MLB games
15
Game IDs, teams, first pitch, venue, and status.
Projections
Daily simulation center
A forecast is a range of game stories, not one final score
The simulator turns current game inputs into transparent win probabilities, score ranges, matchup drivers, bullpen effects, and player opportunities, with uncertainty visible everywhere.
Scheduled-game laboratory v4 Today’s slate is running in development modeActual starters, posted orders, active rosters, weather, empirical runner profiles, and named bullpens · calibration pending
Live input assembly · July 26, 2026
Today’s projection inputs
BaseballR supplies the schedule, probable starters, posted orders, and active rosters. Park coordinates route each game to its Open-Meteo window, and the development simulator now consumes the resulting game rows.
Complete games
0/15
0 confirmed batting-order rows.
Verified players
780
248 workload-qualified bullpen rows remain after the roster join.
Resolved games
15/15
15 of 15 park environments resolved
| Away | Home | Away starter | Home starter | Away order | Home order | Park weather | Temp | Wind | Rain | Input state |
|---|---|---|---|---|---|---|---|---|---|---|
| Cleveland Guardians | Tampa Bay Rays | Parker Messick | Drew Rasmussen | missing | missing | St. Petersburg, Florida | Indoor | - | - | conditional |
| Arizona Diamondbacks | Washington Nationals | Kohl Drake | Miles Mikolas | missing | missing | Washington, District of Columbia | 69 F | 2.3 mph | 0% | conditional |
| Chicago Cubs | Pittsburgh Pirates | Jameson Taillon | Braxton Ashcraft | missing | missing | Pittsburgh, Pennsylvania | 67 F | 1.7 mph | 0% | conditional |
| Toronto Blue Jays | Boston Red Sox | Kevin Gausman | Ranger Suarez | missing | missing | Boston, Massachusetts | 57 F | 3.5 mph | 0% | conditional |
| Atlanta Braves | Baltimore Orioles | Reynaldo López | Shane Baz | missing | missing | Baltimore, Maryland | 68 F | 3.3 mph | 1% | conditional |
| Kansas City Royals | Detroit Tigers | Luinder Avila | Framber Valdez | missing | missing | Detroit, Michigan | 67 F | 4.2 mph | 0% | conditional |
| Los Angeles Dodgers | New York Mets | Emmet Sheehan | Freddy Peralta | missing | missing | Queens, New York | 63 F | 3.5 mph | 0% | conditional |
| San Diego Padres | Miami Marlins | Walker Buehler | Janson Junk | missing | missing | Miami, Florida | 80 F | 2.8 mph | 4% | conditional |
| Athletics | Minnesota Twins | Jeffrey Springs | Connor Prielipp | missing | missing | Minneapolis, Minnesota | 75 F | 3.6 mph | 32% | conditional |
| Houston Astros | Chicago White Sox | Ronel Blanco | Erick Fedde | missing | missing | Chicago, Illinois | 70 F | 5.3 mph | 1% | conditional |
| Colorado Rockies | Milwaukee Brewers | Kyle Freeland | Jacob Misiorowski | missing | missing | Milwaukee, Wisconsin | 74 F | 5.6 mph | 5% | conditional |
| Cincinnati Reds | St. Louis Cardinals | Andrew Abbott | Kyle Leahy | missing | missing | St. Louis, Missouri | 79 F | 2.9 mph | 0% | conditional |
| Seattle Mariners | Texas Rangers | Logan Gilbert | Jacob deGrom | missing | missing | Arlington, Texas | 85 F | 9.4 mph | 0% | conditional |
| Los Angeles Angels | San Francisco Giants | José Soriano | Carson Whisenhunt | missing | missing | San Francisco, California | 56 F | 4.9 mph | 0% | conditional |
| New York Yankees | Philadelphia Phillies | Will Warren | Cristopher Sánchez | missing | missing | Philadelphia, Pennsylvania | 64 F | 2.8 mph | 0% | conditional |
Current boundary: a resolved input row is not the same as a validated forecast. Starter, lineup, active-roster bullpen, and temperature effects now enter the score model; park factors, fresh reliever workload, and out-of-time calibration remain open gates.
Matchup layer waiting: the Phase 1 batter-starter distributions are withheld until the current slate has posted batting orders.
Phase 4 waiting: the detailed state engine requires two complete nine-man orders, active bullpens, current hitter-reliever probability matrices, and the daily empirical baserunning products.
Player simulation board · lineup gate
Player probabilities wait for posted batting orders
The team slate can update in the morning, but hitter and starter-event boards are withheld until current-date orders provide real batting-order opportunity. Yesterday’s player board is never relabeled as today’s.
