Where Cal Raleigh Broke — and Where He Didn't

I’ve watched baseball for a long time, but this is the first year I’ve really followed it — basically every game, every night. And the player I follow closest is Cal Raleigh. He’s my favorite: the switch-hitting catcher who hit 60 home runs last year, the guy everybody said did everything. So the season I finally leaned all the way in is the season he has hit .158. What I have actually watched, night after night, is strikeout after strikeout, the occasional home run, and a month on the injured list — a very different thing from watching him hit home run after home run last year. It is heartbreaking, and I have spent more evenings than I’d like to admit trying to work out what I am looking at.

That is what this piece is: an attempt to understand. Everyone is studying him now — the broadcasts, the swing-mechanics breakdowns, the “pitchers finally figured him out” story. I had the pitch-level data sitting on my laptop, so I asked it the questions a fan asks. Did the pitchers change? Did his swing slow down? Is he chasing? Is he late? Is it the injury? Most of the answers surprised me, and one of them embarrassed me.

The whole thing, in one paragraph. His weighted on-base average has gone from .394 last year to .255, against a league that sits around .325. What pitchers throw him accounts for roughly none of that: two independent estimates put the diet’s share of his lost run value at −4% and +6%, with intervals hugging zero. His swing decisions and his miss rate, together, come out at −0.7%. What happens to the ball when he hits it carries about 95% of the loss, in a model-based decomposition whose interval runs past 100% and whose “contact” includes fouls. The one visible change in his swing — where he meets the ball — has been in place since late August 2025, when he was still hitting everything, and it does not explain the loss; the contact quality itself was already at a career-month low in April, so the oblique strain that cost him a month in May did not start either one. What did, the tracking data cannot see.

The season, in one picture

Rolling 50-event windows across 2025 and 2026, one point per game date: the swing footprint on top (contact depth, attack direction) and the ball off the bat below (hard-hit rate, and the “flush” and squared-up shares — balls leaving at 95%+ and 80%+ of the maximum his bat speed and the pitch allow, on our proxy), with the May 13–June 16 IL stint shaded and the sealed next-80 lines marked. Regular season only; 2026 through Aug 15. The swing footprint (top two panels) was already at its 2026 level by September 2025 — the red late-2025 window — while the ball off the bat (bottom three) fell in 2026 and, through Aug 15, no banked window has crossed the sealed next-80 lines: consistent with a centering loss whose cause tracking data cannot identify.

Read it left to right. In April he was seeing a normal diet, making his normal decisions and striking out at his normal rate (23.7%) — and already hitting the ball hard only 34.6% of the time on 78 balls in play, below every 200-pitch month of his career from 2022 to 2025 (previous floor .352). Contact quality broke first. From May the strikeouts climbed to 31–41% a month and the diet drifted modestly toward hard stuff, never leaving his own historical range.

Then the IL. The footprint does not move across it: his contact depth was 33.0 inches before May 13 and 32.9 after June 16, attack direction likewise sits still, and the one swing measure that changed on his return was bat speed, up 0.8 mph. The IL did not start the footprint. That is a fact about the marker, and it carries no medical implication either way.

What has moved since he came back is the middle of his contact. The top end has stayed where it was. His plain squared-up rate (on our proxy) rose from .529 before the IL to .647 after; hard-hit went .308 to .392, and .426 over the last month; his 26 August balls averaged 93.5 mph. But the flush end has not moved — 5 of 102 post-IL balls, a share of .049, exactly the pre-IL figure — his expected value on contact since June 16 (.352) is below every 102-ball stretch of his 2025, roughly one August ball in three has been a pop-up, and his August wOBA is .200. Shuffle the order of his 208 balls in play five thousand times and a hard-hit rise like this shows up by chance roughly one time in seven to ten in a fixed window — and almost always if you let yourself pick the window. The middle moved; the results have not.

The pitchers didn’t do it

Left: the diet’s share of each season-to-season change in Raleigh’s run value (95% intervals). Right: each earlier Raleigh priced pitch-bucket by pitch-bucket on the pitches he has actually seen in 2026, against 2026’s actual figure. Batter run value per 100 pitches; the location buckets use each season’s zone as tracked (the 2026 rows carry the ABS box), and re-run on Raleigh’s fixed 2026 box the 2025→26 diet share is smaller still (−2% [−10%, +6%]). The pitch mix explains almost none of any year-to-year swing, up or down; hand any earlier Cal the 2026 diet and he still produces (+0.2 to +1.8 RV/100 on the verified pitch-group × location bucketing) against the actual 2026 −1.26.

