Research · pre-registered · permanent
BRIDGE AI turns a trading idea into a precisely specified strategy, backtests it on recorded real market data against a gate whose thresholds were written down before the test ran, and files one verdict — pass or fail — in an archive it can never quietly edit. It does not trade.
Research only · informational only · no broker, no orders, no positions
◀ an illustration: one rule decides, and you can see which
How it works
Most of what enters the funnel has to die in it. If nothing were dying, the gate would be broken — so failure is the ordinary outcome here, and it is recorded as carefully as a success.
Sample floor, expectancy after real costs, profit factor, and a margin over a 20-seed coin-flip control — all fixed before anything is run. A gate you may adjust once you have seen the answer is not a gate; it is a way of writing down what you already wanted to believe.
The control takes the same entries on the same bars with the same stops and targets, and picks its side at random. Most “edges” are geometry: a wide stop and a tight target win often, and so does a coin flip with the same shape. Beating that control is the only evidence the entries knew something.
Pass or fail, each verdict is written to a permanent archive with the exact rules it was judged by and how many strategies had been tested before it. Failures do not quietly disappear, and “we already tried that” stops being a memory.
The vocabulary
Every strategy leaves with exactly one of these, and the word is fixed in advance. A small, countable vocabulary is what makes “how many have we tested, and how many passed?” an answerable question rather than a story.
“Insufficient evidence” is deliberately not a failure — we did not learn anything and we learned this does not work are different facts. “Disqualified” means the test itself was invalid, so its numbers are never shown at all. And a pass is still only a proposal: nothing is shortlisted until a person signs off.
Follow along
The research is private for now. Leave your email and we will be in touch about what we find — including, mostly, what does not work.
✓ You're on the list. We'll be in touch.
Something went wrong — please try again.
No spam. Ever.