第一章神龙摆尾
本章围绕一个结构模型，基于backtrader框架，从策略实现、参数调优等方面探讨量化交易实践。
本模型来自网易公开课，是一个波段上涨模型，名曰亢龙有悔，模型基本示意如下
模型要点如下：
策略实现
策略源代码
初步参数调优-选股参数
测试发现
preboxnohigh参数纯属画蛇添足，遂移除。
最大机会数量
2154 4314 1599 2636 3589.04 {"postboxV":1,"maxuplimit":-1,"needvcd":False,"breakdown":0,"boxR":7}
2154 4314 1599 2636 3589.04 {"postboxV":1,"maxuplimit":-1,"needvcd":False,"breakdown":1,"boxR":7}
2154 4314 1599 2636 3589.04 {"postboxV":1,"maxuplimit":-1,"needvcd":False,"breakdown":2,"boxR":7}
最高收益
1760 2746 1024 1658 19236.8 {"postboxV":1,"maxuplimit":0,"needvcd":False,"breakdown":0,"boxR":7.4}
1760 2746 1024 1658 19236.8 {"postboxV":1,"maxuplimit":0,"needvcd":False,"breakdown":1,"boxR":7.4}
1760 2746 1024 1658 19236.8 {"postboxV":1,"maxuplimit":0,"needvcd":False,"breakdown":2,"boxR":7.4}
907 1128 422 690 19609.18 {"postboxV":2,"maxuplimit":1,"needvcd":True,"breakdown":2,"boxR":7.4}
865 1063 395 652 20718.98 {"postboxV":2,"maxuplimit":1,"needvcd":True,"breakdown":2,"boxR":8.2}
1118 1500 544 936 20830.49 {"postboxV":2,"maxuplimit":1,"needvcd":False,"breakdown":2,"boxR":7}
874 1076 404 656 21041.59 {"postboxV":2,"maxuplimit":1,"needvcd":True,"breakdown":2,"boxR":8}
934 1167 442 708 22042.67 {"postboxV":2,"maxuplimit":1,"needvcd":True,"breakdown":2,"boxR":7}
进一步参数调优-卖出参数
选定{"postboxV":1,"maxuplimit":-1,"needvcd":False,"breakdown":2,"boxR":7}，期待通过止盈止损参数提高收益
ps = dict(
closeup1 = [x/100 for x in range(105, 118)], # 目标止盈
closeup2 = [x/100 for x in range(103, 110)], # 顶部回落止盈
closedown = [x/100 for x in range(72,90,2)], # 止损
)
进一步测试发现，如下参数组合可获得显著的性能提升
2149 4340 2705 1550 2925 31900.29 {"closedown":0.72,"closeup2":1.09,"closeup1":1.05}
2149 4340 2705 1550 2925 31900.29 {"closedown":0.74,"closeup2":1.09,"closeup1":1.05}
2149 4340 2705 1550 2925 31900.29 {"closedown":0.76,"closeup2":1.09,"closeup1":1.05}
其中：
胜率从37%提升到63%，提升了68%收益从220467提升到31900.2提升了47%
三个参数中有两个处于测试范围的边界，下一步可考虑更换范围继续搜索。
所涉策略仅用于技术探讨，不代表作者认同该模型有效性。
文章为作者独立观点，不代表 股票程序化软件自动交易接口观点
相关文章
股民评论