Most pitchers in a slump give you a shrug and a cliché. Shota Imanaga gives you a diagnosis. Last September, with his season ending in a cluster of home runs, he pointed at the homers coming on his fastball and told MLB.com, "I think I need to continue to make adjustments." In early June, four starts into the ugliest stretch of his MLB career, he named a new culprit: the Times Through the Order: How many times a batter has faced the same pitcher within a game..
Here is the interesting part: the data agreed with him both times, on the day he said it. And that is as far as this piece will take either statement, because a comment to the press is a snapshot of the problem in front of a pitcher, not a prediction about where it goes next. Pitching is a game of mutual adjustment. He adjusts, hitters adjust, and a problem that moves afterward is the natural state of affairs, not a verdict on the man who described it. The data even has his side of the exchange on file: the four-seamer he named in September is a smaller share of his arsenal than it has ever been. So instead of grading the quotes, this piece tracks the address: where the problem has lived each year, why the four-start collapse that ended June 4 is a location-and-approach slip multiplied by fly-ball luck of a kind no pitcher's stretch sustains, and the one constant underneath it all, a trait he has never mentioned, which cost him nothing as a rookie and is now collecting.
That is the whole argument. Here it is, point by point.
The collapse is real, and it has a precedent
Through May 13, Imanaga's 2026 looked like a vintage season: nine starts, a 2.32 ERA across 54 1/3 innings, five home runs allowed. And that surface was honest. He held hitters to a .250 wOBA: An overall batting metric that assigns a weight of value to each type of reaching base. Higher is better. in that stretch (wOBA is the all-in damage measure, scaled like on-base percentage), and the contact quality behind it, a .273 expected wOBA (xwOBA (batted balls): A metric that estimates the 'batting results that should originally be expected' from the quality of the batted balls struck by the bat (on contact only)., what Statcast says the contact and the strikeouts should have produced), says the run prevention was earned, not borrowed. Clean, not lucky.
Then came the four starts everyone is asking about: May 18 through June 4, a 10.80 ERA over 21 2/3 innings with 12 home runs, capped by a career-worst four-homer night on June 4.
This piece compares those two stretches constantly, so let's pin the frame now. Call the first nine starts period A and the four from May 18 through June 4 period B. The boundary is the one the public conversation drew at the time, his last four starts, which means it was drawn after the fact, at the observed break in results. Cut a season at its worst seam and the bad side is guaranteed to look extreme. So every A-versus-B contrast below describes the stretch everyone is asking about; none of it tests a hypothesis, and every period-B rate is the kind that regresses. Four starts is a tiny window. Keep that in your pocket for the rest of the piece.
It is also not the first burn. His 2025 ended with a six-start homer cluster of its own: from August 28 on, a 5.97 ERA with 12 home runs in 34 2/3 innings. The symmetry should not be oversold; 2025 was not spotless before its collapse, with mid-season three-homer blips of its own. But the shape, a good run of starts and then a concentrated homer burn, has now happened in back-to-back seasons.
29 starts, 25 starts, 13 starts
HR allowed and runs allowed, start by start, 2024-2026
2024 (29 starts)
2025 (25 starts)
2026 (13 starts, partial)
Bars show runs allowed, not earned runs; the ERA figures in the annotations are the official earned-run numbers.
Period A = starts 1-9 of 2026 (Mar 29 to May 13); period B = the four starts from May 18 to June 4 (starts 10-13). The split is post-hoc, cut at the observed break, so period-B rates read extreme by construction.
2026 is a partial season: through games of June 8, 13 starts.
Regular season only; in-house data.
2024, when the same act was nearly free
What makes the bills of 2025 and 2026 news rather than fate is that the act once worked almost without cost. As an MLB rookie in 2024, Imanaga threw 173 1/3 innings with a 2.91 ERA, tenth among the 120 starters who reached 100 innings (16th of 213 in our broader starter pool; ranks here use the house pool, pitchers with at least five starts and majority-starter usage, not the official qualifier). He held hitters to a .293 wOBA against a league starting-pitcher mark of .322, over 694 plate appearances, and the expected version agreed (.289 xwOBA, 24th of 213). He struck out 25.1% of the hitters he faced and walked 4.0%.
