How finsail reads the market
The whole method, in plain language. If anything here is unclear, that's a bug — write to us.
Read-only by design
Context and probabilities, never trade instructions. Compliance is enforced in the backend, not the UI.
Honest uncertainty
Forecasts are quantile ranges in percent space — no price targets, no certainty theater.
Auditable record
Append-only journal + daily invariant audits. The track record can't be quietly edited.
Does finsail tell you what to trade?
No. finsail provides structured market context — readings, probability fans, and commentary. We never tell you to buy, sell, or hold anything. Where a structure is identified, we describe it without levels or order instructions, by design.
What data does the engine read?
Price and volume across multiple timeframes (1h, 2h, 4h, daily aggregates). We do not use social-media sentiment. Every reading is computed from the same data a market analyst would look at — just every hour, without fatigue.
What is the fan chart?
For each reading, the forecaster produces a distribution of possible 72-hour returns — five quantiles in percent space. The outer band (q10–q90) covers roughly 80% of model probability; the inner band (q25–q75) is the more likely half. The median line is not a prediction; it is the center of an honest uncertainty range.
Why is the system silent most of the time?
Every hour the engine evaluates each asset and, most of the time, deliberately does nothing. Each rejection is recorded with its reason — expected move below cost, median against direction, confidence under threshold. Silence is the system working as designed, and we show you the reasons.
How are models promoted to live?
Candidate models are trained continuously and evaluated against probabilistic backtest-overfitting checks (PBO) and deflated Sharpe ratios (DSR), then run in shadow before any promotion. Most candidates fail. We publish the failures along with the successes.
How is the track record kept honest?
Every identified structure and its outcome is written to an append-only journal, verified by an invariant audit chain. Nothing is deleted or restated. The numbers you see on the landing page reconcile with the audit log, and audits run daily.
How are simulated and live results separated?
Wherever results originate from simulated tracking they are stored and reported separately from live results, and the two are never mixed in any metric. The separation is enforced in the data layer, not by a UI setting.
finsail llc · 1209 Mountain Road PL NE STE N, Albuquerque, NM 87110 · support@finsail.ai