Opening Day in San Diego, third inning. Nick Pivetta starts Javier Báez with a sweeper at 79 mph, and Báez swings through it. Statcast’s newest number puts the bat’s closest pass at about 8 inches from the ball. Pivetta throws the sweeper again. Báez puts it in play at 80 mph, a soft out.
An inning later, Riley Greene swings through a Ron Marinaccio four-seamer at 93 mph on a 3-1 count, and the bat’s closest pass is about a tenth of an inch away. Marinaccio throws the four-seamer again at 3-2. Greene swings through that one too, for strike three.
We picked these two moments with a rule written down before we looked at any 2026 at-bat, and side by side they go against the pattern this article reports: the big miss was followed by contact, the near miss by another whiff. The findings below are rates. Even after a big miss, most swings at the same pitch make contact, and Báez’s did. After a near miss, about one swing in four at the same pitch still misses, and Greene’s did.
A new number, and the next pitch
Baseball Savant began publishing miss distance in June. For swinging strikes since the middle of 2023, it records how close the bat came to the ball at its nearest point. Most swings and misses are close: the typical one passes about an inch and a half from the ball. About one full-swing whiff in seven misses by six inches or more. We wanted to know whether that number says anything about what happens on the very next pitch.
When the pitcher comes back with the same type of pitch, it says something about whether the next swing misses. We took every swinging strike from the second half of 2023 through 2025 that had a recorded miss and was followed by another pitch from the same pitcher to the same hitter in the same plate appearance: 117,104 of them. Then we kept the full swings (bat speed of at least 50 mph on the whiff, which drops most checked swings) whose next pitch was also swung at: 56,997. Finally we split them two ways: by whether the pitcher came back with the same type of pitch, and by how far the first swing missed.
Whiffs per swing on the next pitch after a full-swing whiff, adjusted for the next pitch’s location, the count, the pitch type, the season, and how often that pitcher’s pitch and that hitter usually draw whiffs. Both lines are averaged over one common mix of those situations. Pitchers still chose which pitch to throw using information we can’t see, so the distance between the two lines at any one miss size says nothing about which choice works better, including where the lines cross. Read each line from left to right: one climbs, the other stays flat.
After a full-swing miss of under an inch, the same pitch drew a whiff on 26.0% of swings. After a full-swing miss of six inches or more, 37.6%. The rate rose at every step in between, and the gap of 11.6 percentage points has an interval of 10.0 to 13.3. When the pitcher switched to a different type of pitch, the whiff rate stayed between 26.9% and 27.7% however big the first miss had been.
This hitter, against this pitch
The obvious objection is that a big miss is a sign of a nasty pitch, and a nasty pitch thrown again will draw another miss. We controlled for the measurable versions of that: where the next pitch went, the count, the pitch type, how often that pitcher’s pitch of that type usually gets whiffs, and how often that hitter usually whiffs. The pattern held. It held when we also accounted for where the missed pitch itself had been, and it held, if anything a little stronger, when we compared only the same pitcher and the same hitter within the same game.
The sharper test uses the next batter. If the pattern belonged to the pitch that day, the next hitter to swing at it should show it just as strongly. Put both on a common footing, with the same counts and the same pitch locations, and the next hitter’s association with the earlier miss comes out about a third as strong as the original hitter’s. Both of our independent analyses put it there, and in both the gap between the two hitters is clear of zero. A second check runs time backward, pairing each miss with the hitter who came before; any link there would have to come from the pitch that day. That link came out small and slightly positive in our main version, so some of the pattern may belong to the pitch that day, and this check can’t say how much. The data point most strongly to this hitter, against this pitch.
A season we sealed before looking
Every rate above comes from 2023 through 2025. Before anyone looked at 2026, both analyses froze their code and wrote down a range for the 2026 gap between the biggest and smallest full-swing misses: 8.5 to 14.7 percentage points. Then the 2026 season was built and scored once.
The repeat gap by season, on each season’s own mix of pitches, with 95% intervals; the dashed line is the three earlier seasons pooled (+11.6). The 2023 point covers the second half only, when miss distance begins. The 2026 point was scored once, after the band was frozen. On switches, the 2026 gap was 0.9 points, with an interval spanning zero.
It came in at 13.5 points, inside the band, while switches stayed flat. A sealed pass certifies that the association held in a new season; it says nothing about cause.
