How the readings work

Research and the MCP share one Research API. The browser offers a visual workspace; the assistant retrieves focused evidence and can discuss it with you. New to the terminology? Start with the glossary.

Start with the move

Range interpretation identifies prominent price moves inside your selected A-through-B span. Enabled metrics provide supporting context: participation, positioning, liquidity, momentum and volatility. Findings are calculated by the shared deterministic analysis engine; there is no separate LLM request behind that calculation.

An AI host may add its own explanation. That prose is distinct from the source observations and calculated findings.

Keep the clocks visible

A candle represents its whole interval, not a point-in-time quote. Studies are calculated on the selected 1h, 4h or 1d candles by the shared Research engine. Their periods count observed candles, not a guaranteed elapsed duration; missing intervals, warm-up and source-minute coverage remain qualified. Explicit native-study requests keep their separate source clock, without automatic fallback.

For numerical snapshot and flow layers, the API supplies one reading per selected candle. Chart positions use candle openings, while observation and availability times retain the real source clocks. Snapshot readings are not sums or averages across a candle: funding remains percent per eight hours, open interest retains its matching collection cycle and venue values, and book data retains its usability checks. Flow values sum the observed periods; a partial subtotal is labelled and is not eligible as a complete-period comparison. Consumers do not recompute this alignment.

Market regime and volatility observations use their own hourly clock. They remain hourly when the price chart uses 4h or 1d candles, and their historical percentiles compare this market with up to 90 days of available prior observations—not a promise of 90 complete days per asset. Read the combined guide.

The endpoint comparison and full-range interpretation answer different questions. The first compares selected candles; the second examines the intervening observations. Market readings explains the main metrics.

Compare like with like

Cohorts use a prepared, pinned history and explicit membership controls. Group membership, common assets, weighting, sparse observations and the benchmark affect the result. Refining a comparison should reuse the same snapshot unless you ask for a fresh one. See Explore cohorts.

Published context has its own date

The Market Update preserves the released show and its chapter evidence. Video timelines use publication dates; news context identifies its returned window and source-linked sample. These are context around a market move, not proof of its cause.

Source details

Open the details attached to a reading for its units, timestamps, source and coverage. Gaps remain unknown, rather than becoming zero or a carried-forward certainty. Displayed order-book depth is a sample, not guaranteed execution liquidity.

These are research observations, not personalised financial advice or a promise of future performance. Past performance does not guarantee future results.