And the trait this piece keeps returning to was already fully formed. Hitters were already meeting his pitches well under the ball's center, near the top of the league (the metric is introduced properly below), and his fly-ball rate was already top-decile among starters. The fly balls just were not leaving: his homer-per-fly-ball rate, 0.149, sat below the pool average. That is what "nearly free" means here: extreme fly-ball volume, ordinary homer conversion, elite results.
The arc since: 2.91, then 3.73 in 2025 (mid-pack, 77th of 222), then 4.74 through 13 starts of 2026. The middle year is messier than its ERA. His wOBA allowed stayed genuinely good in 2025 (.296) even as the ERA slipped, while his xwOBA (.313) ran worse than his results. Those numbers describe a tangled season; they do not say which part was him, and we will not pretend otherwise.
What never changes: under the ball, in the air
Contact offset allowed and fly-ball rate by season, vs the qualified starter pools
contact offset allowed (inches; + = bat under the ball's center)
0.411 → 0.594 → 0.407 in; pool means sit around 0.26-0.28
2025 was the spike year (7th of 187 qualified, 96th percentile); persistent, but not flat
fly-ball rate (%)
top of the starter pool every year (93rd / 99th / 93rd percentile)
Contact offset is an in-house model estimate, not a Statcast measurement. The 2026 row rests on exactly 200 batted balls, the qualifying minimum.
2026 is a partial season: through games of June 8, 13 starts.
Regular season only; in-house data.
Diagnosis 1, September 2025: the four-seamer. Right when he said it.
When Imanaga told MLB.com he needed to keep making adjustments, the context was the home runs coming on his fastball (the article counts 22); our own data sharpens the address to the four-seam specifically, and agrees on the day he said it: 22 of his 29 home runs allowed to that point had come on the pitch. That figure is a snapshot, he allowed two more homers on September 25 and the season closed with 24 of 31 on the four-seam, but the snapshot was accurate.
It was not just a counting quirk of throwing the pitch a lot, either. In 2025 the four-seamer was 48.7% of his pitches and 77% of his homers, punished at 0.115 home runs per batted ball against a league four-seam rate of 0.055, roughly 2.1 times league, on 209 batted balls. And 2024 had looked the same: 22 of 27 homers (81%) on the pitch, at 0.083 versus a league 0.053, on 265 batted balls. Per-pitch homer rates on two-hundred-odd batted balls are noisy, but a multiple that large, in both full seasons, is the data agreeing with the pitcher.
Which pitches the home runs came off
Usage share and HR allowed by pitch type, 2024-2026
2024
22 of 27 HR on the four-seam
2025
24 of 31 on the four-seam (22 of 29 at the Sept 19 quote)
2026 (partial)
17 HR spread across four pitch types (FF 6, FS 6, ST 3, SI 2)
2026 per-pitch cells are 19-83 batted balls; every per-pitch rate is directional, not settled.
Bars show pitch types at 1%+ usage. In 2024 one HR came on the cutter (0.5% usage, not shown); the 27 in the annotation includes it.
2026 is a partial season: through games of June 8, 13 starts.
Regular season only; in-house data.
...and by 2026 the problem had left the pitch he named
This season the concentration is simply gone. His 17 home runs are spread across four pitch types: six on the four-seamer, six on the splitter, three on the sweeper, two on the sinker. The burn window is splitter-led, five of its 12 homers on the splitter against three on the four-seamer. The mix has moved underneath the story too: through June 8 of this season he has thrown the four-seamer 41.3% of the time, after 51.9% in 2024 and 48.7% in 2025.
And measured against what each pitch type normally gives up, the pitch he named is, through June 8, the least of his problems. His four-seamer is being hit out at 0.076 per batted ball against a league four-seam rate of 0.053, about 1.4 times league; his splitter runs 0.072 against a league splitter rate of 0.032, about 2.3 times league. Those rates sit on 79 and 83 batted balls. The sweeper's ugly 0.158 comes from all of 19 batted balls and should be read as a footnote, not a finding. Every 2026 per-pitch cell here, 19 to 83 batted balls, is directional rather than settled.