Current state: 0 posted batting-order rows are available for July 26.
The actual slate in one scan
Each game card now comes from today’s schedule and shows the simulated win split, score center, uncertainty, close-game chance, and input status. These are pre-calibration research outputs, not public betting forecasts.
AwayCleveland Guardians43.0%
HomeTampa Bay Rays57.0%
Mean score3.4 – 3.9Model total7.3One-run game39.1%
AwayArizona Diamondbacks64.5%
HomeWashington Nationals35.5%
Mean score5.1 – 4.0Model total9.1One-run game33.8%
AwayChicago Cubs31.8%
HomePittsburgh Pirates68.2%
Mean score4.3 – 5.8Model total10.1One-run game30.1%
AwayToronto Blue Jays40.8%
HomeBoston Red Sox59.2%
Mean score3.2 – 3.8Model total7.0One-run game40.3%
AwayAtlanta Braves41.6%
HomeBaltimore Orioles58.4%
Mean score4.1 – 4.8Model total9.0One-run game35.4%
AwayKansas City Royals39.6%
HomeDetroit Tigers60.4%
Mean score4.2 – 5.0Model total9.3One-run game34.4%
AwayLos Angeles Dodgers64.8%
HomeNew York Mets35.1%
Mean score5.0 – 3.9Model total8.8One-run game34.1%
AwaySan Diego Padres48.7%
HomeMiami Marlins51.3%
Mean score4.5 – 4.6Model total9.1One-run game35.7%
AwayAthletics35.2%
HomeMinnesota Twins64.8%
Mean score4.6 – 5.8Model total10.4One-run game31.3%
AwayHouston Astros40.1%
HomeChicago White Sox59.9%
Mean score5.0 – 5.8Model total10.8One-run game32.0%
AwayColorado Rockies15.5%
HomeMilwaukee Brewers84.5%
Mean score2.7 – 5.6Model total8.3One-run game23.8%
AwayCincinnati Reds41.5%
HomeSt. Louis Cardinals58.5%
Mean score4.1 – 4.7Model total8.8One-run game35.4%
AwaySeattle Mariners47.0%
HomeTexas Rangers53.0%
Mean score3.9 – 4.1Model total7.9One-run game38.2%
AwayLos Angeles Angels54.2%
HomeSan Francisco Giants45.8%
Mean score4.4 – 4.1Model total8.5One-run game36.5%
AwayNew York Yankees42.3%
HomePhiladelphia Phillies57.7%
Mean score3.3 – 3.8Model total7.0One-run game39.7%
Feature simulation · 20,000 draws
San Diego Padres at Miami Marlins
The closest matchup on today’s scheduled slate shows the current score distribution, model drivers, and honest input limitations.
Model leanMiami Marlins51.3%
San Diego Padres4.580% range 2–7
at
Miami Marlins4.680% range 2–7
Winning-margin distributionAway ← outcome → Home
Away by 5+
5.8%
Away by 3-4
12.9%
Away by 2
12.0%
Away by 1
18.0%
Home by 1
17.6%
Home by 2
12.1%
Home by 3-4
14.3%
Home by 5+
7.4%
Why the model lands hereLeague ranks
- Posted-lineup offenseMiami Marlins93 vs 101 blended wRC+ after platoon and recent-form adjustments
- Probable starterSan Diego PadresWalker Buehler 4.5 vs Janson Junk 4.65 form-adjusted run estimator
- Available bullpenSan Diego Padres3.61 vs 4.14 FIP blend with active reliever selector
- Park and weatherRun environment80 F temperature adjustment; park factor 1
Mean total9.180% range 5–13One-run game35.7%Close-game likelihoodExtras13.7%Regulation tie rateBullpen effect-4.4 ptsHome win probability subtracts
Most common exact scoresAway – home
#1 exact score4 – 54.6%#2 exact score5 – 44.5%#3 exact score4 – 34.0%#4 exact score3 – 44.0%#5 exact score5 – 63.3%
Publication gate · July 26, 2026
The scheduled-game engine is running behind a calibration wall
The page can now simulate today’s actual games. Development probabilities stay clearly labeled until current PBP workload, empirical park effects, and a chronological backtest pass.
57.1%activation gates closed
Scheduled games
15 games
Posted lineups
0/15 confirmed
Probable starters
15/15 matched
Active bullpens
Roster verified
Current PBP usage
Current
Park factors
Neutral placeholder
Chronological calibration
Not yet passed
Model fitting · no-lookahead evaluation
Real outcomes change the model only after they earn the right
The fitted score layer learned from prior-only rolling team production, then faced a completely later block of games. It improved the raw score scale and narrowly beat the simple home-team Brier baseline, earning shadow-mode evaluation without overwriting the public board.
