Note: The numbers used in this post do not include last night's start where Martinez was awesome.
Like Alex Reyes now, Carlos Martinez had to spend some time in the bullpen before getting his chance to be a permanent starter. In 2014, he was one of the team’s primary set-up men along with Pat Neshek. The Jason Heyward/Shelby Miller trade before the 2015 season opened up a spot in the rotation and Carlos grabbed it and didn’t let go. He hasn’t been as good this year though. Here’s some selected stats from his 2015 and 2016 seasons:
The biggest difference is probably the huge swing in BABIP, but BABIP is a function of large amounts of variance. The biggest difference you should be concerned about is the drop in strikeouts. First off, let’s look at his Brooksbaseball.net page to get an idea of what his repertoire has looked like this year:
For more context, let's also look at his 2015 numbers:
His pitch mix is very similar, though he’s throwing the slider a bit less. There is no large change in the velocity or movement of any of his pitches, or how often hitters are swinging at them. The only big difference between these charts is how often hitters are whiffing on the change, going from a 43% whiff/swing to 28%.
Since the stuff looks the same, one might assume this has to do with location. Here’s a heatmap of his changeup location in 2015 (left) and in 2016 (right):
Well, that makes a lot of sense. Last year, Carlos buried his changeup under the zone; this year, it’s more in the zone. Now, a changeup down in the zone isn’t horrible, especially when it’s a changeup from El Gallo. However, it definitely makes a lot of sense that it’s going to get more swings and misses when it’s located below the zone.
Really, that might be the whole story. His changeup has experienced the biggest drop in whiff-ability, his stuff is the same, but the changeup is in the zone more often than last year. However, I wondered if Carlos, with his elite stuff, is able to get weaker contact on that stuff. The drop in swings and misses was accompanied by a big drop in BABIP, and while it could easily be random, maybe some of the change is due to a change in strategy.
To see if this was the case, I turned to BaseballSavant.com and their Statcast data on Exit Velocity and Launch Angle. I have used this data before, when looking at Matt Holliday’s batted balls to determine he was getting unlucky on balls in play. I also looked at Kolten Wong’s batted balls to determine that he really did have very weak batted ball authority.
Using some stats I introduced in the Kolten Wong piece, here is how Carlos Martinez has performed in terms of home run Contact Quality:
In case you didn’t read the post linked in the paragraph above this graphic, I’ll give you a rundown of the stats presented here. HRPBB stands for Home Run possible batted balls, which are batted balls with a Launch Angle between 18 and 42, inclusive. The large majority of homers occur on balls in between those angles.
Angle-based takes the average homer per batted ball rate of each HRPBB's angle, to illustrate how well a hitter optimizes the angle of his batted balls towards hitting homers. Velocity is the same, except using velocity of each Home Run possible batted ball. Total takes the specific angle/velocity combination, or vector, into consideration. Actual HR/HRPBB is how the player actually performed. I also track Home-run possible batted balls that are below 90 mph, because those almost never leave the yard.
Carlos is giving up softer contact on homer-possible batted balls, leading to a big reduction in his homer-rate. That appears to mostly take the form of an increase in those batted balls getting hit below 90 MPH. Now let’s look at some stats that consider balls in play:
If you haven’t read the post on Holliday linked above, angle-based xBABIP is created by taking the angle of each batted ball, and comparing it to the average BABIP of all statcast-recorded balls at that angle. Velocity-based xBABIP is done the same way, but just considering a player’s velocity. Total xBABIP takes the specific angle/velocity combination, or vector, of each batted ball. Pop-up% is also included, which is defined by Statcast as a batted ball that is hit with an angle of more than 50 degrees. Nearly all Pop-ups are outs.
On balls in play in 2016, El Gallo has yielded below average contact in terms of angle and velocity. It was a second season in a row of below average velocity-based contact. The result is a total xBABIP well below average.
Now, this isn’t exactly meant to say .276 is Carlos’ true-talent level BABIP. Think of it as peeling one layer of variance off. Instead of taking the actual results of each ball in play, we’re replacing it with the average of balls in play with that specific velocity and angle. Carlos won’t have the exact same assortment of balls in play going forward, so that’s a form of variance we’re not removing from the situation.
