Track Record & Transparency
Prexora’s entire pitch is “don’t trust us, check us.” This page explains exactly how performance is measured and why anyone can verify it — without quoting numbers that change with every resolved bet.
Why there are no figures here. The win rate, bet count and P&L update continuously as bets resolve. Hard-coding them in documentation would guarantee they go stale. The live Track Record dashboard and the /export/track-record API are the single source of truth — always current, always authoritative.
How performance is measured
- Funded signals only. Display-only picks (shown for breadth, zero stake) are excluded from every figure. The win rate reflects bets the agent actually staked.
- Settled against ground truth. Each bet resolves on the official Polymarket outcome, which settles against the Weather Underground station reading for that city — objective, third-party-checkable data.
- Realized, not hypothetical. P&L is computed from actual stake and payout at the bracket’s resolution.
- Broken out by type. The API exposes per-city and per-bracket-kind win rates, so weakness in one market type cannot hide inside the aggregate.
The trust trio
Three independent mechanisms make the record checkable with zero trust in Prexora:
| Mechanism | What it proves |
|---|---|
| Signed feed (ed25519) | Each signal carries a signature — you can confirm it came from Prexora and was not altered after the fact. |
| Daily on-chain anchor (Solana) | A SHA-256 hash of each day’s signals and outcomes is written to Solana as a memo. Falsify any historical signal and the hash stops matching the chain — caught immediately. |
| Team eats last | An 18-month team performance cliff aligns the builders with the long-term record, not a launch-day exit. |
How to verify it yourself
- Pull a day’s signals and outcomes from the public API (
/export/signals,/export/recent). - Re-derive the canonical SHA-256 hash of that day’s data.
- Read the matching memo from Solana via
/export/onchain-commits(or directly on-chain) and confirm the hashes match. - Cross-check any resolved bet against the public weather observation for that city and date.
If the hashes match and the outcomes match reality, the record is exactly what it claims to be.
An honest note
Past performance is never a guarantee. Edges can compress, forecasts can be wrong, and the daily budget the agent runs at changes how the equity curve behaves over time. Prexora reports honestly — including losses, which appear in the public lesson stream. See Risks & Disclaimers.
Where the live data lives
Current numbers: the Track Record dashboard renders them and the public API serves them as JSON. If anything anywhere disagrees with the live API, the API wins.