Guides · Guide 12
Copy trading, trading bots, or doing it yourself
Running a strategy yourself buys control at the price of operations. Copying buys the execution and the operations at the price of a profit share.
- Bugs found in 3 days of demo
- 5
- Slippage saved by fixing exits
- 9–22 bps
- Profit share, copying
- ≈ 10%
The three routes
Manual trading: you decide and place every order. Running a bot: software you own or rent decides and places orders through your API keys. Copy trading: someone else's orders are reproduced in your account. The question is not which is best but what each one actually costs, in money and in attention.
Manual trading
Free of profit share and software, and the most expensive of the three in practice. The evidence on retail futures trading is consistent: most accounts lose, and the reasons are behavioural, not analytical. Position sizing by conviction, moving stops, adding to losers, and trading more in drawdowns. Every risk control in a systematic book exists because a person cannot be relied on to apply it under stress. Guide 04 lists them.
Running a bot
A bot removes the behavioural problem and adds an operational one. It needs a server that stays up, exchange keys with trading rights, monitoring, reconciliation against the exchange, and someone who understands what it does when the exchange misbehaves. Our own bot's first three days on Binance's demo venue produced five findings no backtest had shown:
- Stop orders are rejected on a flat account, so the self-test had to open a real position to verify stops.
- Binance's default acknowledgement reports a market order as unfilled even when it filled; every order now requests the full result.
- Settings changes are refused while stop orders are open, which produced false alarms on every restart.
- Reconciliation misread a suspended asset as a closed position.
- Passive orders on exits cost 9–22 basis points to save 2. Exits now cross the spread.
All five were found before real capital was at risk; that is what a demo stage is for. Anyone renting a bot inherits whichever of these its author has or has not found. Anyone running their own must find them alone. The launch log has the detail.
Copy trading
Copying buys the strategy, the execution and the operations for a profit share, about 10% of profit on Binance's default and on Hyperliquid vaults. What it does not buy is control: you cannot change the rules, and your result depends on the lead's discipline. What it adds: copy drift below about $1,000 on Binance, and the behavioural risk of stopping in a drawdown, which no platform removes.
Costs, side by side
| Manual | Own bot | Copy | |
|---|---|---|---|
| Exchange fees and funding | Yours | Yours | Yours |
| Strategy cost | Your time | Build or rent; your time | ≈ 10% of profit |
| Operations | You, live | Server, keys, monitoring, you on call | None |
| Execution quality | Depends on you | Depends on the code | Depends on the lead |
| Behavioural risk | Highest | Low, unless you intervene | Stopping in drawdowns |
| Control | Total | Total | Start, stop, amount, stop-loss |
Choosing
Run your own bot if you can build, test and operate one and want the control. Copy if you want the outcome without the operations and can accept a profit share and a lead's rules. Trade manually only with capital you have decided to spend on learning. Whichever route, the numbers that decide the outcome are the same: net return after costs, worst drawdown, and whether you stayed in.