We publish the misses too.

Hedge funds and proprietary desks employ teams to read primary sources the moment they publish — SEC filings, analyst moves, supply-chain data — and to connect what they read to the positions they hold. That work is real. What almost never follows it is a published, auditable record of how often the resulting calls were right.

RexUnda is built the other way around. The measurement came first, and the product is what survives it.

We read the public record, correlate events into ranked signals, and show which positions are most likely to be affected. Then every call enters a scoreboard with its sample size attached, and stays there whether it worked or not.

Every signal carries a signal-strength score (0–100): a measure of how much independent evidence backs it, weighted by how each source has performed historically. Corroboration across sources raises it; a lone unconfirmed event scores lower.

What that number is not: the odds that a particular trade makes money. It measures how well-corroborated a claim is, nothing more. Anyone who tells you a confidence score is an expected return is selling something.

The scoreboard is separate, public, and carries its sample size — because a rate without an n is not evidence.

• Small signals matter. The biggest moves rarely come from loud news. They start quietly.
• Context beats volume. We don't flood you with alerts. We show you what actually affects your positions.
• Proof over promises. Every published rate carries its sample size. Thin samples are muted, not quoted.
• The misses stay up. A scoreboard you can edit after the fact is marketing. Ours only appends.
• Rules before results. Tests are written down and locked before the data exists, so a bad outcome can't be reinterpreted into a good one.

People who read primary sources themselves and want the reading organised, scored, and — crucially — kept honest by a record they can check.

If you have ever been shown a track record with no sample size next to it, this is built for you.

Saying so plainly is the point. The scoreboard is live and measuring, but nothing on it has yet cleared the bar we set for calling a result durable — and we would rather state that here than let a visitor infer otherwise.

A sealed test is scheduled. Its question, its scoring rule, and the conditions under which it cancels itself were all written down before any of its data existed. When it resolves, the outcome gets published either way.

Measure first.
Then earn the claim.