Every nightly game from the 2026 season. Each row is a game with its home-plate umpire's accuracy and missed-call count. Filter by date or team. Click any game to open the full pitch-by-pitch report. What's an umpire accuracy / wrong-call number? →
Loading game reports...
Umpire Profiles
Every home-plate umpire’s accuracy, zone style, and run-value impact
Every home plate umpire's zone style, accuracy, and run-value impact. Combines 2025 full-season data with 2026 live tracking. Zone styles → · Run-value impact →
Loading umpire data...
Umpire
Zone Style ⓘ
2026 Acc ⓘ
2026 BSR ⓘ
Wrong/Game ⓘ
2025 Acc
2025 BSR
Games
ABS Challenges & Catcher Framing
Who challenges, who overturns, who still steals strikes
2026 ABS challenge data. Who's challenging, how often the call gets overturned, and how much each overturn was worth (the count matters — a 3-2 overturn is worth ~5× a 1-0). Catchers overturn ~61% of their challenges vs ~45% for batters — the best challengers are catchers. Challenge metrics →
Loading ABS challenge data...
Catcher Framing in the ABS Era
Conventional wisdom said framing was dead under ABS — if any call can be challenged, why bother making a ball look like a strike?
The 2025 data tells a more interesting story: framing didn't lose its value, the mechanism just changed.
Read the analysis →
2026 live leaderboard live · nightly
Borderline strike rate per catcher: % of called pitches within 1.5" of the ABS zone edge that the umpire called a strike. Pulls from /api/catcher-framing (D1 called_pitches), updated nightly.
Loading live framing leaderboard…
2025 league reference static · 2025 full season
Best framer
51.7%
+9.1pp above league avg on borderline pitches
Worst framer
33.0%
−9.7pp below league avg
Framing gap
18.7pp
Best vs. worst on borderline calls
Sample
74
Catchers with 200+ borderline pitches received (2025)
Where framing operates: umpire accuracy by pitch speed
Pitch speed
Umpire accuracy
Framing leverage
Under 85 mph
93.9%
low
85 – 90 mph
93.9%
low
90 – 95 mph
91.8%
highest
95 – 100 mph
91.7%
high
100+ mph
91.9%
mid
The 90–95 mph band is where framing has the most room to operate — fast enough to create umpire uncertainty, common enough to dominate borderline calls. The 2025 framing data lives at the season level; 2026 per-catcher framing is collected nightly but not yet rolled up here.
Zone Explorer
Pitch-by-pitch zone visualization for any game, team, or umpire
Pitch-by-pitch zone visualization. Pick a single game or aggregate by team or umpire. Each dot is one called pitch; color tells you whether the umpire got it right. Useful for “was that strike-three actually a strike?” and for spotting an umpire's zone shape against any pitcher or batter handedness. How we score each pitch →
Select a game to visualize...
Pitcher Profiles
Live leaderboards: tunneling, command, and game stats
Pitcher profiles with tunneling scores, command data, and game stats. Tunnel = divergence (plate separation − decision-point separation). Read The Pitch Tunneling Atlas →
Bars are the type’s average percentile on each axis (high = better). These are descriptive shapes of the current grades, not a stable pitcher typology — we tested whether such types persist year to year and they don’t. Click any arm for its full profile.
Every graded pitcher grouped by the shape of its high-leverage grades; click any arm for the profile. How the grades work →
Loading pitcher data...
The atlas of league-wide pitch tunneling: 654 pitchers, 739,820 pitches. Tunnel divergence is plate separation
minus decision-point separation — how much two pitches diverge after the batter has committed to swing.
Read the flagship analysis →
Both pitches leave the hand from nearly the same point. The batter sees a single origin.
Two pitches leave the hand looking identical. By the decision point (24 ft) they've barely separated — the batter must commit. At the plate, they're a full strike zone apart.
Live 2026 leaderboard live · nightly
Loading live 2026 tunneling leaderboard…
Pulls from the 2026 nightly tunneling pipeline (compute_tunneling.py). Min 100 pitches, 2 games, 2 pitch types. Updates after each night's games.
