python實作馬丁策略

馬丁策略本來是一種賭博方法，但在投資界應用也很廣泛，不過對于投資者來說馬丁策略過于簡單，所以本文將其改進并使得其在震蕩市中獲利，以下說明如何實作馬丁策略，

策略

逢跌加倉，間隔由自己決定，每次加倉是當前倉位的一倍，
連續跌兩次賣出，且賣出一半倉位，
如果爆倉則全倉賣出止損，
初始持倉設定為10%~25%，則可進行2到3次補倉，

初始化馬丁策略類屬性

    def __init__(self,startcash, start, end):
        self.cash = startcash #初始化現金
        self.hold = 0  #初始化持倉金額
        self.holdper = self.hold /startcash  #初始化倉位
        self.log = []  #初始化日志
        self.cost = 0 #成本價 
        self.stock_num = 0 #股票數量
        self.starttime = start #起始時間
        self.endtime = end #終止時間
        self.quantlog = [] #交易量記錄
        self.earn = []  #總資產記錄
        self.num_log = []
        self.droplog = [0]

為了記錄每次買賣倉位的變化初始化了各種串列，

交易函式

首先匯入需要的模塊

import pandas as pd  
import numpy as np
import tushare as ts 
import matplotlib.pyplot as plt

    def buy(self, currentprice, count):

        self.cash -= currentprice*count
        self.log.append('buy')
        self.hold += currentprice*count
        self.holdper = self.hold / (self.cash+ self.hold) 
        self.stock_num += count
        self.cost = self.hold / self.stock_num
        self.quantlog.append(count//100)
        print('買入價：%.2f，手數：%d，現在成本價：%.2f，現在持倉：%.2f，現在籌碼：%d' %(currentprice ,count//100, self.cost, self.holdper, self.stock_num//100))
        self.earn.append(self.cash+ currentprice*self.stock_num)
        self.num_log.append(self.stock_num)
        self.droplog = [0]
        
    def sell(self, currentprice, count):
        self.cash += currentprice*count
        self.stock_num -= count
        self.log.append('sell')
        self.hold = self.stock_num*self.cost
        self.holdper = self.hold / (self.cash + self.hold)
        #self.cost = self.hold / self.stock_num
        print('賣出價：%.2f，手數：%d，現在成本價：%.2f，現在持倉：%.2f，現在籌碼：%d' %(currentprice ,count//100, self.cost, self.holdper, self.stock_num//100))
        self.quantlog.append(count//100)                            
        self.earn.append(self.cash+ currentprice*self.stock_num)
        self.num_log.append(self.stock_num)
        
    def holdstock(self,currentprice):
        self.log.append('hold')
        #print('持有，現在倉位為：%.2f，現在成本：%.2f' %(self.holdper,self.cost))
        self.quantlog.append(0)
        self.earn.append(self.cash+ currentprice*self.stock_num)
        self.num_log.append(self.stock_num)

持倉成本的計算方式是利用總持倉金額除以總手數，賣出時不改變持倉成本，持有則是不做任何操作只記錄日志

資料介面

    def get_stock(self, code):
        df=ts.get_k_data(code,autype='qfq',start= self.starttime ,end= self.endtime)
        df.index=pd.to_datetime(df.date)
        df=df[['open','high','low','close','volume']]
        return df

資料介面使用tushare，也可使用pro介面，到官網注冊領取token，

token = '輸入你的token'
pro = ts.pro_api()
ts.set_token(token)
    def get_stock_pro(self, code):
        code = code + '.SH'
        df = pro.daily(ts_code= code, start_date = self.starttime, end_date= self.endtime)
        return df

資料結構：

回測函式

    def startback(self, data, everyChange, accDropday):
        """
        回測函式
        """
        for i in range(len(data)):
            if i < 1:
                continue
            if  i < accDropday:
                drop = backtesting.accumulateVar(everyChange, i, i)
                #print('現在累計漲跌幅度為：%.2f'%(drop))
                self.martin(data[i], data[i-1], drop, everyChange,i)
            elif i < len(data)-2:
                drop = backtesting.accumulateVar(everyChange, i, accDropday)
                #print('現在累計漲跌幅度為：%.2f'%(drop))
                self.martin(data[i],data[i-1], drop, everyChange,i)
            else:
                if self.stock_num > 0:
                    self.sell(data[-1],self.stock_num)
                else: self.holdstock(data[i])

