Event Contracts you can check before you trade them.
A price on a binary contract is a forecast, and a forecast nobody has scored is an opinion with a number on it. Rivo scores them — against contracts that have already settled, with the sample size attached to every claim — and tests whether an agent’s edge survives the spread before any capital moves. Nothing here needs a wallet to read.
Two problems, both measurable
A probability is not self-explanatory
DreamDEX may say BTC UP 15m is 67%. Nothing on the venue tells you whether contracts that quoted 67% went on to settle true about 67% of the time — which is the only question that number raises.
Rivo measures it against contracts that have already settled, with the sample size attached to every claim.
Prediction accuracy is not economic edge
A model can predict well and still trade badly — being right about direction is not the same as being right by more than the spread you cross. So 7 strategies now run against live Event Contracts, deciding but never spending, each resolved against the venue’s own settlement.
Rivo’s own model is one of them, and it is the one that failed: AUC 0.8158, −6.49% return on stake out of sample. It is marked REJECTED, and the execution path enforces that rather than displaying it.
Understand, validate, prove
and every outcome feeds the next answer- Understand
Live Event Contracts with implied probability, spread, depth and time to expiry — beside how often comparable contracts actually settled true.
- Validate
An agent is replayed against settled history with walk-forward validation, and judged on realised economics rather than accuracy — with the sample size and the interval attached, and no verdict at all until there is enough of both. The result is a state the execution path reads.
- Shadow
It then runs against live markets deciding but not spending. Each decision is recorded and resolved against the real settlement when the contract closes.
- Prove
Only then, and only on an approved testnet, does a decision become a real DreamDEX transaction — with the hash, the receipt and the reconciliation all inspectable.
- Evidence
Every settled contract joins the calibration dataset and the agent’s record, so the next answer rests on one more settled fact than the last.