Transparency
Real-money readiness scorecard
ATLAS trades simulated capital. Before any real money is ever put behind it, every gate below has to be green — measured net of commission and spread. Nothing here is smoothed: the numbers refresh on the nightly scoring run straight from the settled track record.
ATLAS Desk (stocks + crypto)
3/6 gates green- Settled trades92 / ≥ 200
- Out-of-sample expectancy (net)4.71% / > 0%
- Max drawdown62.4% / < 15%
- Profit factor4.36 / > 1.30
- Cost drag on win rate3.3 pts / < 5 pts
- Consecutive green months2 / ≥ 3
Simulator training (Sim Gym)
Every night ATLAS replays years of real daily candles across a broad stock and crypto universe, fires the same chart setups it trades live — trend pullbacks, breakouts, breakdowns and gap-driven news reactions — and settles each one on the true high/low path, net of spread and commission. It is how the bot builds statistical experience far faster than its live book can. Simulated evidence is weighted at a quarter of the real thing and only fills the gaps where live history is still thin.
Last training run · 10/19/2026, 11:50:04 AM
Are the bot's filters worth it?
Every signal the bot refuses is logged and scored as if it had been taken. A negative average means the filter saved money; a positive one means it is costing money, and the learning loop loosens it.
- cash macro shockn=8 · avg -0.95%
- lev risk-off (3 losses in a row)n=8 · avg -7.77%
- lev benchedn=7 · avg -4.79%
- cash no filln=5 · avg +3.99%
- lev macro shockn=1 · avg +0.22%
See the live books: ATLAS desk · full track record. Simulated results only — not investment advice.