Framing Comparisons and 50-50 Calls

If you've followed the recent news in baseball, you'll know that the Gold Gloves were awarded to players with a great defensive record in the past season. This year, the Gold Glove for catchers were awarded to Yadier Molina from the National League and Salvador Perez from the American League. This was Molina's ninth Gold Glove, while it was Perez's fifth, but what makes these two catchers better than the other 28 starting catchers in the MLB? One could say that they are some of the best pitch framers in the game. But let's compare the two side by side to see which is the best.

First we need to create a way to find out where a ball would be called a ball or strike by an umpire. This leads to what I call "50-50 charts", where using functions in RStudio, we can map the behavior of umpires and their tendencies in calling balls and strikes across the plate. Using data from the 2018 season, we first sort it to only have pitches that were decided by the umpire; either a called strike or ball. Having swings in the data would skew it, because it would also show pitches that were strikes but an umpire might not have called. After this, we sort that data into one of four different groups. These groups are right-handed pitcher versus right-handed hitter, right-handed pitcher versus left-handed hitter, left-handed pitcher versus right-handed hitter, and left-handed pitcher versus left-handed hitter. We sort them because later, you'll notice a large difference between each of the four umpire zones. After sorting them, we can finally start model-building. Using the "loess" function in RStudio, we can create a relationship between a umpire call and the x-z coordinate plane right in front of home plate. Using this relationship we can create a way to predict the strike called probability for each of the four groups.

As you can see, there is tilt in the calls for RHP versus LHH (2) and LHP versus RHH (3), almost exactly opposite to each other. This tilt is also visible in RHP versus RHH (1) and LHP versus LHH (4), but less extreme. Now, we can use these graphs to visual the catching data. Let's look at Molina first.

Yadier Molina

RHP vs. RHH

The first graph we'll look at is the calls Molina was able to get with a strike probability of less than 50%, using the relationship we built from the "loess" function. Of all the balls thrown with a less that 50% strike probability, Molina was able to get a strike call 3.3% of the time. 72.2% were fastballs, cutters, or two seams. As you can see on the graph, a lot of these strikes are being called on the outside part of the plate to RHH. 11.1% of these pitches were called third strikes.

Now, let's look at the exact opposite. This next graph is balls with a strike probability above 50%. Of all balls thrown with greater than 50% strike probability, Molina gave up a ball 15.3% of the time. Again, a lot are on the outside part of the plate, a majority of which are sliders, curveballs, and two-seams. Now, if you look deeper, you can see that 22 of the pitches shown have a strike probability greater than 75%. Of these 22, nine were sliders. These pitches are mostly lower in the zone, around 6 o'clock. These misses could be because the pitcher missed his spot or the umpire was maybe in the wrong position to make a call.

RHP vs. LHH

This time we'll look at the situation of a right-handed pitcher versus left-handed hitter. This is where you can really see the difference between the different pitcher-hitter states. Of the pitches, 57.4% were fastballs, 17.4% were curveballs, and 9% were sliders. For this situation, you see a lot of the strikes are being called on the outside part of the plate for the hitter. You see a cluster of fastballs at 8 o'clock and 10 o'clock, with the rest of the pitches spread out pretty evenly. Comparing the first graph with this one, we see that it has the same idea, where the outside pitches are being framed better than inside pitches, resulting in more strikes outside.

Again, we can compare how he and how many balls he gave up. Of all pitches with a strike probability above 50%, 18.3% of these pitches were balls. A lot of these pitches are on the inside part of the plate to the hitters, different from right-handed pitcher versus right-handed hitter. Of the pitches, 60.1% were fastballs, 14.6% were sliders, and 12.6% were curveballs. The fastballs were around 5 to 6 o'clock, while most of the sliders and curves are between 3 to 4 o'clock. Like right-handed pitchers versus right-handed hitters, 22 pitches were inside the 75% strike probability range. of these pitches, 59% were fastballs. A question I have here is what happened on the two pitches near the center of the 75% circle. How were two pitches that were right in the middle of the strike zone called balls, and who was the umpire who called these two pitches (probably Joe West or Angel Hernandez).

LHP vs. RHH

Now, we're going to switch handedness for pitchers, going from right-handed pitchers to left. We see here, there aren't a lot of pitches in this category. Of all the pitches outside the 50-50 zone, 5.2% were strikes. Almost all of the pitches are on the top part of the zone, with barely any on the sides of the plate. Of the pitches, 60% are fastballs.

Like the other two situations, we can compare and see where he lost strikes at. Of all the pitches inside 50% probability, 19.9% were called balls. Of the pitches, 53.8% were fastballs. In this situation, we see a lot of balls low, but we do see small clusters at 3, 9, and 12 o'clock. A lot of the offspeed pitches low, the exact opposite of where Molina got strikes.

LHP vs. LHH

The last situation we'll look at for Molina is left-handed pitchers versus left-handed hitters. Like the other left-handed pitcher situation, there aren't a lot of pitches to make conclusions off of. Of all pitches with less than 50% strike probability, 4.5% were strikes. As I already said, there isn't a lot to say about this graph, the pitches are spread out and their isn't really a pattern.

