The Verified Prediction Standard

§1

Preamble

Anyone can claim to have predicted an outcome after it is known. The credibility of a forecast lies entirely in when it was made, whether it can still be changed, and whether every forecast — not a chosen few — stands in the record. Yet in nearly every field where predictions are sold, none of these three things can be established by the public.

This standard defines the minimum requirements for a prediction to be called verified: sealed before the event, anchored in time beyond the forecaster's control, graded against the official result, and preserved — hits and misses alike — in a record that no one, including its author, can rewrite.

The standard certifies honesty of process, not quality of prediction. A forecaster of any skill may conform. What conformance guarantees is that the published record, whatever it shows, is complete and unaltered — and that anyone can check it.

§2

Definitions

Forecaster: Any party — human, model, or system — publishing predictions about future events.
Event: An occurrence with a scheduled start and an outcome determined by an official source: a race, a match, an announcement.
Prediction: The complete forecast content for one event: selections, rankings, probabilities, and any claims the forecaster wishes to be judged on.
Seal: A cryptographic hash of the complete prediction, computed before the event begins.
Anchor: Proof, held by a system outside the forecaster's control, that the seal existed at a specific time.
Grade: The published judgment of a prediction against the official outcome, under metrics defined in advance.
The Record: The complete, public, chronological body of the forecaster's sealed and graded predictions.

§3

Requirements

The key words MUST and MUST NOT are to be interpreted as absolute requirements of this standard.

1

Sealed before the event

The complete prediction MUST be committed by cryptographic hash (SHA-256 or stronger) before the event begins. Content published after the start of the event MUST NOT be represented as predicted.

2

Anchored beyond the forecaster's control

The seal's existence in time MUST be attested by an independent, publicly verifiable timestamp authority the forecaster does not operate — for example, anchoring to a public blockchain. A timestamp the forecaster could backdate is not an anchor.

3

Graded from the official source

Outcomes MUST be taken from the event's official governing record, and that source MUST be named. Metrics defining success and failure MUST be published before they are applied, and applied uniformly to every event.

4

Complete — no survivorship

Every sealed prediction MUST be graded and published: hits, misses, and voided events alike. A cancelled event MUST be labeled as cancelled, never deleted. Publishing a subset of results, however framed, does not conform.

5

Immutable

Entries in the record — predictions, grades, and timestamps — MUST NOT be edited, re-scored, or removed after the fact, by anyone, including the forecaster. An error discovered later MUST be corrected only by a forward, dated annotation that preserves the original entry in full.

6

Publicly auditable

Seals, anchors, graded outcomes, and metric definitions MUST be publicly accessible, such that any third party can verify any entry — or the entire record — without the forecaster's cooperation or permission.

7

Continuous

Conformance is a property of an unbroken practice, not of an exhibition. A forecaster claiming conformance MUST apply this standard to all predictions published in the conforming record, from a stated start date, without interruption. The record's age is part of the record.

§4

Conformance

A forecaster claiming conformance publishes a statement naming: the start date of the conforming record, the hash algorithm and anchoring method, the official grading sources per event class, and the metric definitions. The claim takes this form:

“This record conforms to the Verified Prediction Standard v1.0: every prediction sealed before its event, independently anchored, graded from official sources, complete, immutable, and publicly auditable since [date].”

Any party may test the claim against §3. A record that fails any requirement, for any entry, does not conform.

§5

What This Standard Does Not Certify

Conformance is not a claim of accuracy, profitability, or fitness for any purpose. It does not certify that predictions are good — only that the record of them is honest. The standard imposes no requirements on how forecasts are produced, and it neither offers nor implies wagering advice. Skill is for the record to demonstrate and the reader to judge.

§6

Provenance

The reference record for this standard is racehp.ai's verified record: 7,000+ predictions sealed, timestamped, and graded in public across seven countries — produced by a single unified neural network (URIN, patent pending) spanning biological racing, where the athlete is a living animal, and mechanical racing: Formula 1, NASCAR, IndyCar, and MotoGP. Its conformance to this standard is what makes that claim testable: one engine, both worlds, right in public.

Every race's sealed top-3 picks are submitted for anchoring to Bitcoin via OpenTimestamps. Full-field hashes are sealed and timestamped but not individually anchored.

AuthorRaceHP.ai — Roger Szymczak, Founder Version1.0, published August 30, 2026 LicenseCreative Commons Attribution 4.0 (CC BY 4.0). Any forecaster in any field may adopt this standard; adoption requires attribution of the standard, not of any record. ChangesFuture versions may add requirements; none may weaken §3. Version history will be preserved in full, in the spirit of §3.5.

