"My win rate is almost 65%. Why am I losing money?"
I stared at the numbers for a full minute. The Scalper, one of our BotLab experiments, wins nearly two out of three trades. By every intuitive measure, it should be printing money.
It's not. It's down -6.9%.
And the reason is the most important lesson in trading that nobody talks about.
The Experiment
In our BotLab, we test trading strategies with virtual money on real market data. One of our bots, The Scalper, follows a simple mean reversion strategy:
- Buy every time BTC drops 3% in a single day
- Sell when it recovers 2% from entry
- Stop loss at -5%
Over 16 months, it executed 54 trades. 35 of them were winners. That's a 64.8% win rate. Sounds great, right?
Here's the problem.
The Fee Math Nobody Does
Every trade costs money. On Bybit (and most exchanges), you pay about 0.1% per trade. That's a buy AND a sell, so each round trip costs 0.2%.
54 trades × 0.2% = 10.8%
That doesn't sound like much. But run the compound math:
After 54 round trips at 0.2% each, you've lost approximately $1,307 in fees alone. That's 13.1% of your starting capital, gone before you've made a single dollar of profit.
The Scalper's gross profit (before fees) would have been around +6%. But after $1,307 in fees, the net result is -6.9%.
Remember: A 65% win rate means nothing if your average win doesn't cover the average loss plus fees. The math is brutal and it doesn't care about your feelings.
The Comparison That Proves It
We ran four bots at different trading frequencies. Same timeframe. Same starting capital. Same market. Only the strategy and frequency differ:
| Bot | Trades | Win Rate | Fees Paid | Net Return |
|---|---|---|---|---|
| The Momentum Bot | 51 | 37.3% | $1,386 (13.9%) | +36.9% |
| The VWAP Bot | 22 | 72.7% | $546 (5.5%) | +15.0% |
| The RSI Bot | 3 | 66.7% | $68 (0.7%) | +0.2% |
| The Scalper | 54 | 64.8% | $1,307 (13.1%) | -6.9% |
Look at this table carefully.
The Momentum Bot has a LOWER win rate (37.3%) than The Scalper (64.8%). It paid MORE in fees ($1,386). And yet it returned +36.9% while The Scalper lost -6.9%.
How? Because when Momentum wins, it wins BIG. The average winning trade is +12%. The Scalper's average win is only +2%. The big wins cover the losses and the fees.
The Rule
This gives us a simple rule that every trader should tattoo on their forearm:
Your average win × your win rate must be larger than your average loss × your loss rate PLUS your total fees.
Or more simply: Big wins beat frequent small wins.
The Scalper wins often but wins small. The Momentum Bot loses often but wins big. After fees, big wins dominate.
Remember: Trading frequency is not free. Every trade you make costs real money. The question isn't "how often should I trade?" It's "does each trade earn enough to justify the fee?"
What This Means for You
If you're day trading crypto and paying 0.1-0.2% per trade:
- 10 trades per month = 2.4% annual fee drag
- 50 trades per month = 12% annual fee drag
- 200 trades per month = 48% annual fee drag
At 200 trades per month, you need to generate almost 50% in gross returns just to break even. Most hedge funds can't do that. You certainly can't.
The BotLab Takeaway
Our 11 bot experiments in the BotLab teach one clear lesson: the best strategies trade rarely but decisively. Our top performers:
- The Momentum Bot (+36.9%): 51 trades, but each one captures a multi-day trend
- The Whale Watcher (+30.1%): only 5 trades, big moves
- The Funding Rate Bot (+28.4%): only 2 trades, perfect timing
The worst performers all share one trait: they trade too often for too little edge.
Remember: Win rate is vanity. Net profit after fees is sanity.
-> See all 11 experiments live — The BotLab Leaderboard
-> Why Daytrading Burns Money — The statistics are devastating
-> Meet our Live Bots — The Watchdog, The Scout, The Genius
Your Dominic, the guy who built a bot with a 65% win rate and watched it lose money. For science.
Disclaimer: This is not financial advice. All results are from backtests on historical data with virtual money. Past performance does not guarantee future results.