2025 reference atlas static · 2025 full season
Most Deceptive
PitcherDiv.Whiff
Craig KimbrelRHP
+13.3"
29%
Kevin GinkelRHP
+12.9"
28%
Trevor MegillRHP
+12.5"
31%
Liam HendriksRHP
+12.0"
29%
Devin WilliamsRHP
+11.9"
37%
Drew PomeranzLHP
+11.9"
24%
Bryan AbreuRHP
+11.6"
42%
Spencer StriderRHP
+10.6"
31%
Blake SnellLHP
+10.3"
37%
Pablo LopezRHP
+10.2"
26%
Least Deceptive
PitcherDiv.Whiff
Taylor RogersLHP
-1.5"
15%
Joey CantilloLHP
-0.9"
18%
Anthony BenderRHP
-0.2"
20%
Luis MayRHP
+0.8"
17%
Cole HoldermanRHP
+1.3"
20%
Relievers
Starters
Full-season 2025 league-wide leaderboard from the flagship analysis. Frozen reference; the live 2026 view above shows current rankings.
Higher divergence (green) = pitches look similar at the decision point but end up in different places at the plate. 2025 season, pitchers with 30+ of each type.
Average divergence by pitch-type pair (2025 full season). Curveball + changeup tunnels best; 4-seam + sinker barely tunnels at all.
Each pitcher tracked against his own season baseline. Velocity and breaking-ball movement are reliable from a single outing — so this board flags the arms whose stuff is off, the leading sign of fatigue or decline. These are all pitchers; start-by-start tracking lives on each pitcher’s profile.
Stuff watch — off baseline live · nightly
Loading velocity watch…
Count Calculator
Outcome probabilities for any ball-strike count
Select any ball-strike count to see outcome probabilities. The count tells you almost everything about what happens next.
0S
1S
2S
0B
1B
2B
3B
No 2026 outcome data available yet for this count. Data is updated nightly.
Glossary:wOBA = Weighted On-Base Average (league avg ~.310; combines all offensive outcomes weighted by run value). Avg EV = average exit velocity on balls in play. False strike = called strike on pitch outside zone. Missed strike = called ball on pitch inside zone. Challenge value = counterfactual wOBA impact of a wrong call.
How this works
Outcome probabilities answer: "Given that the count reached X, what is the probability of each final at-bat outcome?" 2025 Full Season uses complete 2025 Statcast data. 2026 So Far uses the same methodology on 2026 data collected nightly. Umpire Accuracy shows called-pitch correctness by count.
Hot starts · Buy / Hold / Sell
Hitters
April hot starts are mostly noise — only 10% of past top-5 hot starters sustained 85% of their April pace.
We ran a dual-agent projection on every 2026 hitter (50+ PA) and named the six fakes, the six sleepers, and one carry from NPB.
Looking for hitter discipline instead of hot-start signal? Check the Coaching Gap → Hitters registry — quality hitters ranked by chase rate and contact quality.
What you're looking at: Six fakes (sell), one carry (hold), six sleepers (buy). Each row shows the preseason baseline,
April pace, and rest-of-season projection from a dual-agent regression model run on data through April 26.
This is a frozen projection — not an auto-updating leaderboard. Re-runs happen at milestones; the next is queued for the
All-Star break (Jul 14, 2026). Use the Coaching Gap → Hitters tab for live nightly hitter discipline data.
Hover any line for player details. (Tap on mobile.)
All 20 player-seasons that led 22-game wOBA in 2022-2025 (top 5 per season). Lines show the cumulative running wOBA at g22, g50, g100, and g162 — interior points are computed from the actual April + ROS data, not made up. Median full-season decline: -0.135 wOBA — roughly the gap between an MVP candidate and a league-average bat. Source: research/hot-start-half-life/data/noise_floor.json.
Hot-start half-life across history: how many of last decade's top-5 April hitters sustained their pace? Most didn't.
↘sell6 names
hot but regresses to baseline
Andy Pages
LAD
-1
vs prior
Prior .331April .403Proj .330
Ben Rice
NYY
+1
vs prior
Prior .345April .500Proj .346
Mike Trout
LAA
+8
vs prior
Prior .362April .425Proj .370
Aaron Judge
NYY
-7
vs prior
Prior .402April .435Proj .395
Corbin Carroll
ARI
+8
vs prior
Prior .380April .421Proj .388
Max Muncy
LAD
+10
vs prior
Prior .355April .407Proj .365
~hold1 name
above baseline with caveats
Munetaka Murakami
CWS
+57
vs prior
Prior .291April .418Proj .348
NPB rookie; prior is league-average proxy
↗buy6 names
above baseline, under the radar
Jac Caglianone
LAA
+70
vs prior
Prior .240April .320Proj .310
Everson Pereira
NYY
+87
vs prior
Prior .220April .411Proj .307
Jorge Barrosa
ARI
+81
vs prior
Prior .184April .322Proj .265
Samuel Basallo
BAL
+66
vs prior
Prior .246April .334Proj .312
Coby Mayo
BAL
+41
vs prior
Prior .263April .282Proj .304
Brady House
WSH
+45
vs prior
Prior .255April .298Proj .300
Prior (preseason baseline)April (current pace)Projected (R3 model verdict)Badge = Projected − Prior, in wOBA points. The actual signal: how much the model updates above your preseason expectation.
