Evolving a Trading Strategy with Small, Targeted Code Changes
Summary
The document argues that trading systems should be written so that changes to strategy logic require only localized code edits. It illustrates this with an R-Breaker example: if backtesting suggests that holding positions overnight adds risk without enough return, the strategy can be changed to exit before the session closes. The example checks market time, sets the target position to zero near the close, waits for order handling, and then shuts down.
The discussion highlights a practical caveat: a position-target update may not fill immediately, because price movement can require cancellation and resubmission. Waiting for a fixed interval is offered as a simple approach, while checking that the position has reached zero is suggested when greater certainty is needed. The document provides no backtest results showing that the revised exit improves performance; it presents a coding and workflow pattern rather than evidence for the strategy change itself.
Key ideas
- Writing strategy logic in a linear style can make small rule changes easier to implement.
- The example adapts an R-Breaker strategy to close positions before the end of the trading session.
- Setting a target position to zero does not guarantee an immediate fill.
- A strategy can wait for order handling or verify that the position is flat before shutting down.
- The example describes an implementation pattern, not evidence that the revised strategy performs better.
Tags
Full text
# evolution
使用逐步完善的方式构建交易策略
=================================================
`TqSdk`_ 的目标是能尽可能减少编码环节引入的坑，而交易逻辑本身的坑则需要用户自己填，如果每次调整交易逻辑都需要大规模的代码重构会严重阻碍交易系统的演化。
`TqSdk`_ 鼓励使用线性的编码风格，因此可以做到小调交易逻辑只需小改，只有大调交易逻辑时才需要大改。以 `R-Breaker`_ 策略为例，
第一版是留仓过夜，回测下来可能发现留仓过夜引入了很大的风险，却没有获得与风险对应的收益，因此修改交易策略，收盘前清仓离场，
对应代码的修改只需在主循环中加入判断是否接近收盘并平仓::
if api.is_changing(quote, "datetime"):
now = datetime.strptime(quote.datetime, "%Y-%m-%d %H:%M:%S.%f")
if now.hour == close_hour and now.minute >= close_minute: # 到达平仓时间: 平仓
print("临近本交易日收盘: 平仓")
target_pos.set_target_volume(0) # 平仓
deadline = time.time() + 60
while api.wait_update(deadline=deadline): # 等待60秒
pass
api.close()
break
上述代码在行情时间变化时判断是否接近收盘，如果是的话则将目标持仓设为0(即空仓)。由于下单之后不一定能立即成交，价格变化后可能还需撤单重下，
因此等待一分钟后再退出，通常交易的合约不是太冷门的话一分钟应该足够了，如果不放心的话可以改为判断持仓手数是否为0。
更多的用例可以参见: https://doc.shinnytech.com/pysdk/latest/demo.html
.. _TqSdk: https://doc.shinnytech.com/pysdk/latest/index.html
.. _TqSim: https://doc.shinnytech.com/pysdk/latest/reference.html#tqsdk.sim.TqSim
.. _get_kline_serial: https://doc.shinnytech.com/pysdk/latest/reference.html#tqsdk.api.TqApi.get_kline_serial
.. _TargetPosTask: https://doc.shinnytech.com/pysdk/latest/reference.html#tqsdk.lib.TargetPosTask
.. _wait_update: https://doc.shinnytech.com/pysdk/latest/reference.html#tqsdk.api.TqApi.wait_update
.. _DIFF: https://doc.shinnytech.com/diff/latest/index.html
.. _get_account: https://doc.shinnytech.com/pysdk/latest/reference.html#tqsdk.api.TqApi.get_account
.. _get_quote: https://doc.shinnytech.com/pysdk/latest/reference.html#tqsdk.api.TqApi.get_quote
.. _is_changing: https://doc.shinnytech.com/pysdk/latest/reference.html#tqsdk.api.TqApi.is_changing
.. _TqBacktest: https://doc.shinnytech.com/pysdk/latest/reference.html#tqsdk.backtest.TqBacktest
.. _R-Breaker: https://github.com/shinnytech/tqsdk-python/blob/master/tqsdk/demo/example/rbreaker.pyShown in full with attribution under the source's licence. Licence: Apache-2.0
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.