Research ·
We tested take-profits and trailing stops. Mostly, they made things worse.
Take-profits never helped. Tight trailing stops cost up to 9 points a year. Wide ones looked better, but only because of 2021.
The most common question about a trend bot: why doesn't it take profit? Surely it's better to bank +50% than to watch it melt away. We tested it properly instead of guessing.
How the test works
The same bot, the same four coins, the same entries. Only the exit changes. For stops and targets, a 4-hour closing price isn't enough: we need to know whether the price touched the level during the candle. So the test checks each candle's high and low. When a stop and a target are both touched inside one candle, it assumes the stop hit first, the worse case. A price that jumps past a level fills at the opening price. Every exit pays fees and slippage.
Results, January 2021 to September 2026
| Exit rule | Return / year | Sharpe | Exits / year |
|---|---|---|---|
| Now: trend fades, or fixed stop 25% below entry | 45.6% | 1.66 | 1.4 stops |
| Take profit at +30% | 46.9% | 1.70 | 24 + re-buys |
| Take profit at +50% | 45.2% | 1.66 | 8 |
| Take profit at +100% | 45.3% | 1.65 | 3 |
| Take profit at +30%, then wait for the trend to reset | 36.8% | 1.47 | 17 |
| Trailing stop 15% | 36.5% | 1.41 | 96 |
| Trailing stop 20% | 42.2% | 1.57 | 49 |
| Trailing stop 30%, then wait | 49.7% | 1.79 | 10 |
What that means
Take-profits don't pay. Buying straight back after taking profit just adds trades for the same result. Waiting to cool off after taking profit is worse: the bot sits out the rest of the trend. A trend strategy lives on a handful of trends that go much further than anyone expects, and a take-profit is a rule that cuts exactly those short.
Tight trailing stops hurt. Crypto routinely swings 15–20% inside a healthy trend. A 15% trailing stop fired 96 times a year and cost 9 points a year.
Wide trailing stops looked good, and we still said no. A 30% trailing stop was about 4 points a year better overall. Here's where that came from:
All of the gain is in 2021, a once-in-a-cycle mania year. From 2022 to 2026 the two are within 2 points every year, and from 2024 on their risk-adjusted results are identical. It would add about 10 exits a year for copiers, for an edge that only existed in one unusual year.
The one exit we did add
A profit lock: once a position is up 100%, its stop moves to 25% below that price and stays there. It barely changes the backtest (45.7% against 45.6%; it fires about once every two years). It's insurance for the rare coin that doubles and then crashes. We also checked whether it should react faster.