Skip to content

Five feature families

One-liner: Features in this repo are grouped into five families by “who is being measured”. Knowing which family a sentence belongs to tells you which YAML to look in and which court to use.

The five families at a glance

Family Who is measured Typical columns Court
Single-symbol event One symbol on one bar MA, cross, ATR, RSI Event backtest
Cross-section A basket compared every day Momentum rank, amount rank Cross-section backtest
Funding / order flow Perp funding, big orders funding, P99 big orders Event backtest
Calendar / weekday Time: day of week, day of month Monday, month-end Event backtest
Cross-symbol lead-lag One symbol leads another BTC leads AI alts Event backtest

How to use it

When writing a hypothesis, first ask: is this sentence measuring “one symbol on one bar”, or “a basket compared every day”? The former goes on the event axis, the latter on the cross-section axis. Pick the wrong axis and the backtest script is wrong.

Fine print