AI Paper Bot

Virtual money only · no broker connection · times in GMT+4 · refreshes every 5 s

checking… —

Balance (virtual)

—

Net profit / loss

—

Closed trades

—

Win rate

—

Today (GMT+4)

—

Max drawdown

—

Bot — trades automatically, no clicks needed

Loading…

Bot & settings

Change capital and risk in config.json, then restart the bot.

EUR/USD — … —

Data: Yahoo Finance EUR/USD spot (EURUSD=X) — can be delayed and is not a broker feed. Times GMT+4.

Live model probability

waiting for first calculation…

Probability BUY wins

—

Probability SELL wins

—

Model decision now: —

Last ~5 minutes. Green = BUY probability, red = SELL probability, dashed line = trade threshold, thin vertical lines = start of a new candle. The vertical axis is zoomed to 40–70% so small differences are visible.

Model input features (latest candle)

Live estimate: the deployed model run every second on the newest candle (including the one still forming). The bot itself only trades on completed candles. A probability is not a guarantee of profit.

Balance after each trade

Daily result (GMT+4)

Recent signals (completed candles)

Trade history

Research results — every test that was run

Market Show

✔ Edge — profitable in ≥3 of 4 unseen years (2022–2025) and clearly better than break-even ~ Marginal — slightly positive ⚠ Inflated — counted rollover-hour trades the bot cannot take ✘ No edge

Each row is a walk-forward test: the model trains on earlier years and is scored on later, unseen data (cross-data-vendor for EUR/USD). "Binary-style" rows measure direction accuracy at a fixed expiry and need a binary-options broker. "Return" compounds a fixed risk per trade over 4 years and is only for comparing rows. An "edge" here is a reason to run a forward paper test, not a promise of profit.