Sports trading,
reimagined.
A quantitative research lab that treats sports like a market. We ingest everything, test it continuously, and trade only what survives.
What is this?
A research lab, not a tip service. We model sports the way a trading desk models any other market — find where the posted price disagrees with ours, and take the difference.
See how it works →Does it make money?
Some of it does, and we show you all of it. Our best signal returns 17.8% over 42 settled positions. 8 of our 12 tracked signals don't clear the bar, and those are on the same chart.
See the full record →How do I get in?
The research is free and always will be. The live desk — tonight's slate, our number against the market, and which signals are firing — is invite-only while the book stays small.
Enter an invite code →Graduation bar: ROI at or above 6 percent on a sample of at least 40 settled bets. Whiskers are 90 percent confidence intervals on return.
Every signal we track, ranked by return at the prices we actually got — losers included. Bars are the estimate, whiskers the 90% confidence interval. Read the full ledger for the programs behind them, the lab log and the raw data.
The edge isn't in the matchup.
It's in the hours after first pitch.
A pregame price has had all week, and every sharp bettor in the world pointed at it. A live price has had four seconds. 97% of every signal we have graded fired during the game — off a velocity drop, a bullpen door opening, a score that changed what the rest of the game is worth.
The middle innings are where the market is sharpest and where we lose money. The edges live at the edges — early, before the price has caught up with what just happened, and late, when the remaining game is short enough to model precisely and the book is slowest to re-hang. We publish the negative band because knowing when not to trade is worth as much as knowing when to.
No person can watch this.
That is the entire opportunity.
A live slate is thousands of prices moving at once, each re-pricing on every pitch, possession and substitution. The window between a tell appearing and the book correcting is often under a minute. That is not a discipline problem or an effort problem — it is past what a person can physically do, which is exactly why the opportunity is still here.
The moment the game state changes, the rest of the game is simulated again from scratch and every market attached to it is re-priced. Not a refresh on a timer — a new distribution off the new state.
Hundreds of live prices across five sports, monitored simultaneously and continuously. A desk of analysts covers a handful of games; the constraint on an agent is compute, not attention.
The best live prices turn up at 1am on a west-coast game while you are asleep or at dinner. The agents are watching then, and the alert reaches your phone the moment a graded signal fires.
This is how a quant desk has traded every other market for thirty years: model continuously, size by edge, execute without deliberating. Sports are the last liquid markets where that machinery is still rare. Today the agents monitor and alert; automated execution is being built and is not yet live. We will say so here the day that changes.
How the lab runs: the agents, the loop from feed to verdict, and the pre-registration record we commit before any test begins.
Every graded signal, every open program, the dated lab log, our calibration curve — and the raw file to check it yourself.
The working paper and everything we publish. Free, no account.
We build agent systems that have to be right.
This lab is the portfolio piece. It was built for a domain where being confidently wrong costs money the same day. If you want the same thing pointed at your data, we take a small number of builds.