Sleeper relievers (4)
Antonio Senzatela (COL) • Daniel Lynch (KC) • John King (TEX) • Caleb Kilian (CHC)
Buy / Hold / Sell scoreboard for April 2026. Delta column = projected vs prior. SELL = April was noise, BUY = real above-baseline player.
Coming at the All-Star break. The April Sell List made specific calls on 13 hitters. At the break we'll have ~80 more games —
enough to test whether the six sleepers actually broke out, the six fakes regressed as projected, and whether Murakami's NPB-to-MLB signal held.
April Sell List re-check
Due: All-Star break (Jul 14, 2026)
Six fakes (Pages, Rice, Trout, Judge, Carroll, Muncy). Six sleepers (Caglianone, Pereira, Barrosa, Basallo, Mayo, House).
One NPB carry (Murakami). Tracked in the research queue.
Analysis as of April 26, 2026 · 50+ PA threshold · Dual-agent projection (Claude + Codex, R3)
Live tracker · Three rounds of dual-agent research
The Walk Spike
Walks are up roughly +0.68 percentage points in 2026 vs. the same calendar window in 2025 —
the highest league walk rate since modern memory. We've decomposed the cause across three rounds: the new ABS zone
shape explains about +26% of the spike; pitchers absorbing the top-edge change explain the rest.
The spike is fading week-to-week (P=89% it's regressing).
Current league walk rate vs 2025 same-window reference. Pulls from /api/count-outcomes, updated nightly.
Live weekly trajectory live · nightly
Loading live weekly trajectory…
2026 weekly walk rate (red) extends each week as the season progresses. 2025 dashed line is a fixed reference for the same calendar weeks. Pulls from /api/walk-spike/weekly.
Live walks by count live · nightly
Loading live walks-by-count…
Where the walks live, by count. 3-0 and 3-1 dominate; 0-0 walks are mathematically impossible from a single pitch. The 2025 reference column lets you spot which counts moved most in 2026.
Round 3 research findings static · May 14, 2026
What you're looking at: The weekly trajectory of 2026 walk rate vs. 2025, the multi-season history,
and where the spike concentrates by count. Through Week 7, the 2026 line has converged toward 2025 —
the spike is real but bending down.
Round 3 — eight findings, three rounds, six counterfactual implementations, two cross-reviews per round
The spike is fading in real time. Bayesian P(regressed within 2026) = 89%.
2026 weekly walk rate (red) starting March 27 vs. the same calendar windows in 2025 (gray). W1-3 averaged 9.93%; W5-7 averaged 9.07% — a within-season drop of 0.86pp.
League walk-rate history. 2026 sits above every full season since 1950.
Hover a count for the per-bin walk-rate Δ. Cochran's Q across all 12 counts: p = 0.67 — no per-count concentration.
Where the extra walks come from, broken down by count.
What you're looking at: The ABS strike zone shrunk at the top edge by 7-8 percentage points
in called-strike rate. Replaying every 2026 plate appearance under the 2025 zone shape attributes
roughly +26% of the spike to the zone change — with the rest on pitcher behavior.
Hover a cell for the per-bin Δ called-strike rate and 95% bootstrap CI. Dashed box: ABS rule zone for a 6-ft batter.
Called-strike rate by zone location: 2026 (top) vs. 2025 (bottom). The top edge is where the ABS zone gave back to hitters.
Counterfactual: walks attributable to the new zone vs. walks attributable to pitcher behavior.
What you're looking at: The pitchers paying the bill share an archetype — top-of-zone command guys
whose 2025 strike calls disappeared in 2026. R3 triangulated this finding across two independent statistical pipelines.
Each dot is one pitcher (≥40 IP in 2025, ≥200 pitches in 2026). The dashed line is the Bayesian model fit. Three pitchers cleared bootstrap stability in both pipelines; Mason Miller cleared only Claude's.