§7

Audit This Yourself

Requirement 6 says a record must be checkable without its author's cooperation. Here is how to audit the reference record — the last step uses only open-source tools and the Bitcoin blockchain, neither of which RaceHP operates.

1

Pick any race

Open the verified record and choose any race marked ✓ anchored — the live share of graded races carrying a completed proof is stated in §6. Its card shows the sealed picks with their anchor state, the grade against the official result, and the prediction's SHA-256 hash with its lock timestamp.

2

Check it against the server

Use Verify Hash on Server on the race card. This confirms the displayed hash matches the stored record. It is a convenience check — it still trusts RaceHP, which is why step 3 exists.

3

Verify the Bitcoin anchor — without us

Open Verify on Blockchain on the same card and copy the OpenTimestamps proof (a base64-encoded .ots file). On your own machine:

echo '<paste the base64 proof>' | base64 -d > prediction.ots ots info prediction.ots # sealed digest + calendar attestations ots upgrade prediction.ots # fetch the Bitcoin attestation from the calendars ots verify -d <the race's hash> prediction.ots # check it against the Bitcoin blockchain

The digest inside the proof must equal the race's displayed hash. Read the times exactly: the lock timestamp and the calendar submission (about one second later) are the pre-race receipt, roughly 15 minutes before post. The Bitcoin block that seals the proof usually follows within the hour and, in most races, lands after the race has run (measured 2026-08-31: 36 of 56 blocks came after post). Bitcoin proves the sealed prediction could not have been altered after that block; it does not by itself prove it existed before post. The open-source OpenTimestamps tools — CLI or the drag-and-drop verifier at opentimestamps.org — are maintained by the OpenTimestamps project, not by RaceHP. Nothing in this step touches our servers.

Two honest notes: a freshly sealed proof is calendar-pending until it is aggregated into a Bitcoin block (usually within the hour; measured median 26 minutes), after which the attestation is permanent. And today the anchor proves when the sealed prediction existed; publishing each race's sealed payload for byte-exact recomputation of the hash itself is the planned next step of this audit trail.

4

Or watch a real verification — replayed here

Below is a recording of exactly those commands, run against the actual sealed proof of one race in the record — the sealed #1 pick for the Sir Barton Stakes at Laurel Park, May 16, 2026, locked 15 minutes before post. It is a replay, not a live check: a page can never prove its own honesty, which is why every command is shown so you can run it yourself.

opentimestamps — recorded sessionDemonstration

Show the full proof (base64, 888 characters)

AE9wZW5UaW1lc3RhbXBzAABQcm9vZgC/ieLohOiSlAEIaJ7Uz3vz0Y6DWuEb5bWBguwcQoM6uFuSqDPBZYtdpXDwEPelcGt0GQixRm/MRtb8jxgI//AI/wKM1mRc+BMI8BCrgERBhfpzYmjAFyFl18aeCPEgf7qjxsn/gdyfwDtLz4voWvoCQD998rP6KPKpYfmZg+sI8CAzcUetZZvEi+21s/nrFaCdaCHuv9bj9BcKHUyfARpqKAjxBGoIpZ/wCMrRa1vSLeadAIPf4w0u+QyOLi1odHRwczovL2FsaWNlLmJ0Yy5jYWxlbmRhci5vcGVudGltZXN0YW1wcy5vcmf/8AjTkonSXhNibQjwEPC4rnf3Opnh6u6NODAZP10I8SCuiA5pPstWV2vBiIVb/16xRPiuaBtnvlUEos1t+XERnQjxBGoIpZ7wCDE/tBTeAGg+AIPf4w0u+QyOLCtodHRwczovL2JvYi5idGMuY2FsZW5kYXIub3BlbnRpbWVzdGFtcHMub3Jn//AQGcZaP9vMMe0dt/UgeL5dFgjwIECrxPG0JZLWJiLh95RBL1hfkVcDrXJglZmZsVUjIjhhCPAgacc7SGMNCDYPxyKPo9RQ5ChwYkdE0mB2KIhwXF8KpKgI8QRqCKWf8Aj3bZzYRi4y9QCD3+MNLvkMjikoaHR0cHM6Ly9maW5uZXkuY2FsZW5kYXIuZXRlcm5pdHl3YWxsLmNvbfAQkEyA+9kKdXpcrxwkHVxZ2wjxIEeIzJHsv3c5AG2JJoU2IT6bzW8macFBUmF+NTr50HEHCPEEagilnvAI+nnopbXg5toAg9/jDS75DI4jImh0dHBzOi8vYnRjLmNhbGVuZGFyLmNhdGFsbGF4eS5jb20=

Recorded 2026-08-31 · OpenTimestamps JavaScript client v0.4.9 (npm install -g opentimestamps) · output unedited · this demo makes no network requests

See the Record That Conforms →