用python炒股

時間 2020-01-19

標籤 python 炒股欄目 Python 简体版

原文原文鏈接

前言

因爲筆者並沒有深厚的數學功底也無深厚的金融知識, 因此不會在本文中引用各類高深的投資模型或數學模型，參考書籍主要是《海龜交易法則》《以交易爲生》。html

交易系統

在交易以前，咱們應該首先有一個交易系統用於指導咱們本身交易，不必定有什麼規範，可是能夠做爲一個交易的依據，至於這個依據可不可行，科不科學那就見仁見智了。python

固然了,這裏的交易系統不必定是程序，只是指你本身的交易原則或者遵照的一些技巧或者方法，你能夠手動執行也能夠藉助編程語言，編程語言不就是一套用來使用的工具麼.git

這裏參考海龜交易法則裏面的交易體系(這裏只是參考大方向).
創建一個完善的交易體系，咱們至少應該思考一下六個方面。github

一、市場----買賣什麼
二、頭寸規模----買賣多少
三、入市----什麼時候買入
四、止損----什麼時候退出虧損的頭寸
五、止盈----什麼時候退出盈利的頭寸
六、離市----什麼時候離市

簡單的示例

買賣A股
全倉
當日漲幅超過3%買入。
當持有頭寸虧損超過3%，平倉
當日跌幅大於3%或者三個連續陰線

分析: 這個交易策略其實只有在行情以波浪形狀向上的行情時候才能獲利，若是是盤整的狀況下，怕是會虧的很慘。這裏之因此寫的這麼簡單粗暴是爲了後面策略測試擼代碼簡單。編程

數據獲取及處理

由於這裏說的是用python炒股，因此應該採用程序的方式去獲取數據，若是人工炒股，下載任何股票行情軟件都是能夠的，可是人工的執行是須要花費比較多的精力的。api

而python語言中用於獲取股票行情數據的庫，最有名莫過於tushare了。微信

這裏以上證樂視的股票爲例吧。框架

python環境安裝

安裝Anaconda(python2版本)

下載地址:https://www.anaconda.com/download/
注:若是沒安裝過這個環境的經驗，就百度或者谷歌一下吧,若是不是安裝anaconda則須要艱難的自行解決依賴。編程語言

安裝tushare

pip install tushare

參考: http://tushare.org/ide

獲取行情數據

import pandas as pd
import tushare as ts

# 經過股票代碼獲取股票數據,這裏沒有指定開始及結束日期
df = ts.get_k_data("300104")

# 查看前十條數據
df.head()

# 查看後十條數據
df.tail()

# 將數據的index轉換成date字段對應的日期
df.index = pd.to_datetime(df.date)

# 將多餘的date字段刪除
df.drop("date", inplace=True, axis=1)

注:關於股票數據的相關處理須要由pandas,matplotlib的知識，參考:http://pandas.pydata.org/pandas-docs/version/0.20/10min.html

計算經常使用指標

# 計算5,15,50日的移動平均線, MA5, MA15, MA50
days = [5, 15, 50]
for ma in days:
    column_name = "MA{}".format(ma)
    df[column_name] = pd.rolling_mean(df.close, ma)

# 計算浮動比例
df["pchange"] = df.close.pct_change()
# 計算浮動點數
df["change"] = df.close.diff()

最終處理完成後的結果以下:

df.head()
Out[13]: 
             open  close   high    low    volume    code     MA5  MA15  MA50  \
date                                                                           
2013-11-29  9.396  9.741  9.870  9.389  146587.0  300104     NaN   NaN   NaN   
2013-12-02  9.298  8.768  9.344  8.768  177127.0  300104     NaN   NaN   NaN   
2013-12-03  8.142  8.414  8.546  7.890  176305.0  300104     NaN   NaN   NaN   
2013-12-04  8.391  8.072  8.607  8.053  120115.0  300104     NaN   NaN   NaN   
2013-12-05  7.983  7.366  8.108  7.280  253764.0  300104  8.4722   NaN   NaN   

             pchange  change  
date                          
2013-11-29       NaN     NaN  
2013-12-02 -0.099887  -0.973  
2013-12-03 -0.040374  -0.354  
2013-12-04 -0.040647  -0.342

