Trust · auditable
Methodology
Educational radar — not investment advice. We publish how the score is built so you can distrust us productively.
One public score
The only number we call Score on the product surface is the charts engine
mechanical posture score (primary_score / score.final, 0–100).
Sub-indicators may exist in the engine payload for research, but the UI must not show multiple “Composite” scores that disagree.
Mechanical posture ≠ direction
primary.action is a mechanical posture gate (Full / Build / Probe / Wait / No / Put-bias),
not a “go long / go short” direction order. Summary text must match that primary — no “buy” next to “Wait & Watch”.
PUT ≠ sell. A Put / hedge bias label means defensive posture in the mechanical system — it is not a broker sell order and not an invitation to short.
What the score is (and is not)
- Is: a deterministic technical posture of the current setup given OHLCV + environment inputs.
- Is not: a forecast that the stock will rise tomorrow, or a guarantee of edge.
- Same inputs → same score (regression-tested in the engine).
Gates & primary action
Market → sector → stock context feed a single primary action. Gate lamp colors follow server
gate.*.status (pass / warn / fail / unknown). Unknown is not pass — we fail closed on bad data.
Data quality
If volume fields are zero or missing, we withhold volume narratives. If options data is simulated or unavailable, options are marked non-actionable. Unknown tickers fail closed — no fabricated buy score.
Demo vs Pro (supporter until mount)
Demo chips (e.g. INTC) load frozen artifacts — zero credits, no live market API.
On today’s production host, Pro is a supporter plan ($29/mo · $249/yr): higher limits now;
live analyze auto-unlocks when /health.charts_status == mounted — not sold as available before that.
Pattern “success rates”
Pattern labels and historical efficiency ranks, when shown, must cite the data series and period. We do not invent social proof (user counts) or fake testimonials. Sparse operator track notes (when published) live on /track — outcomes stay null until verified. Calibrated private edges publish only after verification.
What we do not publish
We do not ship reproducible formula weight tables, on-disk analysis paths, or host filesystem layout in the public API. The desk shows one mechanical score, gates, and honest freeze labels — enough to audit the decision, not enough to clone the private engine.