Prototype environment

All market values are illustrative demonstration data. This site does not execute trades or connect to third-party marketplaces.

Steam MarketDATABASE
Watchlist 0
Search databaseItems, collections, and games
↑↓ Navigate↵ OpenPrototype index

Methodology & trust

Evidence first.
Interpretation second.

The product is designed to make market conclusions inspectable without presenting buy/sell recommendations or pretending observed data is complete.

01 · Principles

Describe conditions, not actions.

Steam Market Database is an information and insights product. It does not execute transactions, route users to trading venues, calculate gambling outcomes, or issue buy/sell recommendations.

Observed ≠ totalMetrics describe captured observations unless completeness can be established.
Signal ≠ recommendationA signal describes a condition relative to a documented baseline.
Inference is labeledFacts, derived metrics, and interpretation remain visually distinct.
Freshness is visibleTime-sensitive values carry timestamps and data-quality context.

02 · Pipeline

Keep every transformation inspectable.

01Raw observations

Captured market pages, listing snapshots, price-history points, and timestamps.

→
02Normalized entities

Canonical game, item, collection, container, and relationship identities.

→
03Derived metrics

Ranges, baselines, breadth, observed supply, activity, and index values.

→
04Deterministic events

Thresholded, versioned conditions with supporting observations.

→
05Context

Plain-language explanation generated only after evidence exists.

03 · Signals

A signal must show its work.

Each event should include an entity ID, event type, observation window, baseline, magnitude, timestamps, coverage metadata, confidence, and links to underlying observations.

{
  "eventType": "ACTIVITY_ANOMALY",
  "entityId": "item-dn-case",
  "window": "24H",
  "baseline": "30-day hourly median",
  "magnitude": 3.8,
  "confidence": "high",
  "observedAt": "2026-08-17T21:59:22Z"
}

04 · Data quality

Coverage and confidence travel with the data.

  • Record observation timestamps, fetch status, source identity, and parse version.
  • Separate unavailable data from a legitimate zero.
  • Detect stale listings, duplicate captures, schema drift, and partial windows.
  • Suppress or downgrade events when coverage falls below documented thresholds.
  • Expose methodology versions so historical events can be reproduced.

05 · Coverage

Model the Steam item world as relationships.

The intended entity model links games, applications, item definitions, market hash names, collections, containers, capsules, variants, observations, metrics, and events. The prototype contains a small curated dataset to demonstrate those relationships.

06 · Limitations

State what the product cannot know.

  • Observed market activity may not represent every transaction or listing.
  • Historical captures can contain gaps, delays, or source-side changes.
  • Price movement does not reveal participant intent.
  • Correlation among related items does not establish causation.
  • Prototype values are fabricated demonstrations and must not be used for decisions.

07 · Independence

No implied Valve affiliation.

Steam Market Database should clearly identify itself as an independent information service. Branding, marks, colors, copy, and product behavior must not imply that Valve endorses or operates the service.

08 · Privacy

Collect less by default.

The prototype stores watchlist state in browser local storage and contains no analytics. A production implementation should document telemetry, retention, account data, alert delivery, and deletion workflows before launch.