None of that is a grade on the September statement. The statement was checked against the data of its moment, and it passed; a snapshot is not a prediction. In a game where he adjusts and hitters adjust right back, a problem that relocates afterward is the natural state of affairs, and the usage numbers above are his side of that exchange, on the record. The question worth the data's time is not whether he was right then. It is where the problem lives now.
Diagnosis 2, June 2026: the third time through. The burn talking, not the season.
After the June 4 game, Imanaga offered his newest read. Per Nikkan Sports, he said he has to create a pattern for retiring lineups the third time through the order.
Again the surface agrees with him. Through June 8, his season third-trip wOBA is .491, with 8 home runs in 74 plate appearances. That is a grim split. But cut it once more, at the period boundary, and it stops being a season-long story. In period A his third trip allowed a .376 wOBA over 49 plate appearances with two homers: mildly above the league third-trip baseline of .335, and well within what 49 PA of noise can do. Period B's third trip is .716 over 25 plate appearances with six homers, a number that describes a catastrophe without measuring anything stable.
Here is the detail that undoes the theory as a mechanism: in those same four bad starts, he was getting hit immediately. Four home runs in 36 first-trip plate appearances, a homer-per-PA rate around three times the league's first-trip rate. In period A, meanwhile, his first two trips beat the league baselines. The pattern is not "fine until the third trip." It is "fine, then bad on every trip," with the third trip amplifying it. And the honest resolution limit: the cells behind this paragraph run 25 to 36 plate appearances, far below any stability threshold, so the adjudication is the modest one. His third-trip number is the burn talking, not the season.
The third trip, split by period
wOBA allowed per times through the order, period A vs period B, with league SP baselines
season 3rd trip: .491 over 74 PA; decomposed, period A .376 (49 PA), period B .716 (25 PA)
period B was hit from the 1st trip on (4 HR in 36 PA)
the gap between period A's 3rd trip and the league baseline (.335) is within small-sample noise
Period-B cells are 25-36 PA, far below stability; the .716 is descriptive only.
wOBA uses Statcast's own weights; league baselines are computed on the same date spans.
Period A = starts 1-9 of 2026 (Mar 29 to May 13); period B = the four starts from May 18 to June 4 (starts 10-13). The split is post-hoc, cut at the observed break, so period-B rates read extreme by construction.
2026 is a partial season: through games of June 8, 13 starts.
Regular season only; in-house data.
In 2025, the same theory would have been backwards
If the third trip were the structural flaw, 2025 should show it. It shows the opposite. In the season that prompted his first public comment, 18 of his 31 home runs came the first time through the order, nine of them in the first inning and six more in the second. Per plate appearance, his first-trip homer rate ran about two and a half times the league's first-trip rate. And his third trip that year was his best split of the season, a .271 wOBA over 120 plate appearances. (2024, for contrast, drifted upward by trip in the mild, conventional way: .250, .283, .372.)
So, plainly and with respect: each statement matched its moment. In September 2025 the four-seamer really was carrying his homer problem; in early June, the third trip really was where the carnage concentrated. Between those two snapshots the problem itself moved: the 2025 damage came earliest in games, the 2026 damage comes on every trip, and the punished pitch changed underneath the whole conversation. In a game of mutual adjustment that movement is the game's story, not a grade on the speaker. It is also the reason the next question matters more than either quote: if the symptoms keep moving, what stays still?
Damage by times through the order, by season
wOBA allowed per trip vs league starting-pitcher baselines (Statcast weights)
2024
mild, conventional drift: .250 / .283 / .372
2025
18 of 31 HR came the 1st time through (9 in the 1st inning)
the 3rd trip (.271) was his best split of the year
2026 (partial)
wOBA uses Statcast's own weights, not an in-house version.
2026 is a partial season: through games of June 8, 13 starts.
Regular season only; in-house data.
The constant he has never named
Underneath the moving symptoms sits a trait that has never appeared in any of his press-conference diagnoses: every year of his MLB career, hitters meet Imanaga's pitches farther under the ball's center than they meet almost anyone else's.
