For sports analysts and research desks

Know which side the evidence favors in every game

An AI analyst picks a side in each game and explains why, and every pick is scored against the final result.

Guardians vs. White Sox · 2026 postseason

Both models picked Cleveland, and Cleveland won 2–0.

  • Win probability56%
  • Head-to-head19–11
  • White Sox in Cleveland5–10
  • News foundNone
Each pick is logged on game day and scored after the final out.
Track record of every logged baseball pick, scored against final results

Each pick is built from facts you can check.

The Bundle checks each club’s form, head-to-head record, park history, and forecast weather, and records the source of each.

Guardians vs. White Sox · 2026 postseason

51.5%

Cleveland’s expected win rate

From runs scored and allowed, against 49.3% for Chicago. Home teams won 54% of games last season.

Wikipedia game results

19–11

Head-to-head record

Cleveland’s record against Chicago across the 30 games in the Bundle’s schedule data.

Wikipedia game results

5–10

Chicago in Cleveland

Chicago’s record at Cleveland’s park, where it lost by 0.6 runs a game on average.

Wikipedia game results

6 mph

Forecast wind

With a 2% chance of rain at an assumed 7 p.m. start, so the weather did not affect the pick.

Open-Meteo forecast

Every pick is scored against the final result.

Two pick models run side by side, and every pick is marked correct or incorrect once the game is final.

2026 MLB season

Check-based pick

Full season

Five pass-or-fail checks, with the pick and reasoning written by an AI analyst

59%

183 of 310 scored picks correct

42 picks that favored neither side are not counted.

Probability model

Live for the postseason

A win probability for every game, with settings re-checked each week

65%

26 of 40 scored picks correct

Replayed over 2,456 earlier 2026 games, it picked 55.6% correctly.

GameDodgers – Brewers
Check-based pickDodgers Incorrect
Probability modelBrewers, 54% Correct
FinalBrewers 3–2
GameWhite Sox – Guardians
Check-based pickGuardians Correct
Probability modelGuardians, 56% Correct
FinalGuardians 2–0
GameWhite Sox – Guardians
Check-based pickWhite Sox Incorrect
Probability modelWhite Sox, 52% Incorrect
FinalGuardians 9–5
GameBraves – Dodgers
Check-based pickBraves Incorrect
Probability modelDodgers, 52% Correct
FinalDodgers 4–1
GameBrewers – Padres
Check-based pickPadres Incorrect
Probability modelBrewers, 52% Correct
FinalBrewers 3–1

Code limits what the AI can change.

An AI agent reads the day’s injury and lineup news, and code limits how far that news can move a probability.

News limits

News can change a close pick, within a fixed limit.

In the regular season, news can move a win probability by at most 3 points. In the postseason the limit is 8 points, because one missing player matters more in a short series.

0%50%100%

Cleveland on game day: 56%, with room for news to move it from 48% to 64%. No news was found.

of 26 final picks had news
14
of 26 final picks had news
were adjusted
3
were adjusted
pick changed sides
1
pick changed sides

One week of tuning

Settings change only when the improvement is larger than random variation.

Every Monday, code tests the model’s settings against nearby alternatives on games it has not seen, and keeps a record of every decision.

Best alternative’s improvement0.00107
Improvement needed to change0.00144

The best alternative fell short, so the settings stayed the same.

games tested
493
games tested
alternatives tested
26
alternatives tested
settings changed
0
settings changed

Add the data your team already pays for.

Clairvoyint connects licensed feeds, internal databases, and APIs to the Bundle, and the same checks and models run on them.

  • Licensed lineup feedProbable starting pitchers for every game
  • Official league scheduleReal first-pitch times for the weather check
  • Ballpark orientation dataWind direction relative to each park
  • Full game schedule feedA separate pick for each game of a doubleheader
  • Team-tagged news feedNews matched to the right club every time

Use the same method for your own sport or market.

Connect your own schedules, results, and news sources to the Bundle. It runs the same checks, models, and AI analysts on them, and keeps scoring every pick.

  1. 01

    Choose your data

    Start with Wikipedia schedules and scores, Open-Meteo forecasts, and a daily news search, or connect your own feeds. The calculations and views stay the same.

    Wikipedia · Open-Meteo · web search

  2. 02

    Document how your team decides

    Set the check thresholds, how far news may move a probability, and when settings may change. The calculations run as code you can read, rather than as an estimate from an AI model.

    Matchup checks · news limits · tuning rules

  3. 03

    Get picks that cite their sources

    Daily picks with written reasoning, a track record scored against final results, and a log of every weekly settings decision.

    Timeline · track record · tuning

data sources
3
data sources
pick models
2
pick models
AI analyst agents
2
AI analyst agents
dashboard views
3
dashboard views
Early access

Use this Bundle with your own data.

Tell us which sport, league, or market you need to analyze. Join the beta and we’ll help you set up the Bundle with your data.

Examples use public, modeled, or fictional data. Results are research and are not investment, trading, betting, or operational advice or forecasts.