Ask market history in plain language - no code
Type a question like "every time BTC fell 10% in a week, what happened next week?". AI turns it into a fixed query, you check it, and the server computes the answer on real candles: a summary next to the normal baseline, a chart and every occurrence with its date. Historical statistics, not a prediction.
Sign in to ask the data
Sign inHow it works
- 1
Write your question in any of 12 languages. AI only translates it into a fixed query - it never computes the answer.
- 2
Check "Here is how I understood your question" and edit any part before running.
- 3
The server scans real closed candles and shows each occurrence, the result after each look-ahead and the normal baseline.
These are counts of what happened in the past. They are not a forecast and not a probability that it will happen again.
Free and Pro
Free: 2 questions per UTC day, full summary and the 10 most recent occurrences.
Pro: 30 questions per UTC day, every occurrence and CSV export.
See Pro plansQuestions
Do I need to know Python or Pine Script?
No. You write a normal sentence; the result is built from the same rule engine the screener and backtester use, on real exchange and market data.
Can the AI make up numbers?
No. The AI only produces the query you see and can edit. Every number is calculated by the server from candles, and the query is re-checked before it runs.
Why is there a "normally" column?
A market that rises most of the time makes almost any event look bullish. Comparing with every bar in the same period shows whether the event really changed anything.
How far back can I ask?
Up to about 6,000 candles per question - years of daily or weekly data, months of hourly data - limited by what the data provider returns. The result always states the exact dates covered.