Later games only
282
Training ended June 26
Brier score
0.249
Naive home baseline 0.251 - Shadow-mode eligible
Team-runs MAE
2.63
Raw scale 2.93 before fitting
Eligible forecasts
36
264 more game forecasts before live probability calibration
Freeze before first pitch
Every game and player-event probability is archived with its inputs and model version. Late builds are permanently excluded.
Settle against final PBP
Team scores, winners, hitter events, and starter strikeouts are joined by exact game and MLBAM player IDs.
Diagnose the misses
Brier score, log loss, run MAE, calibration bins, and player-event residuals reveal whether the model is too aggressive or too conservative.
Refit in chronological blocks
Only coefficients that improve later unseen games enter shadow mode; public probabilities require a second approval gate.
Why today does not count: the first ledger snapshot was taken after first pitch, so all 15 games were correctly marked late and excluded. The automated daily job will archive future boards before games begin.
Starter decision path
The handoff becomes part of every game simulation
A fixed tied-game workload path shows when the pooled hook model begins shifting probability toward the bullpen.
Hold the starter
72 pitches · 19 BF · 2x through
First major decision
88 pitches · 24 BF · 3x through
Bullpen strongly enters
105 pitches · 28 BF · 3x through
Descriptive pooled-model scenarios, not causal manager tendencies or a forecast for a named starter.
Reliever chain planner
Handedness, leverage, and fatigue change the next arm
Once a starter exits, each three-batter pocket reranks the available bullpen and carries planned workload into the next decision.
Defending Milwaukee Brewers
Scenario
New York Yankees bullpen path
-
7inning
Bridge pocket · vs LHB91.8 Ryan Yarbrough LHP
Availability 100.0% · matchup 100.0% · planned 12 pitches
-
8inning
Setup pocket · vs RHB80.4 Paul Blackburn RHP
Availability 100.0% · matchup 70.0% · planned 12 pitches
-
9inning
Finish pocket · vs RHB68.2 Paul Blackburn RHP
Availability 72.5% · matchup 70.0% · planned 12 pitches
Defending New York Yankees
Scenario
Milwaukee Brewers bullpen path
-
7inning
Bridge pocket · vs LHB77.3 Grant Anderson RHP
Availability 100.0% · matchup 35.0% · planned 12 pitches
-
8inning
Setup pocket · vs RHB82.2 Craig Yoho RHP
Availability 100.0% · matchup 70.0% · planned 12 pitches
-
9inning
Finish pocket · vs LHB64.5 Grant Anderson RHP
Availability 72.5% · matchup 35.0% · planned 12 pitches
Selector score, not probability: the displayed score transparently combines active-roster eligibility, availability, role fit, performance, handedness, and leverage. A confirmed batting order and same-day transaction check remain required before operational use.
Slate table
Every game, every uncertainty signal
The live development run is shown here. Public release remains blocked until calibration and the remaining data-quality gates pass.
| Away | Home | Away win | Home win | Away runs | Home runs | Total | One-run | Extras | Model lean |
|---|---|---|---|---|---|---|---|---|---|
| Cleveland Guardians | Tampa Bay Rays | 43.0% | 57.0% | 3.4 | 3.9 | 7.3 | 39.1% | 15.0% | Tampa Bay Rays |
| Arizona Diamondbacks | Washington Nationals | 64.5% | 35.5% | 5.1 | 4.0 | 9.1 | 33.8% | 12.6% | Arizona Diamondbacks |
| Chicago Cubs | Pittsburgh Pirates | 31.8% | 68.2% | 4.3 | 5.8 | 10.1 | 30.1% | 11.4% | Pittsburgh Pirates |
| Toronto Blue Jays | Boston Red Sox | 40.8% | 59.2% | 3.2 | 3.8 | 7.0 | 40.3% | 15.3% | Boston Red Sox |
| Atlanta Braves | Baltimore Orioles | 41.6% | 58.4% | 4.1 | 4.8 | 9.0 | 35.4% | 13.3% | Baltimore Orioles |
| Kansas City Royals | Detroit Tigers | 39.6% | 60.4% | 4.2 | 5.0 | 9.3 | 34.4% | 12.8% | Detroit Tigers |
| Los Angeles Dodgers | New York Mets | 64.8% | 35.1% | 5.0 | 3.9 | 8.8 | 34.1% | 12.9% | Los Angeles Dodgers |
| San Diego Padres | Miami Marlins | 48.7% | 51.3% | 4.5 | 4.6 | 9.1 | 35.7% | 13.7% | Miami Marlins |
| Athletics | Minnesota Twins | 35.2% | 64.8% | 4.6 | 5.8 | 10.4 | 31.3% | 12.0% | Minnesota Twins |
| Houston Astros | Chicago White Sox | 40.1% | 59.9% | 5.0 | 5.8 | 10.8 | 32.0% | 12.2% | Chicago White Sox |
| Colorado Rockies | Milwaukee Brewers | 15.5% | 84.5% | 2.7 | 5.6 | 8.3 | 23.8% | 8.8% | Milwaukee Brewers |
