How ToteFlow Works

ToteFlow combines a public probability signal with the live tote market to find horses the crowd is under-pricing — then filters aggressively so we only fire when there's real value.

US horse racing runs on pari-mutuel pools: bettors wager against each other, and the platform takes a cut of the total pool (the “takeout,” typically 15–20%). Unlike a sportsbook, the platform is not the counterparty on your bet — so they have no incentive to hide sharp signal from their customers.

That structural quirk means useful probability signal exists in the open, and the market that has to absorb it is famously noisy: recreational money, late swings, favorite-longshot bias, thin pools on smaller cards. If you can identify the right signal, calibrate it against the crowd's view, and be strict about when to actually pull the trigger, positive expected value shows up in specific pockets of the market.

At a high level, for every race on the board we:

1. Ingest live tote odds, pools, and a probability signal.
2. Decide whether that race is even worth having an opinion on.
3. Reconcile the signal with what the market is saying.
4. Ask whether any horse offers positive expected value after takeout.
5. If — and only if — one does, fire a paper ticket.

The interesting work is in steps 2–4. The exact filters, weights, and thresholds are proprietary to ToteFlow and not documented publicly.

The math of blending any conservative-ish probability signal with a takeout-taxed market means positive EV almost never appears at the top of the board. Favorites are priced roughly correctly and the house cut eats any small edge. The pockets of value that survive filtering sit further down the price ladder — and that's where our realized P&L is concentrated.

A consequence: hit rate is low by design. Most tickets lose. The strategy only works if the winners pay enough to cover long losing streaks, which is why disciplined filtering matters more than volume.

Paper trades only. All results are simulated tickets recorded at the odds available at bet time; nothing is placed with real money by ToteFlow itself.

• Low single-digit hit rate, as expected given where the value lives.
• Realized ROI is concentrated in the longer-price buckets.

At the current sample size the confidence interval on ROI is still wide. Realized P&L is the verdict, and the verdict takes hundreds more bets to become statistically meaningful. See the strategies page for live numbers.

• Everything is paper. No real money placed by ToteFlow itself.
• Any edge here has a capacity ceiling — big money into a small pari-mutuel pool moves the odds against itself.
• Upstream data sources can change or disappear at any time.
• Short-run P&L is dominated by variance. The verdict takes time.