We measure that with Contact Offset: An estimate of how far below (+) or above (−) the ball's center the bat met it, in inches., an in-house model estimate, not a Statcast measurement: it inverts a simple two-dimensional bat-ball collision model, using bat speed, attack angle, the pitch's trajectory, and the launch angle, to estimate how far below the ball's center the bat arrived, in inches. Positive means under the ball, the kind of contact that becomes backspin and fly balls.
His yearly averages: 0.411 inches in 2024, 0.594 in 2025, 0.407 in 2026, against pool means around 0.26 to 0.28. Every one of those sits well above the typical starter; the 2025 figure was a genuine spike, seventh of 187 qualified pitchers. The trait is persistent, but it is not flat, and that distinction matters in the next section. His fly-ball rate tells the same story from the outcome side, at or very near the top of the starter pool in all three seasons. (The 2026 offset figure rests on exactly 200 batted balls, the qualifying minimum.)
The engine is the four-seamer: 18.0 to 18.3 inches of induced vertical break by year (the "ride" a pitch shows beyond gravity), among the better-riding four-seamers in the game, delivered at 90.8 to 91.8 mph, bottom-decile velocity for a starter's fastball, and thrown upstairs, with 55 to 59% of them arriving in the upper third of the zone or above. Hitters undercut that pitch by +0.86, +0.90, and +0.75 inches by year against a league four-seam norm of +0.60 to +0.63. The splitter, his other main pitch, gets met near dead center.
One thing this trait is not: the cause of the June burn. Hold that thought, because the next section shows the offset actually fell during the four bad starts. The fingerprint is the standing condition of his whole act, the reason everything he gives up travels through the air. It is not the thing that changed in May.
Which pitches get undercut
Contact offset allowed by pitch type, vs the league mean for the same pitch type
four-seam: +0.86 / +0.90 / +0.75 in by year, against a league four-seam norm of +0.60 to +0.63
the splitter sits near zero: met close to dead center
2024
2025
2026 (partial)
In-house model estimate, not a Statcast measurement; 2026 rests on 200 batted balls, the qualifying minimum.
2026 is a partial season: through games of June 8, 13 starts.
Regular season only; in-house data.
What the burn actually was
Start with what did not happen: nothing physical broke. His four-seam velocity went up across the boundary, 91.8 mph in period A to 92.0 in period B; the burn came at his best velocity as a Cub. The pitch's ride barely moved, 18.4 to 18.0 inches, and movement was flat on every pitch in the arsenal.
On the cause side we report raw spin and movement separately, because spin efficiency is not computable from our data and no claim about it appears here. On his two main pitches, 76% of his usage, everything is flat: four-seam spin down 8 rpm, axis moved one degree, ride down 0.4 inches; splitter spin down 6 rpm, axis moved four degrees, vertical break down 0.8 inches. No meaningful axis move anywhere, and the four-seamer kept carrying 2.1 to 2.5 inches more ride than the average left-hander's fastball in both periods. The pitches that left his hand in late May were the pitches that left his hand in April.
The result side makes the same point from the other end: contact on his big pitches did not get more undercut while the homers piled up; it got less. The four-seamer's offset allowed went from +0.82 inches to +0.60. The splitter's went from +0.25 to essentially zero (-0.03), dead-center contact, in the very window in which it surrendered five of the twelve burn homers after allowing one in all of period A. The offsets did not worsen on the big pitches; the homers did.
The stuff held. The home runs did not
Raw spin, induced vertical break, and contact offset allowed by pitch type, period A vs period B
four-seam: spin -8 rpm, IVB -0.4 in; splitter: -6 rpm, -0.8 in (the two main pitches, 76% of usage)
splitter offset allowed +0.25 → -0.03 in (dead-center contact) while giving up 5 of the 12 burn HR (1 in all of period A)
raw spin (rpm)
induced vertical break (in)
contact offset allowed (in)
Raw total spin only; spin efficiency is not computable from our columns and no claim is made about it.
Spin/axis tracking coverage is partial in period B (FF 95 of 134 pitches, FS 97 of 112, SI 20 of 38); velocity and movement use full n.
Per-type period-B offset cells are 5-28 batted balls (curve 1-2, sweeper 5: reference only).
