Write down the rule first. Then test it on new observations.
Hypothesis and validation
Past moves help form a hypothesis. Testing requires entry, invalidation and exit conditions defined in advance, followed by new signals and outcomes.
Trial mode
A strategy can record model signals without real trades. This is called shadow mode. Its outcomes are kept separate from executed buys and sells.
What counts
- Signal count and unique tokens.
- Losses, missing quotes and delays.
- Fees, liquidity and execution assumptions.
- Observation duration and data coverage.