One more test asked a forecaster’s question. Before the next pitch is thrown, with its location left out because nobody knows it yet, does knowing how far the last swing missed help predict whether a swing at the same pitch will miss again? It does, in 2025 and again in the sealed 2026 season, in both analyses. The gain is small in absolute terms, and it counts as real-time information: the forecast uses only what is known before the next pitch is thrown.
Where the extra misses show up
In the raw shares, the rise in misses on the same pitch is matched mostly by a drop in fouls. After full-swing whiffs, fouls fell from 45% of swings at the same pitch after the smallest misses to 27% after the biggest. Balls in play moved much less, from 30% to 27%, and the contact that did happen was softer: average exit velocity fell from 88.3 to 83.8 mph. On switches, where the adjusted whiff rate held flat, exit velocity rose from 87.6 to 89.6 mph; these raw switch groups mix very different pitches, so that rise says nothing about which choice works better.
Raw shares of next-pitch swings after full-swing whiffs, 2023 (second half) through 2025. Fouls, balls in play and whiffs add up to every swing, so these raw rows imply a bigger jump in misses than the adjusted 26% to 38% above. Much of that difference is the mix: the biggest-miss group is mostly breaking balls, which draw more misses and fewer fouls to begin with. Read these rows for where the change appears and the first chart for how big it is. There were 3,542 balls in play after the smallest misses and 1,295 after the biggest, on the same pitch again; exit velocity averages those with a recorded speed.
What pitchers did, and what we can’t tell you
In the raw counts, after the biggest misses pitchers came back with the same breaking ball 56% of the time, against 45% after the smallest, and with the same offspeed pitch 53% of the time against 43%. Fastballs moved the other way, from 52% to 42%, though after adjustment their direction is unsettled.
Raw share of next pitches of the same type as the whiffed pitch (a slider after a slider, a changeup after a changeup), across all qualifying whiffs, checked swings included: they are 19% of the six-inches-or-more group. These are pitchers’ actual choices, made with information this study can’t see. After adjustment, the breaking-ball and offspeed increases hold up and the fastball decline has no established direction.
The question every reader will ask next is whether pitchers should come back with the same pitch more often after a big miss. This study cannot answer it. We pre-registered a test, and it came back inconclusive. Pitchers decide what to throw using information we can’t see, so the pitches they chose to repeat are a selected sample of the ones they could have thrown. No method available here can fully correct for that, and the adjustments we could make left intervals that include zero.
What to watch for
Savant now posts the miss on nearly every swing and miss, and here is one way to read it. When a hitter takes a full swing and misses by six inches or more, and the pitcher comes back with the same pitch, the next swing at it has missed about 38% of the time after adjustment, against 26% after a full-swing miss of under an inch, in the 2023–25 data, and the gap held in 2026. The pattern is concentrated on this hitter against this pitch, and the choice of the next pitch belongs to teams, who may see things we can’t. These are regular-season numbers; October is a smaller sample that we have left untested.
Methodology
Data, the full-swing rule, the model, both analyses, the sealed 2026 test, and what failed
Data. Statcast pitch-level data with Baseball Savant’s miss distance (published June 2026; available from the second half of 2023). The unit is a qualifying swinging strike: a whiff with a recorded miss distance, followed by another pitch from the same pitcher to the same hitter in the same plate appearance, excluding position players pitching. There are 117,104 in 2023 (second half) through 2025. A whiff with two strikes ends the plate appearance, so every qualifying whiff came with zero or one strike. “Same pitch” means the same pitch-type group (four-seamer, sinker, cutter, slider, sweeper, curveball, changeup, splitter).
Full swings. Every whiff, foul and contact rate in the article uses full swings: bat speed of at least 50 mph on the whiffed swing. Checked swings that count as strikes make up 16% of the six-inches-or-more bin among repeats that drew a swing and show no dose-response of their own, so including them would blur the comparison. The pitch-choice shares in “What pitchers threw next” cover all qualifying whiffs, checked swings included; checked swings are 19% of that chart’s six-inches-or-more group. Statcast moved its recorded pitch location from the front of the plate to the middle in 2026; we converted 2026 to the earlier convention before any comparison.