Contact offset is an in-house model estimate, not a Statcast measurement.
Period A = starts 1-9 of 2026 (Mar 29 to May 13); period B = the four starts from May 18 to June 4 (starts 10-13). The split is post-hoc, cut at the observed break, so period-B rates read extreme by construction.
2026 is a partial season: through games of June 8, 13 starts.
Regular season only; in-house data.
So if the stuff held, what moved? Where the pitches went, and what hitters did about it. In period B he caught much more of the plate: his in-zone rate jumped from 43.6% to 51.8%, and the share of pitches in the fat middle rose from 10.1% to 13.9%. At the same time, hitters stopped offering at anything off the plate: his chase rate collapsed from 39.7% to 26.9%, and the whiffs followed, 30.0% down to 23.9%.
Period A to period B: what held, what moved
The stuff (held)
- four-seam velocity 91.8 → 92.0 mph; the burn came at his best velocity as a Cub
- induced vertical break 18.4 → 18.0 in (flat on every pitch type)
Location, hitters' swing decisions, results (moved)
Period B is four starts and 73 batted balls; every B rate is descriptive, not stable, and regression-to-mean applies. xwOBA moved .273 → .391 alongside (not charted: different scale).
Hard-hit = exit velocity 95+ mph.
Period A = starts 1-9 of 2026 (Mar 29 to May 13); period B = the four starts from May 18 to June 4 (starts 10-13). The split is post-hoc, cut at the observed break, so period-B rates read extreme by construction.
2026 is a partial season: through games of June 8, 13 starts.
Regular season only; in-house data.
That chase collapse could mean two very different things: his misses got less tempting, or hitters changed what they swing at. So it was decomposed, and the answer is lopsided: about 90% of the 12.8-point drop comes from hitters swinging less at comparable out-of-zone pitches. Chase fell in every distance band off the plate (on the most tempting misses, within three inches of the edge, it went from 59.1% to 40.5%), in every count state, and on every pitch type with usable volume. His miss quality barely moved: the average miss was 7.55 inches off the edge in period A and 7.40 in period B, the share of close misses was flat, and that component explains only about 9% of the drop. Count leverage is the last roughly 2%, with his first-pitch strike rate essentially unchanged (56.9% to 57.7%). One more piece of texture: hitters' in-zone swing rate actually rose, 67.2% to 69.8%. Swinging at the hittable ones, laying off the rest.
Honesty about the resolution: period B contains 160 out-of-zone pitches in total, so no single band, count, or pitch-type cell clears significance on its own; the 90/9/2 split rests on the reweighting arithmetic plus the fact that every cut points the same way. Read "hitters adjusted" as a description of these four starts, not as a proven scouting-report narrative.
Where the chase went
Chase rate by distance off the edge of the zone, period A vs period B
chase fell in every band (nearest band, 0-3 in off the edge: 59.1% → 40.5%)
his misses barely moved (mean distance off the edge 7.55 → 7.40 in)
roughly 90% of the 12.8-point drop traces to hitters swinging less at comparable out-of-zone pitches
The 0-3 in band is a house band, not Statcast's shadow zone.
Period B has 160 out-of-zone pitches; no single cell is significant alone. The 90/9/2 split rests on the reweighting plus the uniform direction across bands, counts, and pitch types.
A description of these four starts, not a proven scouting-report narrative.
Period A = starts 1-9 of 2026 (Mar 29 to May 13); period B = the four starts from May 18 to June 4 (starts 10-13). The split is post-hoc, cut at the observed break, so period-B rates read extreme by construction.
2026 is a partial season: through games of June 8, 13 starts.
Regular season only; in-house data.
The splitter deserves its own paragraph, because it leads the burn's homer count and its story is location, not chase. The splitters he buried stayed buried: among his out-of-zone splitters, the share below the zone barely moved, 66.8% to 65.6%. But the pitch as a whole lived higher. Its average height rose by more than an inch relative to the bottom of the zone (+1.40 to +2.57 inches), the share of all splitters below the zone fell from 48.0% to 37.5%, and its in-zone share climbed from 36.7% to 45.5%. That is the location face of the five splitter homers, a thread of the home-run story rather than the chase story. There is also a mix wrinkle: a quasi-new sinker, up from 6.4% to 11.4% usage and poorly located (71% in the zone, 13% middle-middle, on all of 38 pitches), accounts for two of the twelve.