| Cincinnati Reds | St. Louis Cardinals | 41.5% | 58.5% | 4.1 | 4.7 | 8.8 | 35.4% | 13.5% | St. Louis Cardinals |
| Seattle Mariners | Texas Rangers | 47.0% | 53.0% | 3.9 | 4.1 | 7.9 | 38.2% | 14.2% | Texas Rangers |
| Los Angeles Angels | San Francisco Giants | 54.2% | 45.8% | 4.4 | 4.1 | 8.5 | 36.5% | 14.0% | Los Angeles Angels |
| New York Yankees | Philadelphia Phillies | 42.3% | 57.7% | 3.3 | 3.8 | 7.0 | 39.7% | 15.6% | Philadelphia Phillies |
Model card · scheduled-game laboratory v4
What the live and shadow models know
Included in current probabilitiesPosted-lineup wRC+, probable-starter FIP/xFIP/ERA blend and workload, active-roster bullpen quality, temperature, home field, and Monte Carlo score uncertaintyPhase 4 shadow evaluationNamed runners and relievers now drive innings, outs, score, lineup continuity, hooks, player advancement, steal tendency and success, pitcher hold, empirical park movement, walk-offs, and expanded player linesPublication gateUse the earliest 70% of 300-plus settled pregame forecasts for fitting, require improvement on the later 30%, archive versions, and audit misses before any public promotion
From prototype to daily forecast
Three of four activation stages are now underway
The scheduled slate and player simulations now feed a permanent pregame ledger. Version two adds restrained platoon, recent-form, probable-starter, and available-reliever effects; the next work is proving those additions on future games.
-
01
Built
Acquisition spine
Incremental PBP cache, FanGraphs refreshes, weekly award checkpoints, schedule, starters, lineups, rosters, and weather.
-
02
Built in development
Matchup-aware simulator
Current games run through batting-order quality, starter hand and form, hitter platoon and form, available bullpen strength, weather, park, home field, team score, hitter events, and starter strikeouts.
-
03
Shadow mode
Chronological calibration
Future pregame snapshots will test score, win, home-run, hit, total-base, and strikeout probabilities before any public promotion.
-
04
Publication
Automated approved board
Only model versions that repeat their advantage on clean live forecasts can replace development probabilities or supply newsletter excerpts.
Interpretation boundary: these are today’s scheduled matchups and real Monte Carlo outputs, but the probability model is still uncalibrated. Treat the page as a research prototype, not betting advice or a public forecast.
Build path
What the current development model uses
Each refresh rebuilds the slate from posted information. The current version blends batting-order quality, starter handedness, platoon results, recent form, probable-starter run estimators, active-reliever quality and workload, park, weather, and home field before drawing game outcomes.
01
Score distribution layer
Complete-game draws, win splits, score intervals, close-game likelihood, and reproducible seeds.
02
Matchup-aware inputs
BaseballR schedule, probable starters, posted batting orders, active rosters, park coordinates, and Open-Meteo game windows populate the daily contract. Posted hitters now receive restrained platoon and recent-form adjustments.
03
Named-reliever and runner state
The engine carries named runners and relievers through innings, outs, score, batting-order continuity, workload hooks, double plays, productive outs, steals, walk-offs, extra innings, and expanded player counting stats.
04
Empirical baserunning and calibration
Phase 4 learns runner advancement, steal tendency, steal success, pitcher hold, and venue movement from completed games. Every input is league-shrunk and remains in shadow mode while rolling backtests, model-version tracking, misses review, and the projection archive accumulate.
These are development projections rather than betting advice. The probabilities will remain visibly labeled until repeated time-based backtests and the live pregame ledger demonstrate stable calibration.
The first bullpen-entry layer is now measurable
The pooled hook model is evaluated on later game dates that were not used for fitting. These are development diagnostics, not production probabilities.
Validation rows
22,386
Training through 2026-06-29
ROC AUC
0.788
Higher means pulled and retained decisions are ranked more distinctly.