The model. Whiff per swing on the next pitch, from a linear probability model fit to repeats and switches together, with every control allowed to differ between the two: the next pitch’s location, the count, the pitch type, the season, the pitcher’s usual whiff rate on that pitch type, and the hitter’s usual whiff rate (each computed without the event being predicted). The four-bin rates are averaged over one pooled mix of 56,997 full-swing whiffs followed by a swing, so both lines and all bins describe the same situations. Intervals come from a pitcher-cluster bootstrap with 1,000 draws. Repeat gap from the smallest to the biggest misses: 11.6 points [10.0, 13.3]. Switch gap: 0.2 [-1.4, 1.8]. Measured per inch, the relationship is steepest at small misses and flattens at large ones, which is why we report bins.
Robustness. The association stayed positive, with intervals clear of zero, when we added the whiffed pitch’s own location, compared within pitcher-games, compared within batter-season and pitch type, restricted to the first whiff of a plate appearance, used location surfaces fit separately by pitch type, and measured whiffs per pitch in place of per swing. Run value on the next pitch moved in the pitcher’s favor. Within the same pitcher, hitter and game, using the 16.5% of events where that comparison is possible, the association came out somewhat larger than the estimate across all events, with a wider interval that includes it.
This hitter against this pitch. The next batter’s first swing at the same pitch type was compared with the original hitter’s swing on common count-by-location cells. The next batter’s association came out at about a third of the original hitter’s in both analyses (a ratio of 0.32 in one and about 0.34 in the other), and the difference between them was positive with an interval clear of zero. The next batter’s own association is also positive. A time-reversed check pairing each miss with the previous hitter’s swing at the same pitch type gave estimates from slightly negative to slightly positive across constructions. That permits a small component that belongs to the pitch that day and does not bound it.
Two independent analyses. One interpretability-first pipeline (linear, generalized-linear and shrinkage models) and one machine-learning pipeline (gradient boosting, matching and double machine learning) worked from the same shared data build, confirmed by an identical hash of the event list, and were forbidden from reading each other until both had filed. Where they ran the same specification, they reproduced each other’s estimates exactly, which confirms they built and read the data the same way; with their own, different methods, they reached the same verdict on every pre-registered test. Three rounds of cross-review followed, after which both signed the same ledger of claims. Before launch, the research brief was itself stress-tested by two reviewers from different model families, who caught the 2026 change in Statcast’s plate convention, the checked-swing problem, and a pitch-family composition artifact in the raw comparison.
The sealed 2026 test. Both analyses froze their predictions and acceptance bands and hashed them before the 2026 data were built. The season was then built with the same code and scored once, after a comparability report written in advance confirmed that the miss-size bins had shifted by less than a percentage point. Repeat gap 13.5 points [10.7, 16.2] against a band of 8.5 to 14.7; switch gap 0.9 [-1.5, 3.4]. The formal pass rule, also written in advance, used a continuous version of the same relationship and the repeat-versus-switch difference; both landed inside their bands in both analyses. In the forecasting test, pre-pitch models trained on 2023–24 and validated on 2025 forecast the next swing at the same pitch better with miss distance than without it, with the next pitch’s location excluded, and the frozen models repeated that improvement on 2026.
What did not pass. The decision test, which asked whether repeating does relatively better after big misses than after small ones, came back inconclusive: its intervals include zero in both analyses, and the matched version discarded nearly every event. A season-level “early read” found that a pitch’s early miss distances go with its rest-of-season whiff rate, but the gain they added to held-out forecasts did not clear zero in the pre-registered test, so we report that as an association only. We make no claim about deception, pitch tunneling, game-calling, catchers, the postseason, or any individual pitcher or hitter.
The opening scene. The two at-bats were chosen by a rule written and hashed before we searched the 2026 at-bats: the first 2026 regular-season game containing both a repeat after a full-swing miss of six inches or more and a repeat after a full-swing miss under an inch, taking the earliest event of each. Whatever happened next was reported as recorded. The rule selected Detroit at San Diego on March 26.
Prior work. A 2019 FanGraphs Community study by Drake found that repeating the same pitch type and zone after a one-strike whiff went with higher strikeout rates. Miss distance lets us ask the question by the size of the miss, which that work could not measure.
Cite this analysis
CalledThird. "Throw It Again? What the Size of a Miss Says About the Next Pitch." CalledThird.com, October 2, 2026. https://calledthird.com/analysis/throw-it-again
All CalledThird analysis is original research. If you reference our findings, data, or charts in your work, please link back to the original article. For data inquiries: [email protected]