SIGNALS / methodology
A visible method. A bounded claim.
Our first detector describes a past price change on one named source. It does not forecast prices or tell you to trade.
What the detector measures
Up to one hour of timestamped observations. At least three distinct prices covering 15 minutes are required. A draft needs a change of at least 3% from the first price, sustained across the last two observations in the same direction. The actual observed period is shown.
When it stays silent
Prices must be positive, finite numbers. Conflicting prices at the same timestamp, future dates, a short window, a one-point spike or a latest observation older than five minutes are withheld. A feed interruption does not become a market signal.
Evidence and interpretation
Each draft carries the source URL, rights reference, timestamps, method version and an evidence hash. Sample coverage is an evidence indicator, not an accuracy or confidence probability. The next move is unknown.
History and outcomes
Stable identifiers prevent repeat processing. Later matched observations can describe subsequent price movement. This is not a trading return and excludes costs, slippage and investment decisions. No win rate is established.
Publication
The algorithm produces drafts. Source rights, financial-promotion scope, founder review and fresh data must be cleared before publication or notification. Production gates are currently closed.