回测框架PyalgoTrade以及tushare在Python3.6应用

最新推荐文章于 2024-04-22 16:07:30 发布

weixin_34240520

最新推荐文章于 2024-04-22 16:07:30 发布

阅读量589

点赞数

文章标签： python c/c++ 设计模式

原文链接：https://my.oschina.net/lucasli/blog/1791018

版权

2019独角兽企业重金招聘Python工程师标准>>>

原文链接：https://mp.weixin.qq.com/s?__biz=MzIxMzY5ODQxMQ==&mid=2247483666&idx=1&sn=c7a76ca2524e26afed2c65eb12f6a0b2&chksm=97b39316a0c41a0085c42ca29181c5f739fe0abea8efb27c429aa4920244035d56d964f3ba67&mpshare=1&scene=22&srcid=04078Dg5WBaVB1zAl8nP8MRS#rd

个人觉得老外搞的这个PyalgoTrade，优点是按照设计模式的规范编写，扩展起来可能方便些。缺点是代码繁琐，效率较低。

如果你是初次接触还是建议了解国内的吧。

tushare就是国内搞的免费获取股票数据的类库，很好用，代码简介使用方便，一个get_k_data（）函数解决一切问题，函数名称也很秀气。

get_k_data可以获取国内几个大平台的免费股票数据，实时性虽然不行，但用在复盘回测上还是挺有用的。

PyalgoTrade内部提供的默认接口都是google或者yahoo国外的，在国内没法用。

下面首先介绍在pythong3.6(X64）版本上搭建。

1、基础类库安装：numpy、matplotlib、talib等，numpy, matplotlib安装简单可以通过python的pip install方式安装

talib稍微复杂一些，先要下载ta-lib（这个是C库）我使用VS2017编译，然后下载python封住的类库TA-Lib-0.4.16，下载tar.gz版本解压后通过命令python setup.py install方式安装，需要注意的时候c库的ta-lib需要复制到C盘，编译TA-LIB-0.4.16封装类库时需要在vs2017 64位编译环境下进行安装，如果选择默认编译环境会出现页内存不够的错误。

2、tushare和pyalgotrade我分别下载源码tushare-1.1.6.tar.gz和PyAlgoTrade-0.18.tar.gz目的是以后需要根据业务需求修改源码，扩展适合自己的回测平台。tushare采用python setup.py install即可安装成功，但是PyAlgoTrade是在Python2.X环境开发，不适合3.6，此时我们需要修改代码否则安装后测试程序无法运行。

3、PyAlgoTrade代码修改，目前我只简单做了几处BUG

1）首先需要安装3.6升级说明修改代码中关于filter, list,dict等iteritems变成items

2）filter, dict.getValues等这些在3.6版本里面会认为返回iterator,所以需要通过强制list等方式转换一下，否则会出现运行错误。

3）把调用sort的地方lamda参数修改位key=xxx

4）当然还有几处性能上的错误，比如把set()类型改为list类型，减少过多sort调用等

4、tushare增加到PyAlgoTrade

实现类barfeed / tsfeed.py文件，源码如下：

# coding="utf8"

#
# 这个文件是实现Tushare的数据处理，通过调用get_k_data()函数实现数据和Bars的绑定
#

from pyalgotrade.barfeed import membf
from pyalgotrade import bar
import tushare as ts
import datetime

def parse_date(date):
    # This custom parsing works faster than:
    # datetime.datetime.strptime(date, "%Y-%m-%d")
    year = int(date[0:4])
    month = int(date[5:7])
    day = int(date[8:10])
    d = datetime.datetime(year, month, day)
    if len(date) > 10:
        h = int(date[11:13])
        m = int(date[14:16])
        t = datetime.time(h,m)
        ret = datetime.combine(d,t)
    else:
         ret = d
    return ret

class Feed(membf.BarFeed):
    def __init__(self, frequency = bar.Frequency.DAY, maxLen=None):
        super(Feed, self).__init__(frequency, maxLen)

    def rowParser(self, ds, frequency=bar.Frequency.DAY):
        dt = parse_date(ds["date"])
        open = float(ds["open"])
        close = float(ds["close"])
        high = float(ds["high"])
        low = float(ds["low"])
        volume = float(ds["volume"])
        return bar.BasicBar(dt, open, high, low, close, volume, None, frequency)

    def barsHaveAdjClose(self):
        return False

    def addBarsFromCode(self, code, start, end, ktype="D", index=False):
        frequency = bar.Frequency.DAY
        if ktype == "D":
            frequency = bar.Frequency.DAY
        elif ktype == "W":
            frequency = bar.Frequency.WEEK
        elif ktype == "M":
            frequency = bar.Frquency.MONTH
        elif ktype == "5" or ktype == "15" or ktype == "30" or ktype == "60":
            frequency = bar.Frequency.MINUTE
        ds = ts.get_k_data(code = code, start = start, end = end, ktype = ktype, index = index)
        bars_ = []
        for i in ds.index:
            bar_ = self.rowParser(ds.loc[i], frequency)
            bars_.append(bar_)

        self.addBarsFromSequence(code, bars_)

5、下面是一个测试例子

# coding="utf8"

from pyalgotrade import strategy
from pyalgotrade.barfeed import tsfeed
from pyalgotrade import plotter
from pyalgotrade.technical import ma
from pyalgotrade.technical import cross
from pyalgotrade.stratanalyzer import returns

class MyStrategy(strategy.BacktestingStrategy):
    def __init__(self, feed, instrument, smaPeriod):
        super(MyStrategy, self).__init__(feed)
        self.__instrument = instrument
        self.__closed = feed[instrument].getCloseDataSeries()
        self.__ma = ma.SMA(self.__closed, smaPeriod)
        self.__position = None

    def getSMA(self):
        return self.__ma

    def onEnterLong(self, position):
        print("onEnterLong", position.getShares())

    def onEnterCanceled(self, position):
        self.__position = None
        print("onEnterCanceled", position.getShares())

    def onExitOk(self, position):
        self.__position = None
        print("onExitOk", position.getShares())

    def onExitCanceled(self, position):
        self.__position.exitMarket()
        print("onExitCanceled", position.getShares())

    def onBars(self, bars):
        if self.__position is None:
            if cross.cross_above(self.__closed, self.__ma) > 0:
                shares = int(self.getBroker().getCash() * 0.9 / bars[self.__instrument].getPrice())
                print("cross_above shares,", shares)
                # Enter a buy market order. The order is good till canceled.
                self.__position = self.enterLong(self.__instrument, shares, True)
        elif not self.__position.exitActive() and cross.cross_below(self.__closed, self.__ma) > 0:
            print("cross_below")
            self.__position.exitMarket()

def getClose(self):
return self.__closed

code = "603019"
feed = tsfeed.Feed()
feed.addBarsFromCode(code,start='2018-01-29',end='2018-04-04')

# Evaluate the strategy with the feed's bars.
myStrategy = MyStrategy(feed, code, 5)
returnsAnalyzer = returns.Returns()
myStrategy.attachAnalyzer(returnsAnalyzer)
plt = plotter.StrategyPlotter(myStrategy)
plt.getInstrumentSubplot(code).addDataSeries("SMA", myStrategy.getSMA())
plt.getOrCreateSubplot("returns").addDataSeries("Simple returns", returnsAnalyzer.getReturns())

myStrategy.run()

myStrategy.info("Final portfolio value: $%.2f" % myStrategy.getResult())

plt.plot()

运行结果图形如下：

640?wx_fmt=png&wxfrom=5&wx_lazy=1