Polymarket Filtering Markets Explained

The honest order is filter-then-analyze: decide the inclusion rule on the metadata (series, window, resolved), then report how the flag/liquidity filters changed the sample. A sample chosen after peeking at results is a selection called a study.

The archive's market table carries the filter fields (slug, series, window, resolve second, outcome), so the whole recipe is one endpoint call plus one row filter.

Filtering on executed price (vs other coins) biases toward liquid winners; filtering out one-sided frames destroys the venue's texture; and mixing thresholds with Up/Down windows creates a sample with two different settlement clocks.

Each pitfall has a documented field: keep the coin's full series, keep the flags, and tag the window family so the sample's composition is visible in the report.

A filter is a definition, and every honest result names its filter chain: the research methods and methodology pages require it because reproducibility starts at the SELECT clause.

The two filters that change research most are window and resolution state: a 5-minute Up/Down market and a 4-hour threshold line have different mechanics, and resolved versus live markets answer different questions. Discovery through /v1/series and /v1/tags encodes both, so a correct filter is a correct market family.

That is why the page walks the filtering method on the venue's own taxonomy before reaching into the archive — a clean filter at discovery time makes everything downstream cheaper and more auditable.

• Research methods
• Current prices
• Resolved markets