This is Nate Freiman’s third post as part of his August residency. Nate is a former MLB first baseman. He also played for Team Israel in the 2017 World Baseball Classic and spent time in the Atlantic and Mexican Leagues. He can be found on Twitter @natefreiman. His wife Amanda routinely beats him at golf. To read work by earlier residents, click here.
One of my favorite people in baseball is Tom Tornincasa. He was my hitting coach in the Double-A Texas League in 2012. Apart from being a great coach, he kept the clubhouse loose. Ask anyone who played for him; they’ll know what I mean.
At about 6:50, we’d be stretching on the foul line, and he’d walk out with his notebook.
That was the start of our advance scouting meeting.
“Ninety to ninety-four, slider, changeup. Sixty percent fastball, thirty percent slider.”
Dan Straily led the minor leagues in strikeouts that year, spotting his fastball to both sides of the plate and mixing in an almost unhittable slider — unhittable in that it was un-layoff-able — that he’d throw in any count. He was in the big leagues that September.
“One more thing. He sucks.”
I drove in 105 runs in Double-A that year. To paraphrase Hemingway, I do not think that you are very much impressed by that as a baseball accomplishment, but it meant a lot to me.
I drove myself in 24 times, which means I drove somebody else in on 81 occasions. The vast majority of those runners were in scoring position.
The biggest challenge for a hitter when batting with runners in scoring position is to keep his usual approach. Pitchers take advantage of aggressiveness in these situations, trying to get chases out of the zone with a view to inducing the whiffs or weak contact that such swings create.
It is useful, in such moments, to have a sense of what a pitcher is likely to throw. While the scouting reports from my playing days were helpful in a general sense, they typically never included data beyond a pitcher’s overall pitch mix. Sometimes we’d hear that a pitcher relied more on offspeed stuff with runners on, but it was never more than anecdotal.
Times have changed, however. Pitch-type data by count is now available to the public at Baseball Savant. A pitcher’s tendencies are available for everyone to see.
With that in mind, I wanted to examine the kind of data for the 2018 season that was unavailable to us back in my Texas League days — specifically, by looking at MLB pitchers this season and seeing how their tendencies change in different situations. It makes a difference for that hitter in the box.
I was originally just going to analyze each situation and count the percent of fastball and offspeed pitches thrown by each pitcher to see if any pitchers had statistically significant differences. Instead, I got some crucial help from Kyle Burris, a PhD student at Duke with front-office experience. He introduced me to generalized linear mixed models (they are used at Baseball Prospectus) and gave me a crash course in how to implement them, not to mention a peer review.
This kind of model allows one to make a binary prediction in a certain situation that controls for the random effects of different matchups, counts, and outs. It’s a robust way of finding how much the situation dictates the pitch.
I took 2018 pitch data from Baseball Savant and filtered for pitchers who had thrown at least 2000 pitches. This left 78 pitchers with a total of about 173,000 pitches.
I broke the resulting sample into two groups — simply either fastball or offspeed. For the purposes of this post, I included cutters with the other fastballs. The “split-finger fastball” (named for how it is thrown, not for its velocity or movement) was relegated to the offspeed bin.
The first situation I assessed was a generic double-play situation, with a runner on first and less than two outs. Do pitchers start throwing more offspeed pitches to try to get rollovers, or do they pump sinkers?
The fixed-effects coefficient for this model (double-play situation) turned out to be insignificant. In other words, it didn’t tell us anything about pitchers’ tendencies as a whole when there is a double play in order. Certain pitchers do mix up their stuff, though. Here are 10 guys whose predicted odds really do change.
There’s clearly a lot of variation here. It’s not surprising to find Lance McCullers McCullersing in these situations. His curveball is one of the best.
How about pitchers who throw more fastballs in this spot?
In 2018, Kyle Gibson has gone to the fastball more than any other pitcher in a double-play situation. Unsurprisingly, he’s thrown his sinker more than his four-seamer in every one of his major-league seasons.
How about with a runner on third? My experience was that we’d see a lot more offspeed stuff in that situation. With a runner on first, a walk advances everyone. With a runner on third, there’s usually room on the bases, so pitchers aren’t afraid to pitch out of the zone. And lots of times, a walk is better than leaving a pitch over the middle for a crooked number.
Runner on Third
A lot of these names are the same. The percentages also seem a little more extreme across the board. In fact, the fixed-effect coefficient for having a runner on third was statistically significant. Pitchers as a whole have an observable tendency to mix in more offspeed with runners on third.
Are there any pitchers more likely to throw fastballs with runners on third? Actually, yes.
Runner on Third
In addition to throwing his four-seamer higher and his curveball lower since joining Oakland, Mike Fiers has also reached 20% in terms of sinker usage in two of his four starts, a mark he’d hit only three times previously this year.
There is one more scenario I wanted to examine. This one’s the toughest for a hitter because there’s such a tendency to get aggressive. This is when there are runners on and a base open. These days, full counts with a base open are turning into an almost automatic offspeed situation.
In this case, the group of pitchers with a dramatic decline in fastball percentage is basically the same as with a runner on third. But there are some new faces in the group that adjusts least.
Runners On and Base Open
Even the leaders by this measure still throw fewer fastballs. The fixed-effects coefficient for a base open was the most significant, with a p value of zero. The takeaway: be patient in pitchers’ counts when there’s a base open.
Is there something about these pitchers that could predict fastball use? Instead of fastball rate, I looked at the type of fastball. Four-seamers are geared more towards swings-and-misses and balls in the air, while two-seamers and sinkers get contact and ground balls. With a runner on third, I’d want the strikeout if I were the pitcher. In a double-play situation, I might prefer a ground ball. At least that’s the theory.
For each pitcher, I found the percent of fastballs that were four-seamers and ran the correlations.
|Situation||Pearson correlation with FF %||Spearman correlation with FF%|
|Runner on 3rd||0.144||-0.04|
It’s a pretty weak relationship, but there are slight correlations, which suggests two things. First: pitchers who rely the most on the four-seamer might be more likely to switch to an offspeed pitch in a double-play situation. Second: sinker guys might be more likely to go offspeed with a runner on third.
We can plot the results of all three situations together.
In the boxplot above, the stripe inside each box is the median change. The box represents the middle 50% of the data. A wider box means more dispersion. In a double-play situation, the median is basically zero, but the box is the widest of the three, meaning that there is the most variation. With a runner on third, the data is shifted slightly downward. There is still plenty of variation, but on average, pitchers throw fewer fastballs. With a base open, there is a noticeable difference. Not only does the league throw fewer fastballs, but there is less dispersion.
When a hitter’s in the box, knowing that a pitcher throws 56% fastballs in a particular spot vs. 65% won’t lead to dramatically different results. And anyway, at-bats are path dependent. The previous pitch is going to dictate the next pitch. These aren’t random trials.
That being said, tendencies can and should be used to inform a hitter’s approach. Even a marginal advantage can be a difference-maker. Eventually, it will come down to having a plan, sticking to it, and getting on time into a good hitting position. After all, baseball is hard.
Huge thanks to Kyle Burris for the stats help. Apologies to Kyle Burris for the use of the term “peer.”
2013-2014 Oakland A’s