NoticeDomain Migration: This site is moving to predict.miqixai.store (now live). Please update your bookmarks.Open new domain ↗
SponsorSupport Free Data Hosting: Visit our self-service AI store miqixai.store ↗ (ChatGPT Plus/Pro & Gemini plans, 24/7 instant delivery)
BTCMarket DataGitHub ↗中文

DOCUMENTATION

Using a research assistant

Provide the archive README, manifest and a specific research question. Ask the assistant to inspect actual fields, duplicates and gaps, then state timestamp choices and exclusions.

Suggested workflow

  1. Inventory files, schema, row counts and market coverage before writing a strategy.
  2. Check timestamps, windows, pre-open observations and settlement labels.
  3. Show raw rows for one market and explain bid, ask, depth and information availability.
  4. State fees, latency and fill assumptions before backtesting.
  5. Save parameters, hashes and exclusion counts so the result can be audited.

Never use outcomes as a predictive feature, substitute source time for receive time, or treat a quote as an executed trade. Compare results against an explicit baseline.

Backtesting project prompt