When should we challenge, and who should?
Who loses the runs
Pittsburgh vs. league success rate by role, 2026; runs are accuracy in the spots challenged
Defense wins 58% of challenges league-wide (catchers 59%, pitchers 40%); offense wins 49%. Pittsburgh trails in every role. The costliest role: catchers, 3.0 runs. Role runs are accuracy: overturns above or below an average challenger in the same count and game state, at Savant’s run values with the state cost. They add to −5.9; Savant’s −6.4 also counts which calls were challenged and how many (−0.5).
Before and after July 1
Pittsburgh vs. MLB by half: success rate (challenges) and the share made in early counts (0-0, 0-1, 1-0)
| Won | Before 7/1 | After |
|---|---|---|
| Pittsburgh | 41% (159) | 52% (196) |
| MLB | 53% | 54% |
| Pittsburgh catchers (MLB) | 48% (59%) | 54% (58%) |
| Pittsburgh batters (MLB) | 39% (48%) | 49% (50%) |
| Early-count share (MLB) | 40% (42%) | 41% (42%) |
By month, Pittsburgh’s rate and its gap to MLB in points. Across the 30 clubs about 54% of the spread in catchers’ success is real and 10% of batters’; the rest is noise. So Pittsburgh’s 2026 shortfall of 21 overturns comes to about 8 once club accuracy is regressed; with the 2027 catchers’ own records in place of 2026’s catcher shift, about 4, and the second half was already closer to average.
Challenge runs vs. average · all 30 clubs
Savant runs above average: against an average club facing the same pitches. NL Central in bold.
The break-even card
How sure a challenger must be for a challenge to pay: cost of losing ÷ (gain + cost). Cost = the challenge a failure burns, by game state.
| Count | Gain | Two left, before the 7th | Last one, before the 7th | 7th inning on | Old flat cost |
|---|---|---|---|---|---|
| 0-0 | 0.08 | 32% | 53% | 17% | 73% |
| 0-1 | 0.10 | 27% | 47% | 14% | 68% |
| 0-2 | 0.20 | 16% | 31% | 8% | 51% |
| 1-0 | 0.12 | 23% | 43% | 12% | 64% |
| 1-1 | 0.12 | 23% | 42% | 12% | 63% |
| 1-2 | 0.24 | 13% | 27% | 6% | 47% |
| 2-0 | 0.19 | 16% | 32% | 8% | 53% |
| 2-1 | 0.18 | 17% | 33% | 8% | 54% |
| 2-2 | 0.33 | 10% | 21% | 5% | 39% |
| 3-0 | 0.21 | 15% | 30% | 7% | 50% |
| 3-1 | 0.28 | 12% | 24% | 6% | 43% |
| 3-2 | 0.60 | 6% | 13% | 3% | 26% |
The cost is Savant’s state-dependent cost of a failed challenge, averaged by challenges left and inning. An option value measured independently from 4,858 club-games (the overturn runs clubs went on to win with two left vs. one) agrees: r = 0.95 across innings, mean 0.039 vs. Savant’s 0.032. Each club gets two challenges and keeps the ones it wins; in 2026 no club challenged after two failures before the 10th inning, and 67 did in extra innings, so the data show an extra challenge there.
Who may challenge, and when
Expected success for each role, count and game state (league, shrunk toward the role’s count rate), shifted by Pittsburgh’s own accuracy, regressed. ● the 80% range clears the break-even · ◐ toss-up · · hold.
| Role · state | 0-0 | 0-1 | 0-2 | 1-0 | 1-1 | 1-2 | 2-0 | 2-1 | 2-2 | 3-0 | 3-1 | 3-2 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Catchers Two left, before the 7th | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● |
| Catchers Last one, before the 7th | ●† | ● | ● | ●† | ● | ● | ● | ● | ● | ● | ● | ● |
| Catchers 7th inning on | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● |
| Batters Two left, before the 7th | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● |
| Batters Last one, before the 7th | ●† | ◐† | ●† | ● | ●† | ●† | ● | ●† | ● | ● | ● | ● |
| Batters 7th inning on | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● |
The average challenge paid in every cell but batters’ 0-1 with the last challenge (a toss-up). Shift for Pittsburgh’s own 2026 accuracy, regressed: catchers −3.9 points, batters −0.5. † A failure costs more in the early innings. With the last challenge, batters hold on 0-0 and 0-1 through the 2nd inning and on 1-1 through the 1st inning unless the call is clear. Toss-ups: batters on 0-0 and 0-1 in the 3rd–4th, 0-2 through the 2nd, 1-1 in the 2nd–3rd and 1-2 and 2-1 through the 1st; catchers on 0-0 through the 3rd and 1-0 through the 1st. Ranges: every rate is refit on 400 resamples of whole club-games, so the uncertainty in the league rates the cells lean on is in the range. A cell compares the average challenge with the break-even; the card is the bar for the pitch in front of you.
What the policy is worth · forward test
Rules fit on the 5,330 challenges before July 1, scored on the 5,227 after it. Runs per club over a season, 80% range.
