The audit report — not the sales deck
The audited financials of trading signals
Real settlement · No synthetic data · Pre-registered forward tests
Every prediction is settled against the real market price and scored in public — including the verdict “no proven edge yet.” Most of this industry hides its ledger. We built the machine that can't lie to us.
LEDGER HEAD … · … settled · …
What “honest” means here
📈
Settled against real prices
Outcomes resolve at the actual Kalshi / Polymarket close — never against the model's own mark.
🚫
No synthetic data, ever
The engine's own rule: no real order book means the trade is skipped. Nothing is filled against an invented price.
🔒
Pre-registered & frozen
Every pass/fail bar is set before the test. Changing a bar mid-test voids its verdict.
The calibration truth
Market-implied probabilities are well-calibrated. Our model has shown no exploitable edge after costs — and we say so.
Brier — …
The forward tests
Nothing graduates to real money until its gate is met. We show the failure (cult_v1, −67% ) on the same screen as the works-in-progress.
How it's settled
1 Predict — the model logs a probability.
2 Lock the price — the real venue quote at decision time.
3 Settle — at the actual Kalshi / Polymarket close, on the real outcome.
4 Hash — each outcome chains into an append-only ledger (tamper-evident).
The server trusts nothing from the client; verdicts come from settled venue prices only.