Crypto paper trading bot

Paper trading only, permanently. Testnet funds are not real money. Honest numbers, not profitable numbers.

Current paper position

Rules: EMA(20/50) crossover on daily closes, frozen before testing. Simulated orders on the Binance Spot Testnet with valueless test funds. This is a record of what the rules did — not advice, and not a recommendation to buy or sell anything.

Paper period vs buy and hold

bot equity (USDT)buy and hold at the bot's 2.0% capbuy and hold, fully invested

benchmark: first recorded equity split equally across both pairs at the first day's close with entry fees and slippage applied; exit costs would apply on realization. Marked at daily signal closes.

Exposure: the bot risks at most 2.0% of equity per pair, while the fully invested benchmark is 100% committed the whole time, so the two are not like-for-like on risk — the middle line is that same benchmark held at the bot's own cap. The bot held a position on 40 of 61 charted days (66%).

Every tick, every reason

The two year backtest that earned it a paper run

EMA(20/50) crossoverRSI(2) pullbackbuy and hold

BTC/USDT

EMA(20/50) crossover: +8.17% (max drawdown 32.79%, 7 round trips, profit factor 1.16, in market 49.7% of days)

RSI(2) pullback: +12.42% (max drawdown 8.25%, 10 round trips, profit factor 2.48, in market 4.0% of days)

buy and hold: -6.89% (max drawdown 53.04%) — fees and slippage included on every line above.

ETH/USDT

EMA(20/50) crossover: +16.98% (max drawdown 32.63%, 6 round trips, profit factor 1.43, in market 37.1% of days)

RSI(2) pullback: +26.97% (max drawdown 8.01%, 6 round trips, profit factor 13.78, in market 3.2% of days)

buy and hold: -54.41% (max drawdown 67.55%) — fees and slippage included on every line above.

Both rules were pre-registered before they were run, and neither is crowned a winner: these are ~10-trade samples on one window and two surviving pairs. This window was a falling market, so a rule that sits in cash beats buy and hold almost mechanically — that was written down in advance as NOT evidence of an edge.