Track Record
Every prediction, every outcome, and the methodology behind the score. Read the rules or view Desk setups.
Decision board
Verdict, sample, and score gate in one board.
Directional hit rate excludes pending, no-bet (NEUTRAL), and in-deadzone (<2.5% over 7d) moves. Small sample — below ~30 directional calls it is not statistically distinguishable from a coin flip.
Live calls grade after 7 days. BUY/SELL prints only when layers agree; NEUTRAL is a no-bet state.
Calibration report card
Brier score and reliability bins — the machine-readable half of public proof.
Calibration loads from the public report card. If this stays empty, the ledger is still maturing.
Call history
Recent resolved calls. Full history stays available below.
| Date | Call | BTC at call | 7d move | Result |
|---|
Independent research calls
Secondary model-generated theses, published before their outcomes and graded separately.
These are experimental cycle and trade hypotheses. Use Today’s Read for the deterministic portfolio decision; use this record to judge whether the research engine earns a place in that decision.
Loading current research calls…
Field audit
Nine forecast fields ranked by trade-grade evidence.
| Field | Resolution | Correct | Total | Accuracy | Grade |
|---|---|---|---|---|---|
| Loading per-field accuracy… | |||||
Trade-grade = n≥5 graded calls and accuracy ≥60%. Click a row to see that field's last 30 calls below.
Show recent calls for selected field
Methodology
Scoring rules and risk-ratio math.
How outcomes are scored
Every prediction is logged at issue-time and graded 7 days later. A directional call (BUY / SELL) is correct if BTC moved in the predicted direction by ≥2.5% over the 7-day window (the deadzone tracks BTC's weekly volatility floor). Moves inside ±2.5% are inconclusive. NEUTRAL calls are graded separately — by design, NEUTRAL means "the layers conflict, don't bet" — and a flat or sideways week confirms the framework's read. The 7-day horizon matches the cycle intelligence timeframe. Full methodology.
Sortino & Calmar — risk-adjusted quality
Hit rate alone doesn't separate "right and lucky" from "right and consistent". The risk-adjusted ratios treat each directional call as a 7d position: signed return = pct_change_7d × sign(direction); NEUTRAL calls are excluded because they're "don't bet". Sortino = annualized mean return ÷ annualized downside deviation (penalizes only losses, not upside variance — preferred for crypto by institutional allocators). Calmar = annualized return ÷ max drawdown of the equity curve. Both are shown only once n≥10 directional positions (BULL/BEAR) — in-deadzone (sub-2.5%) calls count here because their realized 7d return is real, even though they're excluded from the hit-rate sample. Below n=10 the cells display "n=X/10" so the sample is honest about itself. Standard tiers: Sortino >2 strong, 1-2 moderate, <1 weak; Calmar >3 strong, 1-3 moderate, <1 weak.