Proof of Performance

Each signal receives a composite score (0–100) based on weighted factors: technical alignment (chart structure, support/resistance, trend direction — 35%), risk management quality (stop placement, R:R ratio — 25%), market context (volatility regime, volume, funding rates — 20%), and provider track record (historical accuracy, consistency — 20%). Scores update in real time as market conditions change.

Outcomes are resolved by comparing the signal's entry, TP, and SL against actual exchange price data from Kraken and Binance. A trade is marked HIT_TP if price reached the take-profit level, HIT_SL if stop-loss was hit first, or EXPIRED if neither was reached within the signal's TTL (default: 72h for futures, 168h for spot). An automated inference pipeline runs every 6 hours to resolve pending trades.

All return calculations include estimated trading costs: 0.1% round-trip fees (maker/taker average) and 0.05% slippage per side. Leverage is factored into R-multiple calculations but not compounded — each trade is treated as an independent position. No reinvestment or position-sizing assumptions are made. Drawdown is calculated as the peak-to-trough decline in cumulative R over a rolling 30-day window.

Every signal is SHA-256 hashed at ingestion time. The hash covers the signal's content (pair, direction, entry, TP, SL, provider ID, timestamp). Once recorded, the hash serves as a tamper-proof receipt — any modification to the signal would produce a different hash, detectable by anyone who verifies it. Hashes are stored alongside the signal record and are publicly queryable via the verify endpoint above.

This is not a backtest. All data reflects live, forward-looking signal tracking. However: exchange price data may have brief gaps during outages; the 0.1%+0.05% cost assumption may understate costs on illiquid pairs or during high-volatility events; signals with missing TP/SL fields are excluded from R-multiple calculations; and the inference pipeline's 6-hour resolution interval means some fast-expiring trades may be resolved after the fact. Sample sizes for newer providers may be too small for statistical significance — we flag these with warnings.