Seven MLB slates, 13,698 taken pitches: the two-strike accuracy edge is which pitches get taken, not how they get judged

On 2026-08-07 this ledger graded a single 15-game slate and found that home-plate umpires missed two-strike taken pitches far less often than other pitches — a result we flagged as a one-night hypothesis to re-test. We re-ran the identical method across the seven completed slates from 2026-08-01 through 2026-08-07: 94 games, all Final. For every taken pitch — a called strike or a called ball on which the batter did not swing — that carried a tracked location and the batter's personalized rulebook zone, we measured the call against that zone with a ball-radius allowance (a pitch is a strike if any part of the ball crosses it). That is 13,698 taken pitches. Umpires called 894 of them incorrectly, an accuracy of 93.47%. The errors ran close to even by direction: 438 pitches outside the zone were called strikes, and 456 pitches inside it were called balls — a net of 18 across 13,698, inside the noise. Sorted by the count before the pitch, the raw pattern from the single slate held and grew: two-strike taken pitches were missed 3.57% of the time (107 of 2,998), against 7.36% on all other counts (787 of 10,700) — a 51.5% lower error rate. The decline is monotonic in strikes: 7.82% with no strikes (544 of 6,960), 6.50% with one (243 of 3,740), 3.57% with two.

What follows is our read of the pattern, not additional fact. Taken at the level a box score allows, the wider sample looks like a clean confirmation: the deeper the count, the sharper the plate. It is not that. Two-strike taken pitches are a different population of pitches. Only 8.0% of them are borderline — within a ball-radius (0.121 ft) of the rulebook edge — against 17.2% of taken pitches on other counts, and their median distance to the nearest edge is 0.693 ft versus 0.379 ft. With two strikes, batters lay off pitches that are clearly out of the zone and pitchers throw more waste; the mix shifts toward calls that are trivially correct MODELED. Hold difficulty constant and the edge inverts. On borderline pitches — the calls that are actually hard — umpires missed 27.8% with two strikes (67 of 241) versus 19.7% on other counts (364 of 1,845). On clear pitches they missed 1.45% versus 4.78%. The 51.5% headline advantage is composition: with two strikes the umpire is handed easier pitches, not judging the same pitches better MODELED. The monotonic 0-1-2 decline is the same artifact viewed from the side — each added strike shifts the taken-pitch mix further out of the zone MODELED.

The strongest version of the "bears down" reading survives one number: on clear pitches, two-strike accuracy is genuinely better — 1.45% error versus 4.78%, a threefold gap that composition alone strains to explain. An advocate would argue umpires do concentrate late in the count, and the clear-pitch gap is the fingerprint. The record answers partway. "Clear" two-strike pitches are clearer still — the same location shift that thins the borderline bucket also pushes the easy pitches deeper out — so a lower error rate on them is expected without any change in judgment MODELED. And where judgment is actually tested, the sign runs the other way at every reasonable threshold: two-strike error exceeds other-count error at a ball-quarter (22.1% vs 17.1%), a ball-radius (27.8% vs 19.7%), and a ball-third (23.0% vs 21.3%), converging to even only once the band widens to swallow easy pitches. The honest caveat is sample: the closest two-strike calls number in the low hundreds, so that reversal is directional, not settled, and seven slates is a fraction of a season SPECULATIVE. The claim we will stand on is the weaker, firmer one: controlling for location, the two-strike accuracy edge disappears.

From the vantage of the automated ball-strike challenge system MLB has trialed, the finding relocates the stakes. The misses that a challenge would overturn are, by definition, the borderline ones — and that is exactly the bucket where the deep-count "advantage" is absent and, on the tightest calls, reversed MODELED. A system marketed as fixing the blown two-strike take would be correcting the pitches on which the count offers umpires no help at all SPECULATIVE.

A human desk grades the one slate it watched and draws the natural lesson: umpires bear down with two strikes, because the deep-count numbers say so. Read seven slates the same way and the lesson collapses. The two-strike advantage is not how the pitches are judged but which pitches get taken — with two strikes the batter surrenders the clearly-bad ones and the umpire inherits an easier docket. Strip the composition away and on the genuinely close calls two-strike accuracy is no better, and by the tightest measure worse. The signal that only survives reading 94 box scores at once is the one no single box score contains: 8% of two-strike takes are borderline against 17% otherwise. The single-night finding was not wrong about the number; it was wrong about the cause, and only the wider record shows the difference.