The diet has changed a little, and it is real once measured properly: fastballs .342 → .377, offspeed down to a career-low .179, first-pitch strikes .600 → .654. Priced, it is worth almost nothing. Give the 2025 version of Raleigh, bucket by bucket, the pitches of 2026 and he produces +1.77 runs per 100 pitches, against the +1.59 he actually produced last year. Every earlier Cal back to 2022, put on the 2026 diet, still hits — between +0.2 and +2.3 runs per 100 (only one noisy fine-grained bucketing of the 2023 version dips below zero) — while the actual 2026 figure is −1.26.

A model-based decomposition built independently by the other analysis puts the diet at 6% of the loss, interval [−3%, +19%]; ours says −4% [−16%, +4%]. Our pre-registered line for “the pitchers adjusted” was 25%; neither interval gets near it. The pitchers followed the wound; they did not inflict it.

Now the embarrassing part. In 2026 the public Statcast feed reports the strike zone as the fixed ABS challenge box, one top and bottom per batter, and it sits about 2.6 inches lower at the top than the average hitter’s tracked 2025 zone — 3.8 inches lower for Raleigh. Compare 2025 to 2026 as-tracked and last year’s “up in the zone” becomes this year’s “above the zone”: every chase and high-fastball number jumps without anyone changing anything. Our own first pass said high fastballs were up 50% (12.5% → 18.7%) and that he was swinging at 45% of pitches above the zone, up from 29.5%.

Put his 2026 box on every season and the high-fastball share goes .183 → .187, chase .371 → .382 (inside his 2022–24 range of .352–.383), zone rate .386 → .388. “They went upstairs and he’s chasing it” is, like-for-like, not there. The same artifact runs through much of this year’s discourse — and through a piece this site published in June on chase rates rising under ABS, whose +2.3 points becomes +0.5. That correction is coming.

His swing didn’t slow down, and he isn’t missing more

Bat speed 73.1 → 72.5 mph — a drop of 0.6, an ordinary year-to-year change league-wide; he is still in the 84th percentile (he was 91st). Swing length 7.84 → 7.77 feet, attack angle 17.4 → 16.5 degrees, swing-path tilt 34.3 → 34.2: all inside his 2023–25 range. Whiff rate per swing .315 → .317. Swing rate .506 → .508. Nor is he tired: bat speed by inning and by consecutive games caught is flat, and DH days are no better than catching days.

His decisions have eroded at the edges — in-zone swing rate .750 → .709, a point below his 2022–24 range; in-zone whiff .198 → .222, half a point above it — but his swing rate over the heart of the plate (.804) is inside his band, and a formal decomposition assigns essentially none of the hard-hit drop to swinging at different pitches. From the right side, on 66 balls in play, contact frequency is slipping too (strikeout rate .367, a career high) — a cell too small to build on.

Month by month, 2022 through August 2026 — hard-hit rate on top, wOBA below — against the range of his 22 qualifying 2022–25 months. Each dot is one regular-season month (faded = under 200 pitches) set against the lowest-to-highest month of his 2022–25 seasons; his hard-hit rate had already left that range in April 2026 (.346 on 78 balls in play vs a floor of .352) while his April wOBA (.306) was still inside it — before the diet, the strikeouts or the IL.

So was 2025 the outlier, or is 2026 the break? Both, on different families. 2025 was an outlier on outcomes and batted-ball conversion — a .195 barrel rate against a prior range of .126–.154, 60 home runs — with decisions and whiffs where they always were. 2026 is a break confined to contact quality: hard-hit .351 against a 2022–24 floor of .410 (a career low that ties his 88-ball rookie half-season), average exit velocity 89.0 against a floor of 89.5, expected wOBA on contact .363 against .420. On decisions and contact frequency, 2026 is mostly the 2022–24 hitter again. The ball off the bat is somewhere he has never been.

What broke

Hard-hit rate .496 → .351. Average exit velocity 91.3 → 89.0. Expected wOBA on contact .470 → .363. Barrels .195 → .125. Left-handed hard-hit .491 → .373, right-handed .508 → .303. Both swings broke.

The pattern is consistent with a centering loss — the barrel finding the middle of the ball less often — and we are choosing that phrase deliberately. On Baseball Savant’s squared-up measure he fell from 27.8% of contacts to 21.4%; on the version we pre-registered, the drop is p ≈ .09 with an interval that crosses zero, so it does not clear on its own.

The top end is more striking: balls that leave within 5% of the maximum his swing and the pitch allow — the flush hits — went from about one in seven (.135) to one in twenty (.049). We chose that cut after looking, and we say so: adjusted for the seventeen related cuts we inspected it survives at p ≈ .015, on a proxy, on 206 tracked balls, and its real test is the sealed rule below. Held at his 2025 centering, his 2026 bat speed would produce an average exit velocity near 90.8 mph rather than 89.0; the reverse swap gives 89.4. The point estimate attributes more of the drop to centering than to bat speed; the interval is wide enough that “consistent with” is as far as we go.

The shape of the miss, as far as it can be seen: he is getting under the ball more — the “under” share of his contact rose from .358 to .438 — with the same loft (average launch angle 25.2 → 25.4) and the same swing plane, concentrated in the upper half of the zone, where his hard-hit rate went .512 → .297 on 91 balls against .481 → .393 below. That reads as bat height against the ball, with the swing plane itself unchanged; it describes a small sample and stops short of a confirmed mechanism.

And he is not late. His contact point on four-seamers is where it always was (27.9 → 27.4 inches out front), and his whiff rate on 96-plus is unchanged. Where he has changed is on everything but the four-seamer: holding the pitch fixed, he now meets sinkers, cutters, breaking and offspeed pitches about 2.4 inches deeper (95% interval [1.4, 3.4]; the breaking balls most, about 3). That is real, and it costs him nothing we can measure: control for depth and the hard-hit deficit grows slightly (−.144 → −.162), and at the same bat speed, depth, direction, pitch and location his 2026 balls still leave the bat about 3 mph slower.

Every consecutive-season pair since 2021 with 300 plate appearances in the first year and 200 in the second (n = 1,124), as two strips: change in wOBA, and change in hard-hit rate. Standard, IBB-excluded wOBA. Every dot is one hitter’s consecutive-season pair since 2021; Raleigh’s −.139 wOBA drop and −.145 hard-hit drop each rank second of 1,124, behind only Yasmani Grandal (2021→22) and Cavan Biggio (2023→24) respectively.

His wOBA drop of −.139 is the second-largest of 1,124 consecutive-season pairs since 2021, behind only Yasmani Grandal’s 2021–22 (−.145); his hard-hit drop of −.145 is also second (Cavan Biggio, −.161). The league is hitting the ball less hard this year than in 2025 (.409 → .388), roughly two of his fourteen and a half points. Grandal — also a catcher, then 33 — is the one near-analog who paired a drop this size in results with a drop this size in contact quality: .414 to .269, and the following year he got back only to .301. So not “no analog.” One, and it did not recover.

When it broke

Two things in his swing did move, and they are the same move: his contact point on all swings came back from 35.5 to 33.0 inches in front of him, and the barrel points about three degrees less to the pull side at impact (−7.4 → −4.4). Both analyses dated that footprint independently and landed in the same place: it was already there in late August 2025. The best single break in the raw swing series falls on August 26 (the other method says September 16); mix-adjusted, his contact depth over the last 258 swings of 2025 was 33.6 inches, against 33.7 in 2026, and essentially every bootstrap break date falls before the IL. He hit with it, too: over that stretch his hard-hit rate was .475 and his expected wOBA on contact .498. The swing marker moved roughly seven months before the ball off the bat did, and it does not carry the contact loss now.

That is also where our own first draft was wrong. It said the footprint appeared “from opening week” and “looks deliberate, not injured.” The review found the late-2025 signal; both agents withdrew it. What we can say: a swing-marker shift already in place by September 2025, unchanged since and unchanged across the IL, that does not explain the loss. What we cannot say is why.

The reported wider stance is real on Baseball Savant’s stance leaderboard — about 3.6 inches wider than 2025 from both sides (24.8 to 28.4 inches), though narrower than he stood in 2023 — but that describes how he stands without explaining the collapse: Statcast never sees hands or load, and nothing here says what the stance did. A physical issue already present in March cannot be excluded either; the marker’s flatness across the IL speaks to the marker alone and carries no medical implication. The cause is not identifiable from tracking data.

What to watch the rest of the way

I want him back. So here is what I am going to watch, written to keep that hope honest. Thirty-seven games remain — roughly 150 plate appearances and 75–85 balls in play at his active-period rate.

First, what not to watch. His batting line: at 150 plate appearances a wOBA carries about a fifth of a hitter’s true talent (split-half reliability .19–.21; you need 580–660 PA for a coin flip), and even a full return to his 2025 self would produce a rest-of-season figure anywhere from .320 to .471 (the sealed 80% band at 117 PA). Home runs are worse (reliability .36; four the rest of the way fits the full-season-2026 level, ten fits “2025 is back,” and the bands overlap). Strikeout rate is a symptom of weaker contact. The one thing that has already “recovered,” the plain squared-up rate, is the weakest discriminator we tested. And bat speed never broke, so a change there would tell us about something else.