可視化

走勢圖

所謂一圖勝前言，將數據可視化能夠很是直觀的感覺到股票的走勢。
我的以爲，若是用程序炒股仍是應該一切都量化的，不該該有過多的主觀觀點，若是過於依賴直覺或者當時心情，那麼實在不必用程序分析了。

df[["close", "MA5", "MA15", "MA50"]].plot(figsiz=(10,18))

效果以下:

k線圖

import matplotplib.pyplot as plt
from matplotlib.daet import DateFormatter
from matplotlib.finance import date2num, candlestick_ohlc

def candlePlot(data, title=""):
    data["date"] = [date2num(pd.to_datetime(x)) for x in data.index]
    dataList = [tuple(x) for x in data[
        ["date", "open", "high", "low", "close"]].values]

    ax = plt.subplot()
    ax.set_title(title)
    ax.xaxis.set_major_formatter(DateFormatter("%y-%m-%d"))
    candlestick_ohlc(ax, dataList, width=0.7, colorup="r", colordown="g")
    plt.setp(plt.gca().get_xticklabels(), rotation=50,
             horizontalalignment="center")
    fig = plt.gcf()
    fig.set_size_inches(20, 15)
    plt.grid(True)

candlePlot(df)

效果以下:

注: 這裏只是一個示例，說明matplotlib的強大以及小小的演示，若是遇到什麼奇怪的問題就查api或者google吧。

策略測試

手動擼代碼

這裏用最近買過的一隻股票吧，京東方A(000725)。

# 導入相關模塊
import tushare as ts
import pandas as pd

# 獲取數據
df = ts.get_k_data("000725")

# 處理數據
df.index = pd.to_datetime(df.date)
df.drop("date", axis=1, inplace=True)

# 計算浮動比例
df["pchange"] = df.close.pct_change()
# 計算浮動點數
df["change"] = df.close.diff()

# 查看當前數據數據前五行
             open  close   high    low      volume    code   pchange  change
date                                                                        
2015-07-20  4.264  4.234  4.342  4.165  13036186.0  000725       NaN     NaN
2015-07-21  4.136  4.195  4.274  4.096   8776773.0  000725 -0.009211  -0.039
2015-07-22  4.175  4.146  4.214  4.067   9083703.0  000725 -0.011681  -0.049
2015-07-23  4.136  4.254  4.283  4.096  12792734.0  000725  0.026049   0.108
2015-07-24  4.224  4.136  4.254  4.106  13009620.0  000725 -0.027739  -0.118

# 設定回撤值
withdraw = 0.03
# 設定突破值
breakthrough = 0.03
# 設定帳戶資金
account = 10000
# 持有倉位手數
position = 0

def buy(bar):
    global account, position
    print("{}: buy {}".format(bar.date, bar.close))
    # 一手價格
    one = bar.close * 100
    position = account // one
    account = account - (position * one)

def sell(bar):
    global account, position
    # 一手價格
    print("{}: sell {}".format(bar.date, bar.close))
    one = bar.close * 100
    account += position * one
    position = 0

print("開始時間投資時間: ", df.iloc[0].date)
for date in df.index:
    bar = df.loc[date]
    if bar.pchange and bar.pchange > breakthrough and position == 0:
        buy(bar)
    elif bar.pchange and bar.pchange < withdraw and position > 0:
        sell(bar)

print("最終可有現金: ", account)
print("最終持有市值: ", position * df.iloc[-1].close * 100)

輸出以下:

開始時間投資時間:  2015-07-20
2015-07-29: buy 3.83
2015-07-30: sell 3.653
2015-08-04: buy 3.752
......
2018-02-27: sell 5.71
2018-03-06: buy 5.79
最終可有現金:  333.3
最終持有市值:  7527.0

結論: 經過上面的測試發現資虧了兩千多...

