Location, location, location
I noticed that last week that Jim Albert wrote a piece which included some incredible heatmaps that plot where pitchers spot pitches based on the pitch-count. Albert’s CalledStrike package reduces the options to fastball and non-fastball (so, off-speed) and doesn’t yet allow for a look at individual off-speed pitches. I’ll leave that to someone much, much more competent than me to figure that out.
Suffice it to say, I got hooked by how beautiful those density plots looked and I decided to tool around with the data and apply it to the Yankees’ pitching staff. Why the Yankees? First, my friend is a Yankees fan and I promised I’d write something Yankees related. Second, Yankees fans have been clamouring for more starting pitching, particularly with the ever-injured James Paxton (whom I love), and the (possible?) departure of Masahiro Tanaka. Third, there was considerable variation in pitcher success for the Yanks last year. Albeit a shortened season, Gerrit Cole, Masahiro Tanaka, and Jordan Montgomery accrued 1.4, 0.8, and 0.9 WAR respectively. I’m focusing on these three guys, each making at least 10 starts. So Happ has been excluded, as has Paxton.
The crux of Albert’s first post was a look at the change in pitch-counts and probability that an AB ended at a given pitch-count over the past 20 years. It also segued into Maddux’s recent spiel about never wasting an 0–2 pitch. Albert presents interesting data and I don’t deviate too much from it. In the context of this analysis, I’m mostly interested in variation between the three Yankees across pitch-counts with a focus on 2-stike counts.
Cole threw 1,203 pitches in 2020, 52% of them fastballs, 24% sliders, 17% knuckle-curves, and 5% changeups. He threw a first pitch strike 55% of the time, a first pitch ball 36% of the time, and 8% of first-pitches were hit in-play. Tanaka, in contrast, threw 769 pitches, 37% sliders, 24% splitters, 24% fastballs, 7% sinkers, and 5% curveballs. He threw a first-pitch 50% of the time, a first-pitch ball 36% of the time, and induced in-play contact 13% of the time. Montgomery, similarly to Tanaka, threw 747 pitches, 26% sinkers, 25% changeups, 22% curves, 19% FB, 6% cutter. Montgomery started ahead in the count 49% of the time, fell behind 42% of the time, and induced contact 8% of the time. Interestingly, this is fully something I wasn’t expecting, Cole faced only 6 3–0 counts all season, Tanaka 3, and Montgomery 4.
Two-strike counts are not, theoretically, all equal. Perceived logic dictates that pitchers can waste pitches on 0–2 counts to induce a batter to swing at a pitch they normally wouldn’t. Yet the preponderance of whiffs come in 0–1 counts for both Cole and Tanaka (roughly 20% of whiffs for each pitcher occur at 0–1) and on 1–0 counts for Montgomery (17% of whiffs). 7.6%, 10.5%, and 7.3% of each pitcher’s total whiffs come off 0–2 counts for Cole, Tanaka, and Montgomery respectively.
So Cole, Tanaka, and Montgomery aren’t necessarily inducing more whiffs on 0–2 counts. The plots below use Jim Albert’s plotting function I briefly described above. The figures are from the catcher’s perspective.
The plots demonstrate the preponderance of pitch-location based on RHB or LHB batters. Note that Cole is likely to bust RHB in at the hands on 0–2 pitches with his fastball but increasingly works outside as the number of balls increase. He generally works away to lefties with the exception of the more varied 1–2 count.
Across off-speed pitches, Cole seemingly has a preferred location — down and away to righties and low and inside to lefties. It seems, however, that Cole tries to induce swings on 0–2 counts and slowly works his way back into the strike zone.
Tanaka’s a bit more varied than Cole in that it appears that there’s more variation based on RHB/LBH. In 0–2 counts, Tanaka tries to get hitters to chase a high fastball, regardless of stance. He’s more varied in 1–2 counts across hitter — preferring middle- middle-out to RHB and high-and-inside and below the zone to LHB. 2–2 counts appear to be symmetric across handedness and so is, generally, 3–2 counts.
Tanaka clearly has a preferred spot location for his off-speed pitches, however, preferring to throw down and away to RHB and low and inside to LHB, edging into the strike zone as the ball count increases.
Montgomery offers more variation regarding his fastball and batter handedness. He’s clearly trying to induce chasing pitches on 0–2 and 1–2 counts with his fastballs to both RHB and LHB. 2–2 counts for both batters are more mixed — it seems he has a preference for strike-zone fastballs for RHB but prefers the outside corner for lefties. When facing 3–2 counts, Montgomery clearly prefer high and outside to RHB but varies location to LHB, ranging for top of the zone to low and away.
Like the other guys, Montgomery wants hitters to chase his curveball on 0–2 counts, slowly reeling it into the zone with each additional ball. The propensity to get lefties to chase on two-strike counts is evident — but at 2–2, he switches location towards the inside corner and varies it considerably on 3–2 counts.
On the face of it, at least with this small sample of both pitches and pitchers, it does seem like these guys do try to induce swinging-strikes on 2-strike counts, moving closer and closer to the strike zone with each passing ball.
It doesn’t point to anything causal but demonstrates interesting patterns nevertheless. Obviously, whiff rates, contact, the things Albert points to, need to be incorporated to make this have more teeth. Swing probability needs to be worked in here. It’s possible that batters likely know a wasted pitch is coming as well. There’s a bunch of moving parts here.
You can get absolutely lost in this data for hours. It’s a ton of fun.
You can find Albert’s post here: