Forward ledger - valuations committed before the sale
A back-test is scored on sales the author already knew about. This ledger commits a valuation for a name while it is still for sale, publishes only a salted hash of it, and reveals the figure after a registered source records the sale, so the result can be checked against the committed figure. Each commit time is recorded by the committer and is not yet witnessed by anyone else: until the chain head is anchored with an external timestamp service, a reader has only this site’s word that a figure was committed before the sale, not proof of it.
Status
The ledger is empty. No valuation has been committed yet.
Candidates must come from a live-listing source recorded in the data provenance register. None is recorded yet, so the tool refuses every candidate rather than using names from anywhere else.
How it works
- Offline, the engine values each candidate once under a pinned methodology revision. The record holds the name, the listing source, the engine id, the revision, the appraisal hash, the value and its low and high bounds, and the commit time.
- Each record gets its own 32-byte random nonce. Only the hash of the nonce and the record is published; the record and nonce stay sealed off this site.
- When a registered source records a sale dated after the commitment, the record and nonce are published below and the prediction is scored. A name is never revealed before that, never committed twice, and never removed because it missed.
commitment = SHA256("odvs-forward/v1:" + nonce + ":" + canonicalJson(record))
leafHash = SHA256(canonicalJson({ seq, committedAt, rev, commitment, prev }))The nonce matters. The candidate list and the engine are both reproducible, so without it anyone could hash every candidate and read which names are in the ledger. Each leaf also commits to the leaf before it, so removing or reordering a row breaks every hash after it.
When the recording source does not permit its prices to be republished, the reveal shows only the source and whether the prediction was within 2x. An exact error would give the price back.
Revealed predictions
No prediction has been revealed yet.
Commitments
No commitments yet.