comp_v2 — end-to-end sold-comp aggregates
green · 249yellow · 686red · 8,078total · 9,013
What this dataset is
comp_v2 is the deep end-to-end comp pipeline output: every card in this dataset went through the 5-source aggregator (eBay Firecrawl, Fanatics Sales History, Heritage MOTO, PSA APR, VCP free) and the condition-NLP raw-band classifier. Each row carries a confidence tier and the "why GREEN was blocked" reason on YELLOW rows so a consumer can decide whether to trust the median.
Schema fields
| Field | Description |
|---|---|
| slug | Canonical card slug |
| card_id | Stable 16-char card hash from cards.parquet |
| confidence_tier | green / yellow / red |
| green_blocked_by | Yellow-tier reason: no_raw_sold / thin_sample / single_source |
| summary.n_total | Total sold rows aggregated |
| summary.median | Median sold price (USD) |
| populations.base.graded_by_grade | Per-PSA/SGC/BGS/CGC-grade breakdown (n, median, min, max) |
| populations.base.raw_by_band | Per-condition-band breakdown for ungraded sales |
| source_breakdown | Per-source row counts (eBay Firecrawl / Fanatics / Heritage / PSA APR / VCP) |
| honesty_flags | graded-only, stale-180d, floor-price, thin-data, reprint-flag, defect-flag, lot-flag |
Download
- /data/comp_v2/_index.json — slug → {tier, blocked_by, n_total} map (compact site-wide index)
/data/comp_v2/{slug}.json— per-card JSONs, one file per card
License: CC-BY 4.0. See /data for the license terms and citation.