Bot dashboard
This page shows the live PAPER-trading performance of a self-learning AI trading bot on Binance USDT-M perpetual futures. No real orders are sent: the bot uses real market data but trades a virtual $100 account.
Starting balance: $100.00 · Started: September 26, 2026 · Last updated: Sep 26, 11:56 PM
Equity curve
The equity curve will appear here as the first trades close.
Bot settings
- Timeframe
- 2h
- Market scanned
- Top 50 most liquid perpetuals
- Max open at once
- 3 positions
- Relearning
- every 100 trades
- Trades until next relearn
- 100
- Risk per trade
- 0.50%
- Core basket share
- 50%
How does the bot decide?
At every 2-hour candle close the bot scans the day's 50 most liquid perpetuals (ranked by the previous day's 7-day average volume). It evaluates signals from 10 strategies; a signal forms on a candle's close and the trade is entered at the next candle's open. It holds at most 3 positions at a time.
How does it improve itself?
Every 100 closed trades the bot re-plans itself, looking only at trades that have fully closed (it never sees the future). It keeps a strategy only if it was profitable in at least 3 of 4 separate past periods and drops it if it lost in 3 of 4 - one bad stretch does not change its mind. For each strategy it picks, among 36 exit rules (stop distance, target, breakeven, holding time), the one that was consistently best, and bans market conditions (e.g. extreme volatility) that kept losing.
Core basket
Part of the capital sits in a 'core basket': every 7 days it holds, equally weighted, the 5 liquid coins with the strongest 30-day rise that are above their own trend. This basket is the main source of the bot's long-run return, but it can swing hard because it may hold new, volatile coins.
Risk management
Risk per trade is chosen by the bot from its own history, and for caution it uses half of the best value found (currently 0.5% of capital). Every trade has a stop. The exchange leverage setting is at least 10x; that only sets the margin - position size comes from the stop and the risk. A trade is skipped if liquidation would come before its stop. A leveraged core basket wiped the account in testing, so the bot assumes the basket can fall 80% and stays small enough to survive that.
6-year backtest summary
| Timeframe | Bot return | BTC buy & hold | Max drawdown | Trades |
|---|---|---|---|---|
| 15m | +75.3% | +1,049.2% | -86.4% | 7,883 |
| 30m | +367.0% | +1,049.2% | -87.5% | 11,552 |
| 1h | +862.5% | +1,049.2% | -62.7% | 10,252 |
| 2h Live | +2,842.2% | +1,047.2% | -64.0% | 6,794 |
| 4h | +2,056.3% | +979.1% | -71.7% | 3,143 |
| 12h | +1,498.2% | +899.4% | -73.2% | 1,017 |
| 1d | +742.9% | +859.6% | -80.3% | 268 |
The table shows the same system tested from 2020-01 to 2026-09 on 864 USDT perpetuals (delisted ones included), starting with $100 and re-learning every 100 trades. Fees, slippage and funding are included. Caveats: a large share of the gain comes from the 2021 altcoin season; some settings were chosen after seeing results, so real forward performance may be lower; drawdowns reached -60% to -80%.
Members only: positions, trades and strategy details
The rest is for members
Become a free member to read the full post and see detailed data. Sign in with your Google account in one click.
Membership opens very soon.
Glossary
- R
- A trade's result divided by the amount it risked. +2R: won twice the risk; -1R: stopped out, lost the full risk.
- Drawdown
- The deepest fall from the account's peak. -50%: a $200 peak dropping to $100.
- Paper trading
- Live testing with real market data but virtual money.
- Backtest
- Replaying the strategy over history, seeing at every moment only the data available up to then.
- Funding
- The fee paid every 8 hours between longs and shorts on perpetual futures; the bot includes it in its costs.