The claim

Is it really the most advanced baseball simulation ever built?

We say it on the homepage, so it is fair to make us defend it. Here is the case, and then here is where the answer is no.

A games outlet used that phrase in a headline before we did, which was flattering and also not evidence. So set the headline aside. The useful question is narrower and answerable: does more of what happens in a real baseball game get modeled here than anywhere else?

On the engine, we think the answer is yes, and every claim below traces to the code that produces it. On the product as a whole, the answer is more complicated, and that section is at the bottom rather than buried.

The case

  1. A pitch is a confrontation, not a die roll

    Thirty-three separate inputs decide a single pitch. Eighteen choose which pitch gets thrown, thirteen decide what it physically does, and two more, the catcher and the park, get the last word. The leverage input alone is computed from eight game-state variables. Most simulations in this category resolve a plate appearance by rolling against a rating and consulting a table.

  2. Pitches have real shape

    Twelve pitch types, each with its own velocity band and its own vertical and horizontal break, landing in the same thirteen zones Statcast uses. Command decides how far the ball scatters from where it was aimed, and the spread between pinpoint and wild is roughly four to one.

  3. Contact is modeled the way it is measured

    Every ball in play carries an exit velocity, a launch angle, a spray angle, a distance and a hang time. Hard-hit is 95 miles an hour because that is the definition, not a tuning knob. The model is calibrated against roughly 1.1 million real batted balls from 2015 to 2024, and a dedicated check refuses the build if the league that comes out does not resemble the league that went in.

  4. Nine fielders actually stand somewhere

    Defense is not a team rating nudging a batting average. Nine fielders occupy real coordinates and each carries his own range, arm and error. The same ground ball is an out or a single depending on who is standing there.

  5. Players are people, and it shows up in the box score

    Sixty-six named ratings per player before his arsenal, including eight personality traits that move real outcomes. A smart pitcher and a smart catcher execute more of the game plan. A greedy one costs more in free agency. Morale runs off seven drivers, and a player far enough underwater asks to be traded.

  6. The park and the forecast have opinions

    Every ballpark carries its own outfield distance and wall height in seven directions, its own surface, and whether it is open, domed or retractable, with hand-specific effects so it does not treat lefties and righties identically. Twelve months of real climate sit on top, so the ball carries differently in July than in April.

  7. You are not allowed to know everything

    Scouting is fogged. You see an estimate with a confidence band that tightens the longer you watch a player, not a true rating. That holds inside your own organisation and holds harder everywhere else, which is what makes a trade a decision rather than an arithmetic problem.

  8. The front office is populated

    Seven owner archetypes who authorize very different payrolls on identical revenue, five manager archetypes, eight coaching seats, and rival general managers who each value the same player differently.

  9. The history is real and it is deep

    A hundred and fifty-five seasons of Major League history, 1871 to 2025.

Every figure above is traced to the source that produces it, and the full path of a single pitch is broken out on the methodology page.

Where the answer is no

A claim like this is worth nothing if it only comes with supporting evidence. Four honest qualifications.

It is a beta, and betas have holes

The invite-only beta opened on 2 August 2026. Multiplayer leagues arrive this autumn. If you are looking for a decade of accumulated features and community tooling, the established franchise sims have a twenty-year head start and it would be silly to pretend otherwise.

The interface is young

We think it is already clearer than what this genre has settled for. We also know it has fewer screens, because there are fewer things in it. Depth of interface is not the same as depth of engine, and we are honestly ahead on one of those.

Some systems are built and not yet switched on

Several of the deeper models exist in the engine behind feature flags and are being turned on deliberately rather than all at once, because a system that is on before it is calibrated is worse than one that is off.

We have not been stress-tested by thousands of people yet

That is the actual point of the beta. The people most likely to find where this claim breaks are the ones who have played this genre for twenty years, which is exactly who we are asking in.

So, the honest version

The most advanced baseball simulation engine ever built: we will defend that, in public, against anybody, with the code.

The most complete baseball product in this genre: not yet. The incumbents have twenty years of surface area and a lot of it is good. What they have not done in the last decade is move, and that is the gap we are in.

If you find a place where this page overstates it, tell us and we will change the page.

How the engine works Come break it