Like the other three situations, this last graph is balls with a strike probability of greater that 50%. 57.7% were fastballs, with two fastballs and one knuckle-curveball in the 75% strike zone. Like the strike graph, there's no real pattern, making it hard to draw conclusions from the graph.

Now let's look at Salvador Perez and his framing graphs.

Salvador Perez

RHP vs. RHH

Like Molina, we'll first look at Salvador's ability to frame right-handed pitchers. The first situation we'll look at is right-handed pitchers versus right-handed hitters. On the left, we look at strikes outside the zone, which were 3.6% of the pitches outside of the strike zone. Of these pitches, 64.3% were fastball and 17.8% were sliders. As you can see, most of the pitches are low out of the zone. However, you can see a lot of the slider are at 3-4 o'clock, outside to a right-handed hitter. This could mean that Salvador is good at bring sliders back into the zone.

Now we look at the pitches Perez missed. Of all pitches in the zone, 19.4% were balls. Of the pitches, 56.4% were fastballs and 29.2% were sliders with fastballs concentrated at 12, 6,and 9 o'clock. There's a pretty even spread of pitches, but the pitches also are very close to the 50% line, meaning that they are close to being balls. We can also see 4 pitches inside the 75% probability range, but a few of them are close to the line, meaning they were closer to 75% than 80%.

RHP vs. LHH

Here, we exclusively see pitches on the outside of the plate being called strikes. We see almost none on the inside part of the plate for the hitter. This matches Molina's graph, but it's more extreme with the shift of balls and strikes. Of all the balls outside the zone, 7.1% were strikes. Of the pitches, 43.9% were fastballs, 16.7% were change ups, and 13.6% were sliders. Like I said, a lot of pitches outside, with fastballs being on the bottom half of the zone, with sliders being around 9 o'clock.

Again, we see a large number of strikes being called balls. 19.7% of pitchers were called balls but were in the strike zone, with 52.5% being fastballs and 20.7% being sliders. A lot of breaking pitches were at the bottom of the zone, being low and down the middle. Fastballs have a cluster that were inside to the hitter, which is the opposite of where he was getting balls called strikes. We have 17 pitches in the 75% strike probability range, with six curveballs, four sliders and four fastballs. If we compare where Perez got strikes called to where he missed balls, we can that where he got the most called strikes is also empty for where balls could have been called. It looks like Salvador does reasonably well framing pitches low and on his gloveside, versus low and across his body.

LHP vs. RHH

On the left we have balls called strikes, which were 4.5% of the balls outside the zone, and on the right we have strikes called balls, which were 22.2% of the strikes thrown in the zone. Of the strikes outside the zone, they were 48.7% fastballs and 12.8% sliders, where the pitches were outside to the hitter. There's a cluster of curveballs and sliders at 3 - 4 o'clock, which was the same as both Molina and when Perez had right-handed hitters up. For the strikes called balls, 72.7% were fastballs and 11.1% were curveballs. A lot of the pitches were inside to the hitters, and most of these pitches were fastballs.

LHP vs. LHH

The final situation is left-handed pitchers versus left-handed hitters. As always, we first look at balls called strikes. 2.0% of the balls outside the zone were called strikes, with 33.3% being fastball. Unlike Molina, there is a bit of a pattern here, with most of these pitches being low and outside to the hitter, on the glove side of Perez.

Now, we look at where Perez missed strikes. Of all pitches with a strike probability above 50%, 17.1% were called balls, with 67.6% being fastballs and 18.9% being sliders. If you look, you can see most of the fastballs are exclusively inside to the hitter, with some sliders low and down the middle. It's hard to make conclusions base on these two graphs because of the lack of information and lack of pitches in the data set.

Conclusion

If we compare the percentages, Perez was better at getting his right-handed pitchers strikes that were balls (3.6% for Perez versus 3.3% for Molina in R-R situations and 7.1% versus 5.8% for R-L situations), but that is the only situation Perez was better at. Molina is a better catching for left-handed pitchers, and also saved his right-handed pitchers from throwing more balls. Another idea that we saw was that umpires are willing to give a close pitch that's outside to the hitter as a strike, while a close pitch that's inside to a hitter as a ball. The final note from me is that we see a lot of sliders being called strikes when they were balls and called balls when they were strikes. This had lead me to my next project idea, where I'll look at sliders going from in the zone to out of it and the exact opposite.

Quick note: I'll be going to Las Vegas this weekend for the Winter Meetings, and then when I get back I'll have finals, so I won't be posting on here for two weeks or so. Thanks for reading, and I appreciate any feedback messaged to me on Twitter. Thank you!

References

All data was acquired from Baseball Savant

Graphs were made using RStudio and the "latticeExtra" package

Deepayan Sarkar and Felix Andrews (2016).

latticeExtra: Extra Graphical Utilities

Based on Lattice. R package version 0.6-28.

https://CRAN.R-project.org/package=latticeExtra

Link to github and my some of my code is here: https://github.com/wgentz/MLBCatching/blob/master/RCode

Google Sites

Report abuse