though take slightly different approaches they seem to get similar results, which is reassuring. Also they both come to the conclusion that for most parks on most days the data are probably off by less than an inch or two, also reassuring. But still I think the correction process is an important one, and I hope to give it some thought soon.*

Turning back to Santana, although he occasionally throws a change up he is largely a two-pitch pitcher throwing his fastball 59% of the time and his slider 36% of the time (he stopped throwing a curve in 2007) . This is the case even to left-hand batters, to whom he still throws his change up under 10% of the time and throws his slider over 30% of the time. For a right-handed starter is this is a very high percentage, maybe one of the highest in the league, for slider use to left-handed batters.

I think this is a big part of the reason he throws so far inside to left-handed batters. Sliders have glove-side movement compared to fastballs, so from right-handed pitchers they move away to right-hand batters and in to left-handed batters. Thus from a right-handed pitcher they often end up inside to left-handed batters. Since Santana throws more sliders to left-handed batters than the almost any right-handed pitcher he will tend to throw more inside pitches. And his sliders are even more inside than the average right-handed pitcher's:
So Santana is throwing sliders to left-handed batters much more than other right-handed batters do and throwing them even farther inside than others do. Most right-handed pitchers throw fastballs and change ups to left-handed batters and when they do they throw them outside. Santana throws his share of fastballs to left-handed batters, but when he does he throws them slightly more inside than average:
When you add up all these parts — fastball not as extremely outside as other RHPs, lots of inside sliders rather than outside change ups — you get Santana's extreme inside pitches to left-handed batters. But does it work? What are Santana's numbers against LHBs like?

Santana has averaged a 4.7 FIP against lefties versus a 3.94 FIP against righties — a pretty reasonable platoon split and generally an okay performance against lefties. Also his slider is better against lefties than his change is, getting more whiffs (17% versus 10%), more out-of-zone swings (35% versus 24%), fewer in-zone swings (53% versus 72%), and a lower slugging on balls in play (.583 versus .626). So given what he has to work with it looks like Santana is right to go with his inside slider against LHBs so much over his change. And as a result of this he is the only pitcher whose pitches on average end up on the inside half of the plate to left-handed batters.


Interesting article. It helps prove that pitchers have a great deal of influence on how a batter makes contact. Santana's pitch breakdown proves pitchers have a great deal of influence on how a ball is put into play and, hence, fielding.


Out of interest how to you do these graphs and the one which show GB% varying by say pz? I know you fit a LOESS function but that fits a linear regression model to what is a binomial outcome.

Do you create bins first for the data so you can do the frequency curvens and the GB% by pitch height or is there something else you do?

Hi Alex,

LOESS fits many polynomial functions not a linear one. But yeah I just fit that to the binomial data, and interpret the output as the probability of a grounder.

Hope that helps.