Follow one trade from the board to the monthly statement.
One illustrative order, traced through all eight agents of BETTOR, an exchange order book, a live game, settlement and the month’s waterfall, so you can see how every part of the business works together.
Generic teams · round numbers · no real dates · not a real trade
One order, every handoff, in order.
BETTOR is eight AI agents. Each one does one job, and nothing moves between them without a logged handoff. Then the venue runs the game and settles it, and the result rolls into the month. Each step shows what the desk sees and what it decides.
IllustrativeThis is a teaching example, not a trade the desk made. The teams are generic, the numbers are round and there are no real dates. Venue fees are left out of the arithmetic for clarity. Nothing here is a return, a projection or a performance claim.
-
Scout · Research
Tonight’s board.
It is 17:02 on a game night. Dozens of sports markets are listed across the desk’s connected venues. Scout scans every one and keeps the ones with real depth in the book that settle tonight.
It flags a handful as candidates. We follow one: Away at Home, an NBA-style game with tip-off at 19:30, on the market “Home to win.” It is listed on both live venues, Polymarket and Kalshi.
- Game
- Away at Home
- Market
- Home to win (YES)
- Tip-off
- 19:30
- Listed on
- Polymarket, Kalshi
-
Modeler · Pricing
The model’s price against the market’s.
A contract is a claim on one outcome. This one pays $1 if Home wins and $0 if it does not, so its price in cents reads as a probability. Buyers are bidding 61¢ and sellers are asking 63¢. The market is saying about 62%.
The Modeler prices the game with the CHUNKER read: Home wins about 64% of the time, so the contract is worth about 64¢. Bought at 62¢, that is about 2¢ of expected value per contract. Bought at the 63¢ ask, about 1¢.
- Book
- 61¢ bid · 63¢ ask
- Market says
- About 62%
- CHUNKER fair value
- 64¢
- Edge at 62¢
- +2¢ per contract
-
Challenger · Adversarial review
Then it tries to break the case.
Before anything is sized, the Challenger attacks the Modeler’s case on the record. Its only job is to find the reason the model is wrong.
Has injury or lineup news landed since the model ran? Is the line moving against us? Is any data stale or missing? Does this overlap with positions the vehicle already holds? Here all four checks hold. If one failed, the order would stop here and the reason would be logged.
- Checks run
- 4
- Result
- The case holds
-
Risk · Hard limits
The written limits, enforced in software.
Risk applies the vehicle’s hard written limits before anything is sent: a cap per order, per market and per venue, and a daily loss limit. They live in code, not in a judgment call on the night.
This game has room under every one, so the most this order may put at risk is $25,000. The limits shown match the sample statement and are illustrative.
- Per order
- $25,000
- Per market
- $100,000
- One venue
- 65% of capital at most
- Daily loss
- $75,000
-
Sizer · Position size
How much, then the caps.
The Sizer works out a size from the edge and the price, in the style of fractional Kelly. On a $5,000,000 sample vehicle, full Kelly on a 2¢ edge at 62¢ would put about 5.3% on one game. The desk uses a fraction of that: a quarter Kelly is about $65,800.
Then the caps apply. The $25,000 per-order limit is lower, so it wins. The order becomes 40,000 contracts at 62¢, or $24,800 at risk. That is the most it can lose, and it is known before the order is sent.
- Quarter Kelly
- About $65,800
- Per-order cap
- $25,000
- Order
- 40,000 contracts at 62¢
- Most it can lose
- $24,800
-
Router · Venue and style
Where to trade, and how.
Both live venues list the market. Polymarket shows the better bid and more depth near the top of its book, so the Router chooses it. Novig and Fanatics Markets are integration pending, so they are not options yet.
Then the style. Taking the 63¢ ask right now pays the spread and leaves about 1¢ of edge. Resting a bid at 62¢ as a maker earns the spread instead and keeps about 2¢. Tip-off is more than two hours away, so the Router chooses to rest.
On most markets the desk quotes both sides, a bid and an ask, and earns the spread from whoever crosses it. Here the model leans toward Home, so the bid is the side that matters.
- Venue
- Polymarket
- Style
- Maker, resting bid at 62¢
- Resting
- About +2¢ per contract
- Taking
- About +1¢ per contract
-
Executor · Execution
Rest, fill, confirm.
The Executor places the bid at 62¢. It becomes the best bid on the book, and the spread narrows from 2¢ to 1¢.
Sellers who want out now sell into it. The fills land in three pieces over about seven minutes: 15,000, then 15,000, then 10,000 contracts. While the order rests, the Executor manages it, and would cancel or replace it inside the limits if the book moved away.