Run as published (early-inning holds, toss-ups kept), it would have held 20 of the 5,227 second-half challenges, which netted −0.08 runs, so holding them was worth +0.08 league-wide: about 0 per club-season (80% range −0.04 to +0.05). Its value is in what it doesn’t throw away, not in runs added to how clubs already challenged. On point estimates alone it would have held 35, netting +0.48: labeling toss-ups changes little.
It would have held 2,777 of them (early counts on a flat 0.21-run cost; no hold from the 7th inning on with no challenge lost and the score within 2), and they netted +139.5 runs: about 1.0 wins per club a season lost (8.9–9.9 runs on resampled games). The old page didn’t say how late or how close, so the spread across readings matters more than that range: −9.0 to −10.4 runs (from the 7th, 8th or 9th; within 1 to 3 runs), −10.6 with no exception, and −6.8 if it covered every challenge from the 7th. Pittsburgh: 110 of 196 held, netting +5.0.
Scored at 0.21 a failure, the same holds look like a saving: +3.9 runs per club a season, 59 league-wide in the second half. The verdict flips on the cost, which is why the cost model is checked against the club-game option value.
What’s left is accuracy: Pittsburgh’s 2026 club accuracy, regressed, is about 8 overturns short of average, 1.7 runs (0.18 wins). The catcher part (7 overturns lost) comes from the 2026 catchers’ shift. With the 2027 catchers’ own 2026 records instead, weighted by the plan’s plate appearances, the total is about 4 overturns lost (0.8 runs, 0.08 wins). Which holds for 2027 depends on whether the shift was the catchers’ or the club’s; either way it is small. Challenging more might add runs (573 overturnable calls went unchallenged), but the data can’t say how sure players were on calls they let go, so that value isn’t measured. In-sample (whole season), this card would have held 0 of Pittsburgh’s 355 challenges; the old grid would have held 194, and those netted +6.9 runs. Pitchers’ challenges league-wide (78 after July 1) are left as made: the card routes them through the catcher, which the log can’t score. Method: costs, gains, success rates, their ranges and club shifts fit on the 5,330 challenges before July 1; scored on the 5,227 after it. The card is scored as published: holds by game state and through the inning shown, toss-ups kept, pitchers’ challenges left as made. The old grid is scored with its late-game exception (no hold from the 7th on with no challenge lost and the score within 2). Runs = what the held challenges actually netted (Savant’s run value when won, minus Savant’s state cost when lost); a rule’s value is minus that. Per club-season: the second-half total ÷ 30 clubs, scaled to a full season of challenges. 80% range: 2,000 resamples of second-half club-games.
The policy · set before Opening Day
What the staff runs in the game, from the break-even card and the forward test
- 1
Catchers own defensive challenges; pitchers don’t challenge.
Pitchers win 40% of challenges league-wide (Pittsburgh 29% on 17); catchers win 59%. Route every defensive challenge through the catcher, with the pitcher’s read as input.
- 2
With both challenges left, challenge any call you’re at least 32% sure of.
That’s the bar at 0-0 before the 7th (40% in the 1st, when a lost challenge costs most). With two strikes or two balls it’s at most 17%. A failure here costs about 0.04 runs, because the second challenge is still in hand.
- 3
With the last one before the 7th, the bar is 13%–53%.
A failure now costs about 2.4× as much (0.09 runs). It costs more still early in the game. Batters hold on 0-0 and 0-1 through the 2nd inning and on 1-1 through the 1st inning unless the call is clear. Toss-ups: batters on 0-0 and 0-1 in the 3rd–4th, 0-2 through the 2nd, 1-1 in the 2nd–3rd and 1-2 and 2-1 through the 1st; catchers on 0-0 through the 3rd and 1-0 through the 1st.
- 4
From the 7th inning on, use them.
The bar is at most 17% (37% at worst: 0-0 in the 7th with one left), and a challenge left at the end of the game is worth nothing.
- 5
No green lists: everyone works from the card.
A batter’s success hasn’t been shown to repeat (odd vs. even challenges, r = −0.05, 95% range −0.20 to 0.09, 182 hitters). A catcher’s hasn’t been shown to either (r = 0.08, −0.02 to 0.19, 338 player-seasons, MLB 2026, AAA 2025, AAA 2026). Track Triple-A challenges for feedback, not for promotion.
An earlier version of this page said to hold early-count challenges unless a call was obviously wrong, with one exception (“late in a close game with challenges unused, the bar drops”), on a flat 0.21-run cost per failure. The forward test above is why it doesn’t now.
