Public graded history
MLB Model Record
Gambletron AI runs the numbers on every MLB game. Most assessments are research; a few per slate clear the model's qualifying threshold and become Official Picks. Every Official Pick is locked before first pitch and graded against the final score. The record below is every pick since tracking began on May 26, 2026, last graded Oct 3.
Overall record
MLB Model · all bet types
44-19
+3.75u
+6.0% · 69.8%
63 graded picks
By bet type
Moneyline (ML)
37-17
+1.00u
+1.9% · 68.5%
54 graded picks
Runline (RL)
4-0
+2.09u
+52.3% · 100.0%
4 graded picks
Total
3-2
+0.66u
+13.2% · 60.0%
5 graded picks
A unit is one standard stake. ROI is units divided by graded decisions. Pushes are counted but do not move units.
How the record is kept
Scores and lineups lock roughly fifteen minutes before each game starts, so everyone sees the same published numbers rather than a figure that quietly shifts once play is underway. Once a game is graded, its final result replaces the locked numbers. Nothing is edited afterwards and nothing is dropped once published.
A model score and a market edge are two different measurements. The score is the model's view of the matchup; the edge is the gap between a sportsbook's price and the market's own fair price, which Gambletron AI calculates itself. The record above is Official Picks only — the moments where both the model score and the market context cleared the threshold.
Models live in their own repositories. Gambletron AI displays their output; pick logic never lives in the website itself. This page re-renders when the model's plays log updates.
Why we publish the rejected assessments too
Most of what the model scores never becomes an Official Pick. Showing the rejected assessments is deliberate — a record that only lists the bets a gate let through tells you nothing about whether the gate is set well. Publishing both lets a reader see when the model talks itself out of winners, which is a different problem than hitting cold on the picks it kept, and both are visible on the Betting Agents page.
What the record is, and what it isn't
It is a run of every Official Pick the MLB Model has made since May 26, 2026, with the price locked before the game and the result from the final score. It is not an audited account statement, a prediction, or a guarantee of future results. Baseball is a long season; a positive ROI over a few hundred picks still carries meaningful variance, and the right way to read this is as evidence, not proof.
For a sense of what a track record should and should not be used to tell you, see the track-record guide.
Frequently asked questions
- What is the Gambletron AI MLB Model's public record?
- A running tally of every Official Pick the MLB Model has made since May 26, 2026, graded against the final score. The table on this page shows wins, losses, pushes, units and ROI overall and by bet type (moneyline, runline, total). Numbers only, no games or picks — the pick-by-pick calendar with reasoning is on the Betting Agents page with a free account.
- What counts as an Official Pick?
- The MLB Model scores every game 0-100 using form, matchup data, starters, injuries and market context. Only games clearing the model's qualifying threshold become Official Picks. Everything below the threshold is published as research so you can see the calls the model considered and declined. Most games are research; only a few per slate become Official Picks.
- Why publish games the model declined?
- A record that only lists the bets a gate let through tells you nothing about whether the gate is set well. Publishing the rejected assessments alongside the picks is deliberate — if the model also talks itself out of winners, that is a different problem than hitting cold on the picks it kept, and both are visible.
- How is a unit calculated?
- A unit is one standard stake. Wins and losses are summed into units based on the price locked before the game — a +120 winner adds 1.20u, a -150 winner adds about 0.67u, a loser subtracts the staked unit. ROI is units divided by graded decisions.
- Is this the full record or just the highlights?
- The full graded record since tracking began. Nothing is dropped once published. The page is updated after games grade.
- What's the difference between a model score and an Official Pick?
- The model score is the model's 0-100 view of the matchup. The pick is only made when that score, combined with market context, clears the model's qualifying threshold. A game can score high and still not be a pick if the market has already priced it in. The Betting Agents page shows both the score on every game and whether it was an Official Pick.
- Is Gambletron AI a sportsbook?
- No. Gambletron AI is an informational analytics platform. It never accepts wagers, holds funds, or places bets. 21+ only. If you or someone you know has a gambling problem, call 1-800-GAMBLER.
- What about NFL, college football and other sports?
- NFL also publishes graded Official Picks — see the NFL Model Record page for the full record and the by-bet-type breakdown. NFL also has an Upset Tracker, published as research (not picks), that flags underdogs the model's signals favor. NCAAF has live research — rankings, unit matchups, matchup edges, prop notes — but no picks yet. Other sports are not currently published; picks are added only when a model's results support it.
Free account
See every pick, with its reasoning
The record is public. The pick-by-pick calendar — every Official Pick the moment it's made, the 0-100 score and the pitching, lineup and bullpen data behind it, graded day by day — is on the Betting Agents page with a free account. We don't take bets or ask for payment details.
Gambletron AI is an informational analytics platform, not a sportsbook. We never accept wagers, hold funds, or place bets. 21+ only.