因為要計算每日漲跌幅，要計算差分，所以第一天的資料不能計算在for回圈中跳過，accDropday是累計跌幅的最大計算天數，用來控制入場，當累計跌幅大于某個數值且倉位為0%時可再次入場，以下是入場函式：

    def enter(self, currentprice,ex_price,accuDrop):
        if accuDrop < -0.01:#and ex_price > currentprice:
            count = (self.cash+self.hold) *0.24 // currentprice //100 * 100
            print('再次入場')
            self.buy(currentprice, count)
        else: self.holdstock(currentprice)
        

入場倉位選擇0.24則可進行兩次抄底，如果抄底間隔為7%可承受最大跌幅為14%，

策略函式

    def martin(self, currentprice, ex_price, accuDrop,everyChange,i):
        diff = (ex_price - currentprice)/ex_price
        self.droplog.append(diff)

        if sum(self.droplog) <= 0:
            self.droplog = [0]
        
        if self.stock_num//100 > 1:
            if sum(self.droplog) >= 0.04:
                if self.holdper*2 < 0.24:
                    count =(self.cash+self.hold) *(0.25-self.holdper) // currentprice //100 * 100
                    self.buy(currentprice, count)
                elif self.holdper*2 < 1 and (self.hold/currentprice)//100 *100 > 0 and backtesting.computeCon(self.log) < 5:
                    self.buy(currentprice, (self.hold/currentprice)//100 *100)
                    
                else: self.sell(currentprice, self.stock_num//100 *100);print('及時止損')

            elif (everyChange[i-2] < 0 and everyChange[i-1] <0 and self.cost < currentprice):# or (everyChange[i-1] < -0.04 and self.cost < currentprice):
                    
                if (self.stock_num > 0) and ((self.stock_num*(1/2)//100*100) > 0):
                        
                    self.sell(currentprice, self.stock_num*(1/2)//100*100 )


                    #print("現在累計漲跌幅為： %.3f" %(accuDrop))
                elif self.stock_num == 100: self.sell(currentprice, 100)
                else: self.holdstock(currentprice)
            else: self.holdstock(currentprice)
        else: self.enter(currentprice,ex_price,accuDrop)

首先構建了droplog專門用于計算累計漲跌幅，當其大于0時重置為0，每次購買后也將其重置為0，當跌幅大于0.04則買入，一下為流程圖（因為作圖軟體Visustin為試用版所以有水印，兩個圖可以結合來看）：


此策略函式可以改成其他策略甚至是反馬丁，因為交易函式可以通用，

作圖和輸出結果

buylog = pd.Series(broker.log)
close = data.copy()
buy = np.zeros(len(close))
sell = np.zeros(len(close))
for i in range(len(buylog)):
    if buylog[i] == 'buy':
        buy[i] = close[i]
    elif buylog[i] == 'sell':
        sell[i] = close[i]

buy = pd.Series(buy)
sell = pd.Series(sell)
buy.index = close.index
sell.index = close.index
quantlog = pd.Series(broker.quantlog)
quantlog.index = close.index
earn = pd.Series(broker.earn)
earn.index = close.index

buy = buy.loc[buy > 0]
sell = sell.loc[sell>0]
plt.plot(close)
plt.scatter(buy.index,buy,label = 'buy')
plt.scatter(sell.index,sell, label = 'sell')
plt.title('馬丁策略')
plt.legend()



#畫圖
plt.rcParams['font.sans-serif'] = ['SimHei']

fig, (ax1, ax2, ax3) = plt.subplots(3,figsize=(15,8))

ax1.plot(close)
ax1.scatter(buy.index,buy,label = 'buy',color = 'red')
ax1.scatter(sell.index,sell, label = 'sell',color = 'green')
ax1.set_ylabel('Price')
ax1.grid(True)
ax1.legend()

ax1.xaxis_date()
ax2.bar(quantlog.index, quantlog, width = 5)
ax2.set_ylabel('Volume')

ax2.xaxis_date()
ax2.grid(True)
ax3.xaxis_date()
ax3.plot(earn)
ax3.set_ylabel('總資產包括浮盈')
plt.show()

交易日志