Hermēs DocumentationCheck the Data Time Before the Signal
Distinguish quote time, exposure time, session date and replay time, including missing and stale data.
A page refresh does not guarantee a new source observation. Price, exposure and open interest can update on different schedules.
Inspect the timestamps, session dates and status messages actually visible in each panel. In replay, also record playback time. Panels do not expose identical metadata: if source time is unavailable, record it as unverified rather than substituting browser time.
| Field | Establishes | Does not establish |
|---|---|---|
| Quote time | Time of the displayed quote | Every Greek is equally fresh |
| Exposure time | Time of that structural observation | Complete visibility into actual dealer inventory |
| Request / refresh time | The page attempted a retrieval | The response contains newly generated data |
| Historical session | The trading day selected | An arbitrary minute-level snapshot |
| OI date | Date of the position statistic | Every intraday trade is already reflected |
When a heatmap stops changing
- Check historical dates, replay and delayed previews.
- Check whether the underlying dataset is within its update session.
- Inspect timestamps and errors, not only the moving price.
- Separate unchanged values from stale observations; leave the result unresolved if evidence is missing.
- Retain the panel, symbol, settings and times for reproduction.
Four different states
- Zero: a numerical zero from a verified valid response and definition.
- No results: no rows for this request, not proof of no market activity.
- Unsupported / unauthorized: this symbol or account cannot access the capability.
- Failed / stale: retrieval failed or data is old; neither establishes absence of risk.
Some panels fall back to a recent populated session outside trading hours or after empty results. Verify the displayed session before comparing panels.
Exercise
Compare a live and a historical panel. Mark whether each source timestamp is verified. Preserve unknowns instead of guessing. Continue with the six data checks.