Pitcher archetype scatter: which command profiles absorbed the top-edge zone change.
Six independent counterfactual implementations. Five land positive. The grayed-out Bernoulli replay is preserved as a stress test, not a headline.
Two pipelines, same data, different methods — both flag the same archetype of pitcher as the cleanest examples of the pattern.
Coming with R4 at the All-Star break. Round 3 settled the league-aggregate magnitude (~+26%) and identified
the archetype effect. Round 4 will surface per-actor breakdowns: which umpires drove the top-edge first-pitch loss, which
teams adapted fastest, and whether catcher framing still meaningfully shifts the new zone or ABS feedback has washed it out.
Tracked in the research queue. Subscribe to know when it ships.
Data through May 12, 2026 · 46,755 PAs · 28,579 borderline pitches
Live tracker · Updated nightly
The Coaching Gap
The CalledThird research finds a +0.04 wOBA edge for low-chase hitters against predictable MLB pitchers,
replicating across all five seasons since 2022. This tracker follows it through 2026 — at the level the signal
actually lives: batter discipline trajectories over weeks and months, not single-night predictions.
What you're looking at: Year-over-year shifts in plate discipline, and whether that discipline change
actually paid off on predictable pitches. This is where the signal lives — within-batter trajectories across seasons,
not single-night matchup bets.
Loading year-over-year trends...
What you're looking at: Disciplined hitters with strong contact quality. These are the batters
who turn a predictable pitcher's sequencing into damage. low chase = good · high chase = bad
Loading hitter registry...
Hitter
Team
Chase rate ⓘ
Contact quality ⓘ
Whiff %
Type
2026 edge ⓘ
PA
What you're looking at: Pitchers whose next pitch the model can predict at above-average rates,
given count, handedness, and sequence. Higher predictability = juicier target for a disciplined lineup.
more predictable · more deceptive
Loading pitcher data...
Pitcher
Team
Predictability ⓘ
vs 2025 ⓘ
Pitches
Role
Which lineups are built to exploit predictable pitching? Count of quality hitters (low chase + high contact quality)
per roster. Teams stacked with them can systematically extract the coaching gap across a full season.
Loading hitter registry…
Quality hitters (low chase, high contact)
Free swingers
How we know this is real — and where it isn't. Every claim on this tracker is backed by one of
these three tests. We show you all three, including the one that failed.
Loading validation backtest...
How to read this tracker
This tracker follows discipline trajectories — the timescale where the chase-vs-predictability effect actually shows up. Per-night matchup predictions don't (Spearman ~0 across 70K+ historical matchups). Year-over-year change in chase-rate on predictable out-of-zone pitches does (Spearman −0.26 vs Δ predictable wOBA, 267 recent transitions); whiff-rate on predictable swings tracks it even more tightly (−0.29). Each is measured on 3–4× the per-hitter sample of the wOBA column it replaced, which is why the cohort signal survives at the individual-row level.
Chase rate
% of out-of-zone pitches a hitter swings at. League range 0.17 – 0.46. Lower = more patient — this is the trait the data validates as the gap-extracting axis.
Predictability score
AUC of a per-pitcher fastball-vs-offspeed model using count, handedness, and pitch number. 0.5 = coin flip, 1.0 = perfectly readable. Typical MLB range 0.55 – 0.80; a reliever like Tim Hill with one pitch sits near 0.85. Matches the metric used in the flagship article.
Δ chase on predictable bait
Year-over-year change in a hitter's swing-rate on predictable (top-quintile) out-of-zone pitches. Measured on ~150–400 pitches per hitter-season — 3–4× the sample of the wOBA column it replaced. This is the mechanism: the coaching gap only pays off when a hitter actually stops swinging at the predictable bait. Cohort Spearman against Δ predictable wOBA: −0.26.
Δ whiff on predictable swings
Year-over-year change in whiff-rate on swings at predictable pitches. The payoff layer: even when a hitter swings, does the predictability mean they make more contact? Cohort Spearman against Δ predictable wOBA: −0.29 (the strongest single behavioral predictor we tested).
Predictable-pitch wOBA
A batter's wOBA on the top 20% most predictable pitches they faced (within each pitcher-season). Shown at the cohort level only (big improvers +0.012 wOBA, big decliners −0.029) — dropped from the per-row table because individual hitter-season samples (~50 PAs) are too small to read reliably.
Full methodology + 6-round dual-agent research write-up in the flagship analysis. Open-source code in the research repo.