Back to market intelligence

METHODOLOGY

Evidence, Windows & Market Memory

W63N separates provider observations from derived market-structure explanations. Every derived result keeps its data window, provider set and limitations.

Updated: 2026-07-12

Observation memory

Provider rows are normalized into UTC time buckets with unique indexes. Missing intervals remain missing.

  • Binance core spot assets: one-minute collection target
  • Bybit Linear contracts: one-minute collection target
  • Active DEX pools: five-minute collection target

Deterministic intelligence

Transparent rules compare price, volume, open interest, funding and DEX changes. The engine does not output buy or sell instructions.

  • Evidence hashes prevent needless repeat computation
  • Insufficient history produces an explicit empty state
  • Detected events are not rewritten using later outcomes

AI boundary

AI may organize and explain available evidence, but it cannot introduce a data source that the system did not collect. Cached answers remain tied to their evidence and window.

Retention and freshness

Recent high-frequency observations support replay and comparison. Older observations are pruned under a retention schedule, while source timestamps and generated briefs preserve their audit trail.

Risk boundary

Market structure can describe leverage, divergence and activity. It cannot establish wallet ownership, predict returns or guarantee future direction.

W63N Methodology | Evidence, Windows & Market Memory