Benchmark Methodology

The benchmark ran continuously from 2026-07-30 21:00 UTC to 2026-07-31 21:00 UTC, polling each provider's production API on 5-minute cycles with real API keys. Single-endpoint providers received 293 polls over the window; providers benchmarked on two endpoints received roughly double. The exact poll count backing each success rate is printed beside it on every comparison page.

Pricempire was measured on its free tier, the only tier we held a key for. CSGOSkins.gg was not benchmarked at all; its comparison page shows published specifications only and its measured cells say so.

A marketplace counts only if the API returned a real, non-zero price for it during the benchmark. A price: 0 placeholder is not a price, and a marketplace listed in documentation but absent from responses is not coverage. Store and trade variants of the same venue count as separate price streams, not separate marketplaces. By this rule CS2Cap returned live prices from 40 marketplaces; the full list is on every comparison page under the coverage figure.

Freshness is the median age of the price timestamps the provider's own API embeds in its responses, taken at read time. It measures the data, not the marketing page. Two providers could not be measured this way: CSMarketCap's responses carry no price timestamps, and CSPriceAPI's responses carried future-dated timestamps during the benchmark. Both are reported as unverifiable rather than assigned a number.

Reliability is successful polls divided by total polls over the 24 hours. A retry is not a poll: each 5-minute cycle counts once per endpoint, so a provider cannot improve its score by being asked twice.

Anything a provider was not measured on is quoted from its own public pages, copied by hand on the date shown on each comparison page, with the source linked beside the claim. Where advertised and measured numbers disagree, both are shown.