Why Your Crypto Trading Bot Is Losing Money: 7 Checks

Short answer: most losing bots have one of seven problems: an untested strategy, costs that eat the edge, the wrong market, the wrong coins, a size that is too big, a bad stop, or a live bot that trades differently from the backtest. Here is how to check each one, with real examples from our own tests.

1. The strategy was never really tested

One good backtest on one period is not proof. It can be luck.

Check: test the same rules on other periods, nearby settings and other coins. In Algonney, ask Jarvis to "validate" the strategy and it runs all of these for you.

2. Fees and slippage eat the profit

A strategy can win before costs and lose after them. Fast strategies suffer most.

Real example: we tested a VWAP Z-score strategy on ETH, 5-minute chart, for 3 months, using the whole balance on each trade. Before fees, its profit factor was 1.32. After fees, it was 0.88. It made 76 trades and paid 85 USDT in fees on 1,000 USDT capital. The idea worked; the costs killed it.

Check: look at "profit factor before fees" next to the normal profit factor. A big gap means costs are the problem. Trade less often, or use limit orders where it makes sense.

3. The market changed

Trend strategies make money in trends and lose in sideways markets.

Real example: an EMA 50/200 trend strategy on BNB, XRP and AVAX had a profit factor of 0.59 from March 2022 to January 2024, then 1.64 from January 2024 to October 2026.

Check: look at results year by year, not only the total.

4. The wrong coins

The same rules can win on one coin and lose on another.

Real example: in the same test, BNB had a profit factor of 3.33, XRP had 0.27.

Check: read the per-coin results and remove coins that keep losing in every period.

5. The size is too big

Bigger size means bigger returns in a backtest, and much bigger drawdowns.

Real example: a Donchian breakout strategy on BTC, ETH, SOL and BNB since January 2021:

Risk per trade	Max drawdown
1 %	32.1 %
2 %	42.3 %
3 %	53.1 %

Check: choose the drawdown you can live with first, then the size.

6. The stop loss is wrong

Too tight, and normal noise stops you out. Too wide or missing, and one bad trade hurts a lot.

Real example: on the Donchian strategy from October 2021 to October 2026, a 5 % stop gave a profit factor of 1.48. With only a wide 30 % safety stop, it was 1.38. Here the stop helped. On other strategies a tight stop can hurt.

Check: backtest two or three stop sizes and compare.

7. The live bot is not trading like the backtest

Small differences add up:

• Account limits. A max position size can cut every trade smaller than in the backtest.
• Minimum order size. On a small account, 1 % of capital can be below the exchange's minimum order, so the order is refused and the trade never happens.
• Skipped signals. A max number of open positions can skip trades the backtest took.

Check: compare the bot's trades with the backtest's trades for the same dates. Algonney's backtests list skipped signals with their reasons, and each bot's events show refused and rejected orders.

Quick checklist

• Validated on other periods, settings and coins
• Profit factor after fees still above 1.2–1.3
• Results checked year by year
• Losing coins removed
• Size chosen from the drawdown you accept
• Stop loss tested
• Bot settings match the backtest

Find the problem faster

Ask Jarvis to backtest and validate your strategy, then read the costs, per-coin and per-year results. Learn more in Why your backtest lies.

Trading crypto is risky and you can lose money. Past results do not guarantee future returns. Read our Risk disclosure.