csfolder
Sign in

Data methodology

How csfolder collects and calculates CS2 market data

csfolder is a market-data and portfolio product, not a marketplace. This page explains what our numbers mean, where they come from and where an estimate has limits.

Prices, offers and estimated activity

Item pages compare current public offers from supported CS2 marketplaces. “Offers available” is active supply at the end of a chart bucket, not sales. “Sold, estimate” is derived from changes in listed inventory and must not be read as an exact transaction count. The coverage label states the available window, and bars are omitted when the required data is absent.

Freshness and failure handling

Market workers refresh sources throughout the day and keep the last successful value when an upstream request fails. We do not replace a known value with zero during a temporary outage. Stale or unavailable inputs are excluded where the calculation cannot be made responsibly.

Inventory value, rankings and market cap

Inventory value is quantity multiplied by the selected source quote for every matched variant. Coverage shows how many units received a price. Rankings use current positive asks and keep one representative per finish family. Market-cap views use the supply and price fields described on the relevant page; they remain estimates, not audited exchange capitalization.

Known limitations

Base quotes may not capture sticker overpay, rare pattern index, exceptional float, private deals, marketplace fees or withdrawal friction. Currency conversion and Steam Wallet value can differ from cash value. For that reason, csfolder labels calculations as estimates and exposes source-level comparison instead of presenting a single guaranteed sale price.

Quality controls

  • Canonical item and locale URLs are generated from the same route rules used by the sitemap.
  • Public data pages fail closed when their required dataset is missing instead of publishing invented numbers.
  • Automated SEO experiments are limited to existing-page titles and descriptions, require search evidence, keep history and support rollback.
  • New editorial pages, claims, schema changes and body copy require human review; the automation does not publish generic AI articles.
  • Users can report a mismatch through the feedback page. Corrections are made in the source data or calculation, not hidden in generated prose.

Methodology reviewed: 27 July 2026.