藉助測試框架

藉助測試框架纔是正確的回撤姿式,由於框架包含了更多的功能。這裏使用pyalgotrade。

簡單使用

from pyalgotrade import strategy
from pyalgotrade import technical
from pyalgotrade.barfeed import yahoofeed

# 自定義事件窗口類
class DiffEventWindow(technical.EventWindow):
    def __init__(self, period):
        assert(period > 0)
        super(DiffEventWindow, self).__init__(period)
        self.__value = None

    def onNewValue(self, dateTime, value):
        super(DiffEventWindow, self).onNewValue(dateTime, value)
        if self.windowFull():
            lastValue = self.getValues()[0]
            nowValue = self.getValues()[1]
            self.__value = (nowValue - lastValue) / lastValue

    def getValue(self):
        return self.__value

# 自定義指標
class Diff(technical.EventBasedFilter):
    def __init__(self, dataSeries, period, maxLen=None):
        super(Diff, self).__init__(dataSeries, DiffEventWindow(period), maxLen)

# 定義本身的策略
class MyStrategy(strategy.BacktestingStrategy):
    def __init__(self, feed, instrument, diffPeriod=2):
        # 傳入feed及初始帳戶資金
        super(MyStrategy, self).__init__(feed, 10000)
        self.__instrument = instrument
        self.__position = None
        self.setUseAdjustedValues(True)
        self.__prices = feed[instrument].getPriceDataSeries()
        self.__diff = Diff(self.__prices, diffPeriod)
        self.__break = 0.03
        self.__withdown = -0.03

    def getDiff(self):
        return self.__diff

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

    def onEnterOk(self, position):
        execInfo = position.getEntryOrder().getExecutionInfo()
        self.info("BUY at $%.2f" % (execInfo.getPrice()))

    def onExitOk(self, position):
        execInfo = position.getExitOrder().getExecutionInfo()
        self.info("SELL at $%.2f" % (execInfo.getPrice()))
        self.__position = None

    def onExitCanceled(self, position):
        # If the exit was canceled, re-submit it.
        self.__position.exitMarket()

    def onBars(self, bars):
        account = self.getBroker().getCash()
        bar = bars[self.__instrument]
        if self.__position is None:
            one = bar.getPrice() * 100
            oneUnit = account // one
            if oneUnit > 0 and self.__diff[-1] > self.__break:
                self.__position = self.enterLong(self.__instrument, oneUnit * 100, True)
        elif self.__diff[-1] < self.__withdown and not self.__position.exitActive():
            self.__position.exitMarket()

def runStrategy():
    # 下載數據
    jdf = ts.get_k_data("000725")

    # 新建Adj Close字段
    jdf["Adj Close"] =jdf.close

    # 將tushare下的數據的字段保存爲pyalgotrade所要求的數據格式
    jdf.columns = ["Date", "Open", "Close", "High", "Low", "Volume", "code", "Adj Close"]

    # 將數據保存成本地csv文件
    jdf.to_csv("jdf.csv", index=False)

    feed = yahoofeed.Feed()
    feed.addBarsFromCSV("jdf", "jdf.csv")

    myStrategy = MyStrategy(feed, "jdf")

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

runStrategy()

輸出以下

2015-07-30 00:00:00 strategy [INFO] BUY at $3.78
2015-07-31 00:00:00 strategy [INFO] SELL at $3.57
2015-08-05 00:00:00 strategy [INFO] BUY at $3.73
2015-08-06 00:00:00 strategy [INFO] SELL at $3.56
...
2018-02-13 00:00:00 strategy [INFO] BUY at $5.45
Final portfolio value: $7877.30

