One controlled BTCUSDT dataset, from source to model-ready features.
Downloads the last 7 complete UTC days (10,080 one-minute candles), stores them safely, validates every minute, and calculates model inputs.
The first run can take around 20–60 seconds.
Checking…
Run the pipeline to see candles and indicators.
Adds causal Inversion Fair Value Gap state, creates 15-minute profit outcomes, trains on Days 1–3, then runs and retrains across Days 4–7.
No experiment yet.
Actual completed candles showing a first retest or confirmed rejection, with the zone state known at that minute.
Run the model experiment to calculate IFVG history.