Evidence before estimates

How Rumble Deals estimates video game value

Every sold estimate uses one stored condition bucket. Listings are classified from seller titles, which can be wrong or incomplete; check the visible sale examples before using a median. When a bucket has no sold evidence, we show a labeled current-listing fallback instead of inventing a market median.

Catalog coverage: 240 tracked collectibles, 15 new systems, and 11 refurbished systems. The all-results view includes all 266; sold-comp estimates apply to collectibles, not the system shopping links.

Current source boundary

eBay completed and sold listings are the only source currently baked into sold-comp summaries. Current listing signals arrive through the Rumble Deals catalog pipeline. Amazon, eBay, Walmart, Best Buy, GameStop, Target, Costco, and Newegg may appear as shopping links, but those links do not automatically become sold-comp evidence.

Refresh schedule and timestamps

Catalog and estimate fields are baked during the site build. Sold comps are refreshed in explicit batches through either a visible Brave browser session or Firecrawl, then committed as a dated snapshot. The process is not real-time. Every game page shows the catalog refresh date, and each comp entry retains its own fetch timestamp.

What qualifies as a comparable sale

A listing must contain every significant title token after stopwords and Roman numeral equivalents are normalized. It must also be consistent with the product platform and expected US market unless the product name explicitly identifies another region or variant.

Titles naming clearly incompatible regions or platforms are rejected. The current rules include Super Famicom, Famicom, SFC, JPN, Japan, NTSC-J, PAL, EUR, Mega Drive, and Disk System markers where they conflict with the tracked product.

Condition, grader, and variant normalization

WATA, VGA, CGC, PSA, UKG, and generic graded markers route a listing to a graded bucket before sealed wording is considered. A title such as “VGA 85 Factory Sealed” should be graded. Factory sealed, brand new, unopened, and new-in-box markers can enter the raw-sealed bucket when no grading marker is present. Loose, cartridge-only, disc-only, CIB, complete-in-box, and unboxed markers enter the loose or CIB bucket.

The September 21 snapshot predates the UKG and generic graded-marker fix, so a few graded titles may still appear in its raw-sealed bucket until a full comp refresh. Inspect the listed examples and report a mismatched condition.

Clearly labeled lots, bundles, reproductions, replacement packaging, box-only pieces, posters, manuals, and reseals are excluded as non-comparable evidence. Unknown condition titles are not forced into a tier.

Outliers, medians, and visible evidence

Each tier is summarized independently. For four or more prices, a 1.5 times interquartile-range rule removes extreme values from the median calculation. The published low and high retain the observed tier range, and the sample titles shown on the page come only from that same tier.

Sold medians and current listing prices are never averaged. A valid sold-comp median leads. A current listing becomes the estimate only when the matching sold tier is empty, and the page labels that fallback as low confidence.

Confidence calculation

  • High: at least five matching comps and an observed range no greater than 50 percent of the median.
  • Medium: at least three matching comps that do not qualify as high confidence.
  • Low: one or two matching comps, or a current-listing fallback.
  • Unavailable: no matching sold evidence and no current listing price.

Small samples and corrections

A tier with fewer than three sales carries a visible low-sample warning. An empty tier stays empty, even when another condition has expensive sales. If you spot a bad comp, include the game URL, the listing title, and why the platform, region, variant, or condition is wrong.

Report a data issue

Commercial relationships

Rumble Deals may earn a commission from qualifying Amazon and eBay marketplace actions. Affiliate relationships do not change tier membership, sold medians, confidence, or sorting rules. Other retailer comparison links are currently untagged. Read the affiliate disclosure.

How we use AI

We use AI tools to help research, draft, and maintain parts of this site. A person reviews what gets published and we fix errors when we find them, but AI-assisted content can still be wrong or out of date, so confirm the details that matter on the marketplace's own listing before you buy. We do not publish AI-generated reviews or testimonials. No AI system reads or processes anything you submit to us.