Through the 3 August trade deadline, MLB players had issued more than 7,100 automated ball-strike challenges at a 53% success rate. Read as an aggregate, that looks like noise. Read as a distribution, it shows a new and unevenly held skill: catchers like Carson Kelly and Dillon Dingler succeed at rates far above chance while other challengers drag the average down. The challenge system did not remove human judgment from ball-and-strike calls. It relocated it from the umpire to the player and attached a scarce resource to it.
There is now most of a season of data on baseball's automated ball-strike challenge system, and the headline number is deeply unimpressive. As of the 3 August trade deadline, players had issued more than 7,100 challenges league-wide, per ESPN's tracking, and succeeded 53% of the time. Teams average just over four challenges a game. A coin flip, with the challenger holding a sliver of an edge.
Read carelessly, that says the system is useless — that players cannot tell a ball from a strike any better than chance, and the whole apparatus is theatre. Read properly, it says something considerably more interesting, and it is not what either side of the decade-long robot-umpire argument predicted.
Fifty-three percent is an average across everyone permitted to challenge — catchers, pitchers and hitters, in every game situation, including the desperate ninth-inning challenge on a pitch nobody thought was close. Averages across mixed populations hide the thing you want to know.
The Cubs' Carson Kelly is 21-for-25 while catching, per ESPN — 84% — which Baseball Savant translates to 2.3 runs above expectation, the highest mark of any catcher, hitter or pitcher. The Tigers' Dillon Dingler is 15-for-17, 88.2%, described as an MLB high. The Rockies' Hunter Goodman is 20-for-26, 76.9%. These are not the same claim, and it is worth resisting the temptation to rank them cleanly: Dingler's rate is higher than Kelly's but on far fewer attempts, and seventeen tries is few enough that a couple of coin flips would reorder the leaderboard. Kelly's twenty-five attempts at 84% is the more robust of the two lines, which is why the runs-above-expectation figure lands on him.
All figures are as of the 3 August 2026 trade deadline, per ESPN's tracking. ABS data moves weekly and individual player lines will have shifted since.
The point is not that a few catchers are lucky. It is that if some challengers run at 80-plus percent over meaningful samples and the league average is 53%, a large number of challenges are being thrown by people who do not know where the pitch was. The aggregate is not measuring a skill. It is measuring a skill averaged with a guess.
Someone can reasonably argue the opposite: that 53% is barely better than chance, that a handful of high performers is exactly what you would expect from random variation across hundreds of challengers, and that the outliers are survivorship rather than skill. That argument deserves testing rather than dismissal, because it is the sort of thing that is usually right.
What weighs against it is that the gradient was visible before the regular season began and it fell along a predicted line. In spring training, catchers succeeded on 56% of challenges and pitchers on 41%. That is not a random scatter — it is a fifteen-point gap between two groups, in the direction anyone who has thought about the mechanics would have guessed in advance. A catcher receives the pitch, sees it from behind and level with the zone, and has the best available view of where the ball actually crossed. A pitcher is sixty feet away, looking at the pitch from behind and above, and is generally the last person in the ballpark who should be trusted to evaluate his own work.
A prediction made from first principles, confirmed in spring training and holding through August, is not what pure noise looks like. The counter-case is not defeated, but it now has to explain a structured gradient rather than an unstructured one.
For a decade the standard argument against automated strike calling included an elegy for pitch framing — the catcher's craft of receiving a borderline pitch so quietly that the umpire calls it a strike. Framing was the last genuinely invisible skill in baseball, it was worth real runs once teams learned to measure it, and automation was going to erase it.
It did not. It relocated it. The underlying ability framing rewarded was never the smooth glove — it was knowing, precisely and instantly, where the pitch crossed the plate. Under ABS the same knowledge is worth just as much, applied about half a second later and through a different channel: not persuading an umpire, but deciding in the moment whether to spend one of two challenges. Kelly's 2.3 runs above expectation is the same kind of value framing used to generate, produced by the same underlying perception, expressed through a new mechanism.
There is a real difference, and it is that the new version is scarce. A catcher could frame every pitch of a game. He gets two challenges, and a successful one is not deducted — which converts the skill into something closer to card counting. Knowing the pitch was a ball is necessary but not sufficient; you also have to judge whether this ball, in this count, in this inning, is worth the resource, given that being right costs you nothing and being wrong costs you the option for the rest of the game.
The framing that has dominated this debate treats the challenge system as a compromise on the road to full automation — a transitional half-measure until everyone gets comfortable. This season's data is the strongest available argument that it is a different animal entirely.
Full automation removes human judgment from ball-and-strike calls. The challenge system does not remove it; it moves it from the umpire, who was never on anyone's roster, to a player who is — and in doing so creates a new, coachable, unevenly distributed and directly valuable skill. That is a design outcome, and a fairly sophisticated one. It preserves the thing the sport was actually worried about losing, which was not human error but human agency, while removing the specific errors that generated the most anger — the missed call on strike three with the bases loaded, which is now correctable by anyone paying attention.
Whether MLB eventually goes further is a separate question and the league has been careful not to foreclose it. But the case that it should has weakened this season rather than strengthened, because the challenge system turned out to do something full automation cannot: it gave players a new way to be good at baseball.
League-wide and individual figures are as of 3 August 2026 per ESPN's ABS tracking; spring training splits and the rule structure are per ESPN and FOX Sports. These figures update weekly during the season.
Each team gets two challenges per game. Only the pitcher, catcher or batter can initiate one, and it must be done immediately — no dugout input, no replay review. If the challenge succeeds and the call is overturned, the team keeps it; only unsuccessful challenges are deducted. That structure is what makes it a resource-management problem as well as a perception one: being right is free, being wrong costs you the option for the rest of the game.
DrafterDaily reads sports data for the distribution underneath the average — where the skill actually sits and what it is worth.
More sports coverageThe NFL wants a 50-60% raise from its broadcast partners. Fox has said no on the merits. CBS cannot say anything at all, because an antitrust dispute in an unrelated deal has frozen its owner.
Sportradar is now supplying Polymarket with official data, live streaming and integrity monitoring across 20-plus competitions and roughly 300,000 matches a year. The counterparty argument survives. The stack argument is new.
A $245.3 million cap and a $171.2 million floor. Twelve clubs would have to add $617 million; eight would have to cut $578 million. Read those two numbers together and the proposal's real shape appears.