A drawdown ladder can step the whole system down to reduced sizing, and then to a full halt. The validator enforces it; the model does not get a vote.
| Symbol | Entry | Stop | Status |
|---|---|---|---|
| – | withheld | attached | protected |
Entry and stop are placed together, and the stop lives at the broker, so a position stays protected even if the machine running this goes offline.
Most sessions end in no trade at all. This panel is the record of where every candidate stopped, and it is the one people ask about first.
| Candidate | Verdict | Reason |
|---|
Rejections are logged as fully as executions. Roughly nine in ten proposals never make it past the gate.
This preview is deliberately thin. The production system carries considerably more, the full scoring model, the complete gate set, the learning loop that revises it, and real performance data measured in R rather than the blanks above.
Happy to walk anyone through the live console and share the actual numbers.
Request the full demo →It researches, decides and places the orders on its own. Plain code decides whether any of it is allowed to happen.
Every morning it sweeps the market for names worth a second look, and keeps a running memory of the ones that are moving.
Anything promising gets a full written work-up, including an argument for why the idea is wrong. No write-up, no trade.
The work-up becomes a single number. Under the bar, nothing happens, however good the story sounds.
Every order hits one deterministic gate that the model cannot argue with. If the gate says no, it is no, and the refusal is recorded.
Anything that gets through is protected the moment it opens, then graded afterwards. What is learned changes the scoring, and only evidence can change it.
The real thing has considerably more going on underneath: how candidates are actually sourced, what the gate really checks, and how the model rewrites itself. Those stay private.
If you want the detailed walkthrough and the performance figures, just ask.
Request the full demo →