🍄 LOAN-TERMS ORACLE

How much can you safely lend against a trading card? Look one up and watch the answer get built — no black box, every input verifiable on-chain.
The method, in plain English:
1 · Value. We start from the card's real market price — sales-derived, not asking prices.
2 · Tail risk. Our forecast engine publishes a calibrated 99% worst-case move for the loan term (VaR99), validated against reality daily and merkle-committed on-chain BEFORE outcomes exist.
3 · Tail value. Value × (1 + VaR99) = what the collateral is still worth in the bad case.
4 · Buffer. A liquidation must clear 10% below even that — so the max loan is tail value × 0.90.
5 · Liquidity cap. A card that rarely trades can't be liquidated at model price: trading continuity and observed slab supply cap the LTV (55% deep → 15% illiquid).
6 · Result. Max LTV, max loan, liquidation price. If we can't measure a card's risk, we say not rateable instead of guessing — that refusal is part of the method.

The search above covers the full rated universe — 2,000 cards, listed openly, re-ranked daily by market relevance. The top 250 (the free board) show the complete worked math right here; every other rated card, plus graded-slab terms and a suggested interest premium (expected-loss priced from the same calibrated tail), is one call to the API ($0.10). Where prices come from: raw-card values are TCGplayer market prices (sales-derived) mirrored daily — every derivation links to its TCGplayer product page and its on-chain Merkle proof, so you can check both the source and that we committed the number before you asked.
Risk inputs: the daily conformal board, committed to PredictionRegistry on Base · LiteForge twin · prices committed daily to our stats registry (stream tcg_price — one multi-stream contract carries sports stats and card prices side by side) · supply: the slab census · sports · souls · 🍄 The Undesirables