Pittsburgh’s catchers for 2027
Roles from the Winter Plan (decisions feed). 2026 challenges as a catcher, MLB and Triple-A, regressed toward the MLB catcher rate.
| Catcher | Role | MLB | Triple-A | Regressed |
|---|---|---|---|---|
| Rafael Flores Jr. | Starter | 20 of 38 | 19 of 35 | 57% |
| Endy Rodríguez | Backup· 60-day IL | 9 of 22 | 9 of 21 | 55% |
| Derek Berg | Triple-A depth | — | 30 of 46 | 60% |
| Christian Bethancourt | Third catcher (minor-league deal) | — | 44 of 77 | 58% |
| Henry Davis | Trade or release before the Rule 5 filing (Winter Plan) | 56 of 103 | — | 57% |
Henry Davis, traded or released before the Rule 5 filing in the Winter Plan, made 103 of Pittsburgh’s 175 catcher challenges in 2026. The catcher row carries Pittsburgh’s 2026 catcher shift (−3.9 points, regressed), so part of it walks out with the catchers leaving. Recomputed with no shift, and at both ends of MLB catchers’ own regressed shifts (−4.5 to +7.6 points, 91 catchers with 10+ challenges), no call and no early-inning hold changes; only the toss-up innings for catchers on 0-0 and 1-0 with the last challenge move. Who catches doesn’t change what the staff runs.
Pittsburgh catchers
League catcher rate 58.6%
| Player · policy | Ch | Won | Runs |
|---|---|---|---|
| Henry DavisTrade or release before the Rule 5 filing (Winter Plan) | 103 | 54% | −1.1 |
| Rafael Flores Jr.Count policy | 38 | 53% | −0.5 |
| Endy RodríguezCount policy | 22 | 41% | −1.1 |
| Joey BartNo longer with club | 11 | 55% | −0.2 |
| Shawn RossNo longer with club | 1 | 0% | −0.1 |
| All | 175 | 52% | −3.0 |
Pittsburgh batters
League batter rate 48.9%
| Player · policy | Ch | Won | Runs |
|---|---|---|---|
| Marcell OzunaNo longer with club | 23 | 52% | +0.1 |
| Bryan ReynoldsCount policy | 23 | 52% | +1.1 |
| Spencer HorwitzCount policy | 19 | 26% | −1.1 |
| Ryan O'HearnCount policy | 17 | 47% | −0.8 |
| Esmerlyn ValdezCount policy | 16 | 56% | +0.4 |
| Nick GonzalesCount policy | 14 | 50% | 0.0 |
| Jake MangumCount policy | 13 | 31% | −1.0 |
| Konnor GriffinCount policy | 7 | 57% | +0.5 |
| Oneil CruzCount policy | 6 | 17% | −0.4 |
| 10 others | 25 | −0.9 | |
| All | 163 | 44% | −2.1 |
Pittsburgh pitchers
League pitcher rate 39.7%
| Player · policy | Ch | Won | Runs |
|---|---|---|---|
| Gregory SotoLeaving: free agent | 6 | 33% | −0.1 |
| Camilo DovalNo longer with club | 3 | 33% | 0.0 |
| Bubba ChandlerRoute to catcher | 3 | 33% | −0.1 |
| Braxton AshcraftRoute to catcher | 2 | 0% | −0.2 |
| Dennis SantanaNo longer with club | 1 | 0% | −0.1 |
| Carmen MlodzinskiRoute to catcher | 1 | 100% | +0.1 |
| Yohan RamírezRoute to catcher | 1 | 0% | −0.3 |
| All | 17 | 29% | −0.8 |
Players are listed by the club they challenged for, so each table adds up to its role above. Pittsburgh’s failed challenges missed by a median of 1.06″ (league 1.00″); 24% missed by 2″ or more (league 20%), and Savant rated 29% of them reasonable attempts (league 33%). Neither difference is outside the noise, so the data don’t say whether the misses were misreads or close calls. By zone edge, Pittsburgh won 46% at the top (expected 52%, 59), 46% at the bottom (expected 54%, 120) and 48% on the sides (expected 53%, 176); by pitch, 44% on fastballs (52%, 212), 52% on breaking balls (54%, 111) and 50% on offspeed (55%, 32). Expected = the MLB rate for the same roles, counts and game states. Overall, 47% against 53% expected (z = −2.3). No edge or pitch group stands out from the rest of Pittsburgh’s challenges once 6 splits are allowed for (|z| under 2.64), so the shortfall is club-wide.
Who does it best · MLB leaders by challenge runs
Benchmarks, not targets: neither a batter’s nor a catcher’s success has been shown to repeat, so these lists are last season’s results, not a skill to buy.
- Jonah Heim ATH 65% · +4.3
- Hunter Goodman COL 68% · +3.9
- Austin Wells NYY 63% · +3.1
- Joe Mack MIA 64% · +2.8
- Ryan Jeffers MIN 61% · +2.8
- Victor Caratini MIN 59% · +2.2
- Nolan Schanuel LAA 56% · +1.8
- Jose Altuve HOU 57% · +1.7
- Luke Keaschall MIN 55% · +1.5
- Sal Stewart CIN 55% · +1.4
- Bryce Harper PHI 45% · +1.3
- Kevin McGonigle DET 61% · +1.3
Source: Baseball Savant ABS challenge log and leaderboards, 2026 regular season. Leaders are ranked by Savant’s runs above average and tagged with the club they made most of their 2026 challenges for. Method and tests: the ABS model card.