There’s also the fact that hitters are more responsible for contact quality than pitchers. FiveThirtyEight.com concluded that five-sixths of Exit Velocity is due to the hitter. However, Carlos has suppressed velocity for two years now, in terms of both balls in play and homer production. Also, as the article alludes, any real ability to suppress contact can represent a significant gain. Plus, pitchers that can suppress velocity multiple years in a row could be more responsible than just one-sixth.
Now let’s look at the details of Carlos’ Contact Quality with a very pretty visual from BaseballSavant.com :
That big spike at 20 degrees isn’t great, as it’s a good angle for both balls in play and homers. At the least, those balls aren’t hit hard so they’re rarely homers. However, 10 to 25 degrees are the Line Drive angles, where balls in play need very little Exit Velocity to be hits.
Most of the homer suppression comes from the low Exit Velocity averages in the homer-possible angles, 18 to 42 degrees. Most the BABIP suppression comes from that too, as well as weakly hit grounders under -10 degrees, which is a range with little chance of being a hit.
Usually I use this info for hitters. One of the fun things about doing one on a pitcher is that I could break it down based on pitch type. It bears pointing out that BaseballSavant.com calls Carlos’ breaking ball a curve ball. Most everyone else refers to it as a slider though, including the writers and community here, so I’ll refer to it as a slider to avoid any confusion. Here is his 2016 Contact Quality numbers broken down by pitch, first homer production:
And then balls in play:
That fourseamer has really done well in terms of contact. He has a slightly above-average angle-based xHR/HRPBB, but the velocity number brings down the total. It also induces worse than average contact in terms of balls in play, by both angle and velocity. The angle-based xBABIP looks largely influenced by being the main pitch for Carlos to generate pop-ups, which makes sense. It’s the only pitch he’s trying to throw up in or out of the zone a lot.
The sinker is the only pitch with an above-average chance of being a home run when hit at a home-run angle, but it’s also the toughest pitch for hitters to get into a home-run possible angle. I was hoping the sinker would induce a good angle-based xBABIP score, as it’s a pitch where Carlos is trying to get the hitter to pound the ball into the ground. That is indeed the case.
Surprisingly, Carlos’ change induces weaker than average home-run power. You’d think that when a hitter hits a changeup in the air, that it’s probably a really bad thing for the pitcher. Carlos has managed contact on the pitch well though.
The slider is the pitch opposing hitters hit at a home-run possible angle most often off of Carlos, but they have been hit at incredibly low velocities. The same is true of balls in play, though there a strong angle-based xBABIP brings it up to a pitch that has allowed an above-average level of contact on balls in play.
Let’s put it all together now. We’ll take the frequency and whiff/swing numbers from Brooks, as well as the total xBABIP and xHR/HRHRPBB numbers we’ve calculated today. All we’ll add this time is a row for xwOBA on-contact. That’s calculated essentially in the same manner as the xBABIP and xHR methods, but calculating the average wOBA of each velocity/angle combination. Here’s the final chart:
League average wOBA on-contact is .367, so overall, each of Carlos’ four pitches have induced a below-average rate of contact. That seems pretty impressive. We don’t know how likely it is to continue going forward, but at the very least we can say he has limited hitters’ Contact Quality thus far, and has done so with each pitch.
The sinker has the lowest whiff/swing rate, as well as the lowest xwOBA on-contact, which speaks to the contact-management nature of the pitch. Conversely, the slider creates the most swings and misses, but the highest quality of contact when hitters do make contact. The fourseamer and changeup are in between those extremes, with the changeup grading out better, likely due to throwing it less and playing off the fastball.
I like to make definitive conclusions in my posts. However, at this point, we can’t really say for sure Carlos Martinez is actually good at limiting contact. He definitely can't keep it up to the degree he's shown so far. It’s certainly something to watch going forward though. Perhaps, a segment of pitches that were swung at missed at last year are instead generating weak contact. With the strikeouts down, perhaps Carlos is making up for some of the gap by inducing weaker contact. We’ll see if he can keep it up.