Public paper book · transparent rules
How the Oracle picks markets, sets lines and sizes bets
Every number on the live P&L tracker comes out of a fixed rule set. Nothing is picked by feel. This page is the full playbook, so the results sheet reads like what it is: a disciplined process executed in public. Every price quoted here is on the live odds board.
1. Which markets get picked
Full-board scan
Before each slate, the model screens every college football winner market on the board and re-prices each game with the CutbackPredict analytics data — the same software the show credits on air.
The price band
The Oracle only trades contracts priced between 20c and 80c. Below 20c the payout is lottery-style and fees eat the edge; above 80c there is no room for the target to pay. Anything outside the band is skipped automatically.
Two reads must agree
Each game gets independent model reads. When the reads agree tightly, the signal can grade A-conviction. When they disagree, the signal is downgraded or passed entirely — disagreement is a reason to stand down, not to guess.
Passing is a position
Most games are a pass. A slow week on the ledger is the filter working, not the desk being asleep. The Oracle only posts when the numbers clear every bar.
2. How the fair line is set
For each game the model produces a fair probability — the price the contract should trade at. The edge is simply the gap between fair value and the live market price:
edge = model fair probability − market price
Entry requires at least 8 cents of edge. A game the model prices at 57c that the market offers at 49c is a candidate. At 52c market price, the same game is a pass.
When the edge is 12c or more and the two model reads agree within 8c, the ticket grades A-conviction and gets full sizing. Everything else that clears the bar is B-conviction at reduced size.
3. Entry, stop and target
Entry
The ticket is entered at the live market price the moment the signal fires — never a back-filled number. That entry is printed on the open trades list.
Stop: entry − 10c
If the market moves 10c against the position, the ticket is closed. The loss is capped at exactly the pre-committed risk — no hoping, no averaging down.
Target: entry + 12c
If the market moves 12c in favor, the ticket banks the win. If neither rail is touched, the position settles at the final game result.
4. Bet sizing — the 2% rule
Sizing is fixed-fractional. The desk risks 2% of current equity to the stop on every A-conviction ticket, and 60% of that on B-conviction. Contract count is worked out from the risk budget, so every ticket carries the same dollar risk regardless of price:
contracts = (equity × 2%) ÷ (entry − stop)
Worked example:
Equity $5,000 → risk budget $100. Entry 45c, stop 35c → 10c of risk per contract. The ticket buys 10 contracts. Target 57c pays +$120 (1.2R). A stop-out loses exactly the $100 (−1R). Settling at the final can pay more or less, but the planned risk never changes.
This is why the P&L page shows every ticket with its risk in R multiples: the question is never “how much did we bet?” but “how many R did the process make?”
5. Invalidation — when the thesis dies early
Every posted pick carries an invalidation note: the condition where the original read is wrong even if the stop hasn't traded — for example, the backed team trailing by more than a score into the fourth quarter. The Oracle says it out loud on the stream when a thesis is invalidated, because a rule you only follow when winning isn't a rule.
6. How the P&L is computed
Total P&L
Realized dollars from settled tickets, plus mark-to-market on open ones (contracts × current price − entry). Equity = $5,000 starting bankroll + total P&L.
Win rate & average R
Win rate is winners ÷ settled tickets. Average R is the mean of each ticket's P&L divided by its planned risk — the number that tells you whether the edge is real.
The Oracle trades a public paper book for education and entertainment. Nothing on this page is financial advice, and past paper results do not guarantee future performance. Prediction-market contracts involve real risk of loss — never trade money you cannot afford to lose.