The One Idea
Beat the closing line — or don't bet
Here is the whole game in one sentence: you are betting against a crowd that is very, very good. The final odds on the board are the sharpest prediction in all of sports — they fold in money, workouts, weather, and thousands of opinions. If your plan is to out-guess that from scratch, you will lose. We measured it: the closing market sorts winners with an AUC around 0.77; a built-from-scratch handicapping model managed about 0.56 — barely better than a coin. So we don't fight the crowd. We stand on its shoulders.
We anchor, then nudge
Every horse starts at the market's price. Then our signal (speed, form, pace — next lesson) tilts that number a little: model ∝ market · e^(γ · signal). The knob γ is how hard we lean on our read. Drag it below and watch what happens.
Top bar = the market. Bottom bar = our model after tilting by γ. We only bet horses our number rates higher than the crowd by the overlay threshold (2%). Push γ too far and the model just bets against the market everywhere — which is exactly how you lose.
Why "winners" is the wrong scoreboard
New bettors chase win rate. It feels great to cash. But a horse you back at 8-1 that should be 4-1 makes money even though it loses most of the time, while a 60% favorite at even money quietly drains you. Our overlays hit maybe 8–15% of the time — and that's the point. We're paid for being right about the price, not right about the winner.