Methodology before marketing.

Every displayed estimate is generated by versioned decimal-arithmetic rules. Observed facts, modeled costs, assumptions, freshness, confidence, and warnings remain separate.

Spot arbitrage

Capital is walked through buy asks and sell bids. Trading fees, market impact, transfer assumptions, and a risk buffer are itemized. Stale or crossed books fail closed. Buy and sell depth coverage are measured from the observed book. Fill probability remains unavailable until real execution telemetry can calibrate it; depth coverage is not relabelled as probability.

Funding arbitrage

Native venue rates and intervals are retained and normalized hourly. Projections use a finite holding horizon and deduct round-trip fees, four modeled slippage events, and a basis-risk allowance.

Opportunity Score

The score ranks comparable observations using net return (30 points), liquidity (20), data confidence (15), freshness (10), persistence (10), venue quality (10), and execution simplicity (5). It is not a promise or probability of profit.

Persistence uses viable prior snapshots for the same route or provider during the preceding 24 hours. Repeated snapshots are deduplicated into 15-minute windows; recurrence contributes up to six points and observed duration up to four. Results also expose the measured lifetime between the first and last qualifying observations. Open “Why this score?” on a result to inspect every component and cap.

Current calculation versions: spot-arbitrage/1.1.0, funding-arbitrage/1.5.0, and yield/1.0.0. Funding compares a long-spot/short-perpetual hedge with a perpetual/perpetual hedge. The spot leg uses observed order-book depth; the perpetual leg retains an explicit slippage estimate. Funding cash flow counts only native venue settlements inside the selected horizon; it is not continuously prorated. Results expose the projection window, per-leg settlement counts and amounts, borrow-cost assumption, fee source/tier, and risk/slippage parameters. Current score version: opportunity-score/1.1.0.