Brier score
0.074
Lower is better; calibration still requires improvement before publishing forecasts.
Hook rate
8.9% / 9.1%
Observed first, mean prediction second.
How workload changes the decision environment
Representative tied-game contexts scored through the pooled model. These fixed scenarios isolate the shape of the model, not a specific manager.
higher hook likelihood80.2
Starter, deep workload
7 inning | tied game
80.2% modeled hook probability
105 pitches | 28 batters faced
higher hook likelihood29.1
Starter, third time through
6 inning | tied game
29.1% modeled hook probability
88 pitches | 24 batters faced
middle hook likelihood16.8
Reliever, first pocket
7 inning | tied game
16.8% modeled hook probability
12 pitches | 3 batters faced
Hook scenario board
Starter and reliever decision points
| Rank | Situation | Role | Inn. | Pitches | BF | TTO | Hook prob. | Relative |
|---|---|---|---|---|---|---|---|---|
| 1 | Starter, deep workload | starter | 7 | 105 | 28 | 3 | 80.2% | higher |
| 2 | Starter, third time through | starter | 6 | 88 | 24 | 3 | 29.1% | higher |
| 3 | Reliever, first pocket | reliever | 7 | 12 | 3 | 1 | 16.8% | middle |
| 4 | Reliever, extended outing | reliever | 8 | 30 | 8 | 1 | 16.6% | middle |
| 5 | Starter, middle innings | starter | 5 | 72 | 19 | 2 | 4.8% | lower |
| 6 | Starter, early workload | starter | 3 | 45 | 12 | 2 | 1.2% | lower |
What moves the model
Observed manager-decision factors
Coefficient direction is descriptive. Correlated game situations mean these are not isolated causal effects.
| Factor | Coefficient | Association |
|---|---|---|
| Workload beyond 60 pitches | 2.09 | Higher modeled likelihood |
| Starter role | -1.42 | Lower modeled likelihood |
| Reliever role | 1.37 | Higher modeled likelihood |
| Batters faced beyond 18 | 1.19 | Higher modeled likelihood |
| Adverse plate-appearance result | -0.79 | Lower modeled likelihood |
| Third time through order | -0.23 | Lower modeled likelihood |
| Score within two runs | 0.10 | Higher modeled likelihood |
| Team trailing by four-plus | -0.04 | Lower modeled likelihood |
| Later inning | -0.04 | Lower modeled likelihood |
A Boston proof of concept for the broadcast workflow
Left- and right-handed matchup pockets combine recent availability, pitcher role, exact MLBAM identity, handedness, and current results allowed.
| Next batter | Rank | Reliever | Throws | Role | Available | Performance | Matchup | Selector |
|---|---|---|---|---|---|---|---|---|
| L | 1 | Jovani Morán | L | reliever | 87.2% | 78.9% | 100.0% | 89.6 |
| L | 2 | Alec Gamboa | L | reliever | 92.2% | 50.0% | 100.0% | 84.4 |
| L | 3 | Ryan Watson | R | reliever | 92.2% | 30.5% | 35.0% | 66.5 |
| R | 1 | Jovani Morán | L | reliever | 87.2% | 78.9% | 40.0% | 77.6 |
| R | 2 | Ryan Watson | R | reliever | 92.2% | 30.5% | 70.0% | 73.5 |
| R | 3 | Justin Slaten | R | reliever | 89.4% | 34.4% | 70.0% | 73.4 |
Roster gate connected: candidates now must appear on the BaseballR active roster and survive the pregame workload screen. The selector remains a research lead, not a live recommendation.
Calibration audit
Predicted probability against observed decisions
| Bin | Rows | Min predicted | Max predicted | Mean predicted | Observed hook rate |
|---|---|---|---|---|---|
| 1 | 2,238 | 0.4% | 0.5% | 0.5% | 0.8% |
| 2 | 2,239 | 0.5% | 1.1% | 0.9% | 1.5% |
| 3 | 2,238 | 1.1% | 1.1% | 1.1% | 0.3% |
| 4 | 2,239 | 1.1% | 1.7% | 1.2% | 0.6% |
| 5 | 2,239 | 1.7% | 7.3% | 4.7% | 5.0% |
| 6 | 2,238 | 7.3% | 8.8% | 8.0% | 7.3% |
| 7 | 2,239 | 8.8% | 14.8% | 12.3% | 9.6% |
| 8 | 2,238 | 14.8% | 16.7% | 15.3% | 20.0% |
| 9 | 2,239 | 16.7% | 17.5% | 17.1% | 22.1% |
| 10 | 2,239 | 17.5% | 98.2% | 30.1% | 22.0% |