- Bid placed
- 17:02:21
- Fills
- 3
- Filled
- 40,000 at 62¢
- Done by
- 17:09:47
-
The game · At the venue
Tip-off to the final buzzer.
From 19:30 the contract trades live with the game, and its price swings with every run. In this example the desk simply holds the position to the end, and Risk keeps watching the limits the whole time.
The final buzzer goes at 21:51. Home 112, Away 107.Away 109, Home 104.
- Tip-off
- 19:30
- Final
- 21:51
- Final score
- Home 112, Away 107Away 109, Home 104
- Position
- Held to settlement
-
Settlement · Escrow
Paid on the verified final score.
Every contract is fully collateralised. The desk paid 62¢ for its YES side, and the NO side of the same contract is backed by the other 38¢. The venue holds the full $1 in escrow until the result is verified under its published rules. It is an exchange, not a bookmaker: it has no side in the game.
Home won, so each of the 40,000 contracts pays $1. The position made $15,200. Had Away won, each would have paid $0 and the loss would have been $24,800, the amount fixed at entry. Both outcomes were always possible.
Away won, so each of the 40,000 contracts pays $0. The position lost $24,800, the amount fixed at entry and no more. Had Home won, it would have made $15,200. Both outcomes were always possible.
So why trade it? At a fair value of 64¢, buying at 62¢ is worth about 2¢ per contract on average: 0.64 × 38¢ minus 0.36 × 62¢. That is about $800 on 40,000 contracts, before fees. Any one game can go either way. The small edge, plus the spread, only shows up when the same discipline repeats across thousands of short games.
See it settle- If Home wins
- +$15,200
- If Away wins
- −$24,800
- Expected value
- +2¢ per contract, before fees
- Settled
- 21:58
-
Reconciler · The record
Matched to the venue’s own ledger.
The Reconciler checks every fill against the venue’s own records: the order, the fills, the quantity, the average price, the cost and the payout. Here they match, line for line.
Then it writes the audit trail: every handoff from Scout to Executor, timestamped, in one record. If anything ever fails to match, the desk halts until it does.
- Lines matched
- 6 of 6
- Reconciled
- 22:00:15
- Handoffs logged
- 7, across 8 agents
- Payout matched
- $40,000.00$0.00
-
The statement · The month
One line in the month.
The reconciled result joins every other trade the vehicle settled that month. This one adds $15,200.This one takes away $24,800. On its own, one trade says little. The month is thousands of them.
At month end, settled profit runs through the waterfall in a fixed order. First, the partner’s 10% a year preferred return, accrued monthly, is paid from settled profits. If a month falls short of it, the shortfall carries forward. Any losses are recouped before any split. Only then is the remainder split: 80% to BettorToken, 20% to the partner. There is no management fee.
The preferred return is a distribution priority, not a guarantee.
Open the sample monthly statement (PDF)- Paid first
- 10% a year preferred return
- Before any split
- Losses recouped
- Then
- 80% BettorToken, 20% partner
- Management fee
- None
Now run it thousands of times. Every edge, settled by morning.
The same flow runs on every game the desk trades, on both live venues, from the first evening tip-off into the night. Most positions open and close inside a few hours, so capital comes back for the next slate instead of sitting in long-dated bets.
Any one trade can lose. The business is the discipline, repeated across thousands of short-lived games: each one priced, challenged, capped, sized, routed and reconciled the same way.
of the desk’s trades settle within roughly five hours.
Thirty minutes on the live trading floor · Tim Hastings and Dan Reynaud join every call
This page walks through a single hypothetical order for explanation only. The teams, scores, prices, sizes, limits, times, order books and the sample vehicle are illustrative and do not describe any actual trade, position or result; the night illustration is not desk data. Venue fees, which vary by venue and order type, are left out of the arithmetic for clarity. Nothing on this page is a return, a projection or a performance claim. The Prime Partnership is presented for discussion purposes only and is not an offer to sell or a solicitation to buy securities or any other instrument; structure and terms are set only in definitive documents, reviewed by counsel for both parties. Any preferred return describes a distribution priority payable from settled profits of the vehicle: an order of payments, not a guarantee of return or of capital. The settlement-timing figure is a company statement about the desk’s recent trading and may change. Financial records are maintained internally and are not audited. Prediction-market trading involves substantial risk, including the risk of total loss. Past performance does not predict future results. BettorToken Holdings LLC is not registered with the U.S. Securities and Exchange Commission or the U.S. Commodity Futures Trading Commission in any capacity.