More plate plus swings rotated toward the hittable pitches bought exactly the contact you would expect: hard-hit rate up from 38.8% to 46.6%, xwOBA from .273 to .391. But notice what did not move. The barrel-grade rate, the flush, ideal-angle contact, stayed nearly flat (3.6% to 4.1%), and the overall contact offset allowed fell, from +0.49 inches to +0.25. The contact got squarer, not deeper under the ball; this is why the standing fingerprint, which is about undercut contact, is not the burn's cause.
And then the multiplier. In period B, 44.4% of his fly balls left the park, 12 of 27, after 11.4% in period A. No pitcher sustains a 44% homer-per-fly-ball rate; that is a fact about the league's distribution, not a prediction about Imanaga. His actual results in the window, a .470 wOBA allowed, ran about 80 points worse than even the degraded contact quality implies. The homer anatomy agrees: only 4 of the 12 were center-cut mistakes, and 4 came in the lower third of the zone.
The fly balls are constant. Whether they leave is not
Fly-ball rate vs HR per fly ball within the starter pool, by season
2024
a top-shelf fly-ball rate, yet HR per fly ball (14.9%) below the pool average
2025
2026 (partial)
Reference: period B alone ran 44.4% (12 of 27 fly balls), outside the starter pool's distribution entirely.
Pool membership carries a batted-ball floor (150 BBE in 2024-25, 80 in 2026).
2026 is a partial season: through games of June 8, 13 starts.
Regular season only; in-house data.
None of which makes the burn "bad luck," and we are not calling it that. The deterioration is real: more of the zone, no chase, harder contact. The luck is the multiplier sitting on top of it. And every number in this section describes four starts and 73 batted balls, cut after the fact at the observed break. That is the stretch everyone asked about, and it is also exactly the kind of window whose rates regress.
Where the data says the problem lives
Pull it together and the 2026 problem is not living at either address the public conversation has for it.
Not the four-seamer of his September comment: relative to what the league gives up on each pitch, it is the least punished of his main pitches in the data through June 8. Not a structural third-trip flaw: in the four bad starts he was hit on every trip, four first-trip homers in 36 plate appearances, and in 2025 the damage came earliest in games, the opposite pattern.
Where the data does point: a splitter living measurably higher than it used to, with its below-zone share down from 48.0% to 37.5% and five of the twelve burn homers on it; hitters who stopped offering at anything off the plate, chase down 12.8 points with roughly 90% of that drop their own swing decisions on comparable pitches; a poorly located, quasi-new sinker (71% in-zone, 13% middle-middle, on 38 pitches); and the sum of it, 8.2 points more of his pitches in a zone hitters are attacking. What he does with that is his craft, not our call.
Two factual notes are allowed to point forward. A 44% homer-per-fly-ball stretch is a league-outlier rate that does not persist for anyone. And the standing condition underneath, an undercut, airborne profile more extreme than nearly any starter's, is the same one that cost him almost nothing in 2024, when the fly balls stayed in the park and he was a top-ten starter. The act has worked before, at full extremity.
The rest is the game doing what the game does. He has moved: the four-seamer he named in September is down to 41.3% of his pitches, and a quasi-new sinker is in the mix. Hitters have moved: in the four bad starts, the chase he had been living on vanished on their swing decisions, not on his misses. In that exchange a problem does not sit still, and a problem that moved is no verdict on the man who described it; both of his snapshots were accurate when he took them. The constant is the fly balls. The data's contribution is the forwarding address: as of June 8 the problem lives in a splitter sitting higher than it used to, a zone he is catching more of, and hitters who no longer offer off the plate.
Postscript (June 11, 2026)
The numbers in this piece run through the games of June 8, 2026. In his June 10 start, his first after this piece's cutoff, Imanaga faced 19 batters without allowing a home run: seven strikeouts, two walks, two singles.