Getting started
Evaluating signals
Understand coverage, available history, and how to evaluate the feed.
Eclipse provides indicator outputs you can evaluate as inputs to your own strategy. A proprietary model or a high agreement score does not, on its own, establish a profitable edge.
Regime filtering
Compare a strategy with and without a directional market filter.
Candidate ranking
Use indicator agreement to prioritize assets for further research.
Signal confirmation
Test whether an additional indicator improves an existing entry rule.
Check the available history
Universe events are retained for up to 365 days; alerts for up to 90 days. These are retention limits, not a promise that the full period is available. Read history_starts_at for each tier in the public statistics response before planning a study. That date marks the earliest retained event, not proof of uninterrupted coverage for every asset.
The catalog lists configured coverage. Statistics report event counts; they do not measure returns or prove that every reading is fresh. The sample endpoint contains static examples and is suitable for checking your parser.
Make the comparison reproducible
- Define the strategy, asset universe, holding period, and baseline before comparing results.
- Separate development and evaluation periods chronologically. Keep the evaluation period out of parameter tuning.
- Account for when the signal became available. The reported price is not a guaranteed executable fill; include execution delay, fees, and slippage.
- Compare drawdown, turnover, and results across different periods. Do not treat a short sample or a large number of correlated signals as independent evidence.
- Record the data you actually received and any gaps. Validate behavior with paper trading before connecting an execution system.