Apsis Algo

A research company that publishes its failures.

Every trading idea here is written down before it is tested, tested with real commissions and slippage, and judged against market data that was sealed before the idea existed. Most ideas die. The record of every one is public.

How an idea dies here

Five screens. Each idea falls as far as it earns, and pools where it stopped.

An idea is registered with its rules and its pass bar before any data is touched. It is then run with the fees a real broker charges, against a benchmark it has to beat, with its sample size counted against it. The few that pass get one run, ever, on data sealed from 2025 onward. Nothing has passed that last screen.

Three ways in

The same record, drawn three ways.

What we do not do

No signal service. No win-rate claim. No number without its costs.

No win rates

Win rate is a property of the exit, not the signal. The same 656 trades read 79% with a quarter-R target and lose money, or 26% with a five-R target and make it. We publish neither as a selling point.

No hindsight

Rules are frozen in a registry before a test runs, every grid is judged on its median, and the 2025+ data is a vault opened once per rule on a logged decision, never re-run.

No quiet failures

Dead ideas stay in the record with their cause of death. The Head scout reports what killed what, and the bar itself is tested against published premia to prove it can be passed at all.