Validation
Before a signal is trusted it has to beat plain buy-and-hold on data it was never fit on, after costs, after accounting for how many things were tried. So far, nothing has.
The Sharpe ratios in the film show how the gate reads, not live results. The verdict is real: no signal has been certified.
Out of sample, or it does not count.
The best honest annualised result the harness has measured out of sample is about 10.8%. In its own tests nothing has beaten simply holding. We would rather tell you that than sell you a backtest.
Data it was never fit on.
Every signal is fit on one stretch of history and judged on a later one it never saw. Most look brilliant on the first and ordinary on the second. That gap is what Dev, the out-of-sample tester, watches for.
Deflated for everything we tried.
Try enough ideas and one will look good by luck. Priya, the multiple-testing statistician, discounts every result for how many things were tried before it. A Sharpe that survives that is worth a look. None has yet.
What changes if one passes.
Less than you might think. A validated signal can carry more weight in the floor's argument. It still goes through every limit and veto, and on a client desk it still waits for your click.
The rest of the desk.
Back to the overviewThe Floor
Seventy-four agents in seven departments, arguing in the open
Approvals
Every plan waits for your click
Risk limits
The numbers you signed, enforced in code
Practice desk
Real prices, fake money, two weeks
Alerts
WhatsApp, Telegram, iMessage, email or text
Research chat
Ask the floor anything. It cannot trade.
A small number of clients at a time.
Tell us a little about you. We reply personally, usually within a few days. If it is a fit, the questionnaire comes next, and your account after that.
Already a client? Sign in to get your install line