猛地一看會發現，用框架彷佛寫了更多的代碼,可是框架內置了更多分析工具。
下面簡單介紹。

策略可視化

from pyalgotrade import strategy
from pyalgotrade import technical
from pyalgotrade.barfeed import yahoofeed
from pyalgotrade import plotter
from pyalgotrade.stratanalyzer import returns

class DiffEventWindow(technical.EventWindow):
    def __init__(self, period):
        assert(period > 0)
        super(DiffEventWindow, self).__init__(period)
        self.__value = None

    def onNewValue(self, dateTime, value):
        super(DiffEventWindow, self).onNewValue(dateTime, value)
        if self.windowFull():
            lastValue = self.getValues()[0]
            nowValue = self.getValues()[1]
            self.__value = (nowValue - lastValue) / lastValue

    def getValue(self):
        return self.__value

class Diff(technical.EventBasedFilter):
    def __init__(self, dataSeries, period, maxLen=None):
        super(Diff, self).__init__(dataSeries, DiffEventWindow(period), maxLen)

class MyStrategy(strategy.BacktestingStrategy):
    def __init__(self, feed, instrument, diffPeriod=2):
        super(MyStrategy, self).__init__(feed, 10000)
        self.__instrument = instrument
        self.__position = None
        self.setUseAdjustedValues(True)
        self.__prices = feed[instrument].getPriceDataSeries()
        self.__diff = Diff(self.__prices, diffPeriod)
        self.__break = 0.03
        self.__withdown = -0.03

    def getDiff(self):
        return self.__diff

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

    def onEnterOk(self, position):
        execInfo = position.getEntryOrder().getExecutionInfo()
        self.info("BUY at $%.2f" % (execInfo.getPrice()))

    def onExitOk(self, position):
        execInfo = position.getExitOrder().getExecutionInfo()
        self.info("SELL at $%.2f" % (execInfo.getPrice()))
        self.__position = None

    def onExitCanceled(self, position):
        # If the exit was canceled, re-submit it.
        self.__position.exitMarket()

    def onBars(self, bars):
        account = self.getBroker().getCash()
        bar = bars[self.__instrument]
        if self.__position is None:
            one = bar.getPrice() * 100
            oneUnit = account // one
            if oneUnit > 0 and self.__diff[-1] > self.__break:
                self.__position = self.enterLong(self.__instrument, oneUnit * 100, True)
        elif self.__diff[-1] < self.__withdown and not self.__position.exitActive():
            self.__position.exitMarket()

def runStrategy():
    # 下載數據
    jdf = ts.get_k_data("000725")

    # 新建Adj Close字段
    jdf["Adj Close"] =jdf.close

    # 將tushare下的數據的字段保存爲pyalgotrade所要求的數據格式
    jdf.columns = ["Date", "Open", "Close", "High", "Low", "Volume", "code", "Adj Close"]

    # 將數據保存成本地csv文件
    jdf.to_csv("jdf.csv", index=False)

    feed = yahoofeed.Feed()
    feed.addBarsFromCSV("jdf", "jdf.csv")

    myStrategy = MyStrategy(feed, "jdf")

    returnsAnalyzer = returns.Returns()
    myStrategy.attachAnalyzer(returnsAnalyzer)
    plt = plotter.StrategyPlotter(myStrategy)
    plt.getInstrumentSubplot("jdf")
    plt.getOrCreateSubplot("returns").addDataSeries("Simple returns", returnsAnalyzer.getReturns())

    myStrategy.run()
    print("Final portfolio value: $%.2f" % myStrategy.getResult())
    plt.plot()

runStrategy()

圖片輸出以下

注: 這裏的策略測試股票選擇以及時間選擇並不嚴謹，僅做功能展現，測試結果可能有很大的巧合性。Pyalgotrade詳細介紹皆使用參考:http://gbeced.github.io/pyalgotrade/docs/v0.18/html/index.html
上述源代碼:https://github.com/youerning/blog/blob/master/python-trade/demo.py