The sealed watch list. Top: how reliable each contact metric is at the sample he has left. Bottom, for each metric: his 2025 and 2026 levels, the 80% bands a “2025-Cal” and a “full-season-2026-Cal” stretch of 80 balls in play would produce, the pre-registered line, and where the last 80 balls in play banked (June 28–Aug 15) sit. Upper-half hard-hit is a ~35-ball cell at 80 BIP and is a tiebreaker only. Reliability is league split-half (Spearman–Brown) at each sample size; the reject lines, bands and rest-of-season endpoints were sealed on 2026-08-16 before any of the next 80 balls in play were seen — crossing a line rejects “full-season-2026 level” at one-sided α ≈ .10 per metric, and the rule needs two of the three contact metrics to cross.

What to watch is the ball off the bat, and specifically the two things that broke. Flush contact, as a count: 7 or more of his next 80 tracked balls in play (5 of the next 50). Hard-hit balls: 35 or more of the next 80 (23 of 50). Upper-half hard-hit rate at or above .405 is the tiebreaker, on a cell too small to headline. Contact depth (above 34.2 inches over his next 150 swings) is a marker only.

The sealed rule: two of the three cross at the same time, or nothing has been shown. Either of the two big ones alone can be luck about one time in ten; both together, about one in seventy; a hitter with his 2025 contact clears both about four times in five. Even then, crossing rejects “full-season-2026 level” — against the hitter of his last 102 balls in play the same rule fires about one time in five — and “consistent with 2025” is declared only if the crossing metrics also land inside their 2025 bands. The sample already banked crosses nothing: it does not reject full-season-2026 level, and the August uptick is not beyond noise. The clock started August 16.

Post-hoc accounting reweightings: expected-value wOBA with 95% bootstrap intervals (the hollow marker is the realized-outcome version), and each row’s share of the .139 gap. Every bar is an accounting reweighting of the same 208 balls in play — what the 2026 ledger would show if the 2025 outcome distribution returned in that one cell, everything else exactly as it happened — with no fix, no cause and no claim that any of it is reachable; the fast-swing and upper-half cells overlap (together they cover 156 of 208 BIP) and their intervals touch zero. These describe a fixed season and are not instructions.

One more thing, offered as accounting for a season already played, with no instruction in it for Cal Raleigh or anyone else. If his 2025 outcome distribution returned on all of his 2026 contact, the ledger would move about +.063 in expected wOBA (+.120 on realized outcomes) — only 42–45% of the gap on expected value, because a further .056–.058 of it is spray, park, defense and luck Statcast cannot assign. His fastest swings (+.039) and the upper half of the zone (+.035) are the two biggest cells, but together they are three-quarters of his balls in play and both intervals touch zero. Edge decisions are worth +.011, fouls and two-strike exposure +.009, and returning his soft-stuff contact depth to 2025 is worth nothing or less (−.008). The reweightings say where the loss is booked. Getting it back is a different question.

Finally, the sealed numbers, so we can be graded. Rest of season, at 117 plate appearances: if he is the full-season 2026 hitter, a wOBA of .253 with an 80% band of [.199, .312]; if he is the 2025 hitter again, .392 [.320, .471]. The bands do not overlap. For 2027, .307 [.265, .348] — a model-conditional figure from a pooled next-season regression, with his drop at the very edge of the sample and 200 plate appearances assumed. Nothing in that model says .390 next year, and the one near-analog says less. The file itself is published here, hashed (sha256 20095739…9a8de1, in full in the methodology) and untouched — run shasum -a 256 on it yourself. We’ll score it in October.

If you are watching tonight, watch the ball off the bat and count the flush ones. Ignore the box score. His chase rate sits inside his own range, his contact point on fastballs is where it always was, his bat speed is intact, and the pitch diet prices at zero. Something changed in what happens when barrel meets ball — consistent with the barrel finding the middle less often — before anyone noticed, and the honest answer to “why” is that we do not know.

Bring us the next question

If this is the kind of answer you want about a player you are watching, send us the question and we will run it. A hitter or a pitcher you are puzzled by, a claim from a broadcast you want checked, a number that looks too good to be true, a “is this real or is it noise?” — those make the best pieces on this site, and this one started as exactly that kind of question asked about my own favorite player. Write to [email protected], or find us on Bluesky and X. Every chart here comes from code we can point at a new question.