Does This Bet Have Edge?
PREDICTION MARKETS · WORKFLOW · FRACTIONAL KELLY SIZING
Compare your probability estimate against a prediction market's price and size the bet with quarter-Kelly.
You think this is 60% likely · the market prices it at 40%· bankroll $10,000
Get this from a live market on Odds Converter or Market-Implied Odds.
Uses quarter-Kelly by default — a conservative haircut on the full Kelly criterion, since full Kelly is famously volatile bet-to-bet. This is bet sizing, not a probability estimate — it can't tell you if your own probability is actually right.
How much do you disagree with the market?
Enter your own probability estimate and the market's price — I'll tell you if the edge is worth betting on.
Does this bet have real edge?
A prediction market's price is the market's collective probability estimate. If your own estimate is meaningfully different, that gap — the edge — is what you're actually betting exists. This sizes how much of your bankroll that edge justifies staking, using the Kelly criterion: the formula that maximizes long-run bankroll growth for a repeated bet with a known edge.
The formula
Buying a contract at price p that pays $1 on YES: staking f of your bankroll buys f/p contracts, so the net-odds multiple is b = (1−p)/p. Standard Kelly f* = q − (1−q)/b simplifies to:
- f* = yourProbability − (1 − yourProbability) × marketPrice / (1 − marketPrice)
- EV per $1 staked = yourProbability / marketPrice − 1
- recommendedStake = f* × kellyFractionCap × bankroll (default cap: 0.25, quarter-Kelly)
Sanity check: if your probability exactly equals the market price, f* comes out to exactly zero — no edge, no bet, which is the whole point of the formula.
Worked example
You think an event is 60% likely; the market prices it at 40%:
- b = (1 − 0.40) / 0.40 = 1.5
- f* = 0.60 − (1 − 0.60) / 1.5 = 0.60 − 0.267 = 0.333 (33.3% of bankroll, full Kelly)
- quarter-Kelly stake = 0.333 × 0.25 = 8.3% of bankroll
- EV per $1 staked = 0.60 / 0.40 − 1 = +50%
Why quarter-Kelly, not full Kelly?
Full Kelly maximizes long-run growth in theory, but it's extremely sensitive to overestimating your own edge — a slightly-too-confident probability can turn a real edge into a bankroll-damaging stake. Quarter-Kelly (a 0.25 multiplier on the full formula) is a standard, widely-used haircut for exactly this reason — stated explicitly here as a convention, not hidden inside the math.
Get a market price first
This tool needs your own estimate and a market price to compare it against. Pull a live one from Market-Implied Odds or Odds Converter.
Use via API or MCP
Available as a deterministic API call for agents that need to translate a probability disagreement into an actual position size.
What if my probability estimate is wrong?
This tool can't validate your estimate — it only tells you how much to stake if your estimate is right. Garbage in, garbage out is the biggest real risk here, which is exactly why quarter-Kelly (not full Kelly) is the default: it limits the damage from an overconfident estimate.
Why is the recommended stake sometimes $0?
The Kelly fraction floors at zero whenever your probability is at or below the market's price — there's no mathematical edge to size a bet on, so it says skip rather than force a number.
What's the difference between "edge" and "EV per dollar"?
Edge is the simple percentage-point gap between your probability and the market's (e.g. 60% vs 40% = 20 points). EV per dollar staked accounts for the payout structure too — how much you actually win per dollar risked at that specific price, which is what the Kelly formula uses.
Can I change the Kelly fraction cap from the default 25%?
Yes, via the API (kellyFractionCap, 0.01-1). The web widget uses the standard quarter-Kelly (0.25) default; the API exposes the parameter for anyone who wants a different risk tolerance.