After a day of tweaking, I replayed the historical candlestick data from January 2026 to now, and the results are as follows:
Current v007 backtest:
Period: January 2026 to present
Assets: BTC / ETH / BNB
Capital: 1200U
Position: 10% equity margin per trade
Leverage: 10x
Rules: Orders that haven't been filled within 6 hours are automatically canceled.
Results:
Return rate +72.39%
Win rate 71.62%
Max drawdown -10.15%
PF 2.75
Average of about 15 trades per month
A few key designs were implemented in the system:
1. Multi-version strategy registration mechanism
v001 / v002 / v007 can be plug-and-play, no need to rewrite the main process for backtesting and live trading.
2. Decoupling strategy from execution parameters
The strategy only handles direction, entry, stop-loss, and take-profit; capital, position, leverage, and withdrawal time are managed centrally in the runtime profile.
3. Shared position calculator for backtesting and live trading
Avoids having "one setup for backtesting and another for live trading." Contract size, notional position, and margin are calculated uniformly.
4. Order expiration handled at the system level
Since the exchange doesn't support custom expiration times, we created a watcher: orders are automatically canceled if not filled within 6 hours.
5. Separation of notifications and order state machine
Using REST polling + private WebSocket bypass to avoid false reporting of terminal orders.
The focus now is not just on achieving a nice backtest, but on making strategy development, backtesting, execution, risk control, and notifications into an iterative system.
#大漠茶馆 #量化交易BTC #OPENCLAW
Current v007 backtest:
Period: January 2026 to present
Assets: BTC / ETH / BNB
Capital: 1200U
Position: 10% equity margin per trade
Leverage: 10x
Rules: Orders that haven't been filled within 6 hours are automatically canceled.
Results:
Return rate +72.39%
Win rate 71.62%
Max drawdown -10.15%
PF 2.75
Average of about 15 trades per month
A few key designs were implemented in the system:
1. Multi-version strategy registration mechanism
v001 / v002 / v007 can be plug-and-play, no need to rewrite the main process for backtesting and live trading.
2. Decoupling strategy from execution parameters
The strategy only handles direction, entry, stop-loss, and take-profit; capital, position, leverage, and withdrawal time are managed centrally in the runtime profile.
3. Shared position calculator for backtesting and live trading
Avoids having "one setup for backtesting and another for live trading." Contract size, notional position, and margin are calculated uniformly.
4. Order expiration handled at the system level
Since the exchange doesn't support custom expiration times, we created a watcher: orders are automatically canceled if not filled within 6 hours.
5. Separation of notifications and order state machine
Using REST polling + private WebSocket bypass to avoid false reporting of terminal orders.
The focus now is not just on achieving a nice backtest, but on making strategy development, backtesting, execution, risk control, and notifications into an iterative system.
#大漠茶馆 #量化交易BTC #OPENCLAW