Use our models, or build your own.
Start from a real projection model and make it yours. Bounded sliders let you lean it harder on recent form, an opponent matchup, the venue, the weather, or the market. You then save the version and replay it against history to see how it lines up with the house model.
Inputs you tune
Tune a real model, then see how it grades
Bounded sliders
You adjust how much the model leans on each category, including recent form, opponent matchup, venue, weather, and market gravity. Every slider stays inside a sensible range, so you personalize a model without breaking it.
Save your versions
You can keep more than one build, such as a form-heavy model, a matchup-heavy model, and a market-anchored one, and each saved model is yours to revisit and refine.
Replay against history
You run a saved model back over past games and grade it the same way the house model is scored, which shows how your weights would have lined up with real results.
Leaderboard vs. the house
Your model lands on a replay-graded leaderboard next to the house row, so you can see where your build ranks against the house model and against other members' ideas.
Categories, not the recipe
You shape the weight on named categories of input, while the formulas and thresholds underneath stay the model's own. You steer it rather than rewrite it.
Starts with strikeouts
The first builder is built around the strikeout model, and more markets follow as the same idea carries across the sports we price.
Your read on the game gets graded honestly
Everyone weighs a matchup a little differently. Maybe you trust recent form more than the season line, or you think a certain venue and weather combination moves a game more than the market gives it credit for. The builder lets you turn that instinct into an actual model by adjusting how much weight each category of input carries, and you start from a real projection rather than a blank page.
The important part is what happens next. A saved model gets replayed against history and scored the same way the house model is, then placed on a leaderboard beside it. That is what keeps the exercise honest, because instead of arguing about whose read is better, you can see how a set of weights would have lined up with games that already happened, side by side with the house version.
One honest caveat applies here, the same one we hold the house model to: a strong replay score is backward-looking context, not a promise. It describes how a model would have graded on past games, and any model can be wrong on a given day. Bounded sliders and replay grading exist so you can learn and compare, not so any build arrives with a guarantee. See how our models work for the categories every model shares.
Building your own model, answered
What does it mean to build your own model?
You start from a real, working projection model, the same kind that powers our boards, and you adjust how much weight it gives to different families of input. Bounded sliders let you lean a model harder on recent form, or on the opponent matchup, or dial the market's influence up or down, and then you save that version as your own. You are shaping an existing model rather than writing one from scratch.
Which inputs can I adjust?
You tune the weight on category-level inputs, which include recent form, opponent matchup, venue, weather, and market gravity. The builder names the categories and does not expose the underlying formulas, and every slider stays inside a bounded range, so a model can be personalized without being broken.
Why are the sliders bounded?
Bounds keep every custom model in a sensible, gradeable range. Rather than letting a single input swamp the projection, each slider moves within limits, so what you build stays a real model you can compare fairly against the house version and against other members' models.
What is a replay-graded leaderboard?
Once you save a model, it can be replayed against historical games, scored the same way the house model is, and ranked on a leaderboard alongside the house row. It looks back at how a set of weights would have lined up with real results, so you can see how your version compares, and it is not a promise about future games.
Is a custom model that grades well a guarantee?
No. A replay score describes how a set of weights lined up with games that already happened, so it is context and a way to compare rather than a prediction. Any model, yours or the house one, can be wrong on a given day, which is why the builder is framed around comparison and learning rather than guaranteed outcomes.
Can I save more than one model?
Yes. The idea is to try different ideas, such as a form-heavy build, a matchup-heavy build, and a market-anchored build, save each one, and see how they stack up against each other and the house model on the leaderboard. Saved models are yours to revisit and adjust.
Which sports and markets does the builder cover?
It is coming to members starting with the strikeout model, and the same tune, save, and replay idea carries to further markets across the sports we price as those builders come online. Every version is graded the same way, whichever market it is built on.
Be first to build your own
The custom model builder is coming to members. Start with the live boards now, and you will be set the moment it opens up inside your account.