(Python을 이용한 Quant 투자) VAA 백테스트 : using python

Posted by : at

Category : Python


필요 데이터 추출

import pandas_datareader as pdr 
import pandas as pd # from datetime 
import datetime, timedelta 
import matplotlib.pyplot as plt 
import seaborn as sns 

pd.options.display.float_format = '{:.2f}'.format 

start_day = datetime.date(2009,1,1) # 시작일 
end_day = datetime.date(2021,11,13) # 종료일 

# 공격자산 
SPY = pdr.get_data_yahoo('SPY', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
VEA = pdr.get_data_yahoo('VEA', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
EEM = pdr.get_data_yahoo('EEM', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
AGG = pdr.get_data_yahoo('AGG', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 

# 수비자산 
LQD = pdr.get_data_yahoo('LQD', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
SHY = pdr.get_data_yahoo('SHY', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
IEF = pdr.get_data_yahoo('IEF', start_day - datetime.timedelta(days=365), end_day)['Adj Close']

df_VAA = pd.concat([SPY,VEA,EEM,AGG,LQD,SHY,IEF],axis=1) 
df_VAA.columns = ['SPY','VEA','EEM','AGG','LQD','SHY','IEF'] 
df_VAA.head(5)

print(df_VAA)

모멘텀 계산 및 파일 추출

import pandas_datareader as pdr 
import pandas as pd
import datetime, timedelta 
import matplotlib.pyplot as plt 
import seaborn as sns 

# 모멘텀 지수 계산 함수 
def get_momentum(df_VAA, x): 
    momentum = pd.Series([0,0,0,0,0,0,0], index=['SPY','VEA','EEM','AGG','LQD','SHY','IEF']) 

    try: 
        # df_VAA[x.name-datetime.timedelta(days=35):x.name-datetime.timedelta(days=30)].iloc[-1] 하는 이유?
        # X달 전 데이터를 잘못 가져올 경우를 대비 5일치를 가져와서 마지막 행만 읽어준다
        before1 = df_VAA[x.name-datetime.timedelta(days=35):x.name-datetime.timedelta(days=30)].iloc[-1] 
        before3 = df_VAA[x.name-datetime.timedelta(days=95):x.name-datetime.timedelta(days=90)].iloc[-1] 
        before6 = df_VAA[x.name-datetime.timedelta(days=185):x.name-datetime.timedelta(days=180)].iloc[-1] 
        before12 = df_VAA[x.name-datetime.timedelta(days=370):x.name-datetime.timedelta(days=365)].iloc[-1] 
           
        momentum = 12 * (x / before1 - 1) + 4 * (x / before3 - 1) + 2 * (x / before6 - 1) + (x / before12 - 1) 
    except: 
        pass 
    
    return momentum

def main():
    pd.options.display.float_format = '{:.2f}'.format 

    start_day = datetime.date(2009,1,2) # 시작일 
    end_day = datetime.date(2021,11,13) # 종료일 

    # 공격자산 
    SPY = pdr.get_data_yahoo('SPY', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    VEA = pdr.get_data_yahoo('VEA', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    EEM = pdr.get_data_yahoo('EEM', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    AGG = pdr.get_data_yahoo('AGG', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 

    # 수비자산 
    LQD = pdr.get_data_yahoo('LQD', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    SHY = pdr.get_data_yahoo('SHY', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    IEF = pdr.get_data_yahoo('IEF', start_day - datetime.timedelta(days=365), end_day)['Adj Close']

    df_VAA = pd.concat([SPY,VEA,EEM,AGG,LQD,SHY,IEF],axis=1) 
    df_VAA.columns = ['SPY','VEA','EEM','AGG','LQD','SHY','IEF'] 
    #df_VAA.head(5)

    # 각 자산별 모멘텀 지수 계산 
    df_VAA[['SPY_M','VEA_M','EEM_M','AGG_M','LQD_M','SHY_M','IEF_M']] = df_VAA.apply(lambda x: get_momentum(df_VAA, x), axis=1) 
    df_VAA.tail(10)

    print(df_VAA) # print to Console
    df_VAA.to_csv('./outPut.csv', sep=',', na_rep="NaN") # printf to File


if __name__ == "__main__":
    main()

말일 데이터만 뽑아보자

import pandas_datareader as pdr 
import pandas as pd
import datetime, timedelta 
import matplotlib.pyplot as plt 
import seaborn as sns 

# 모멘텀 지수 계산 함수 
def get_momentum(df_VAA, x): 
    momentum = pd.Series([0,0,0,0,0,0,0], index=['SPY','VEA','EEM','AGG','LQD','SHY','IEF']) 

    try: 
        # df_VAA[x.name-datetime.timedelta(days=35):x.name-datetime.timedelta(days=30)].iloc[-1] 하는 이유?
        # X달 전 데이터를 잘못 가져올 경우를 대비 5일치를 가져와서 마지막 행만 읽어준다
        before1 = df_VAA[x.name-datetime.timedelta(days=35):x.name-datetime.timedelta(days=30)].iloc[-1] 
        before3 = df_VAA[x.name-datetime.timedelta(days=95):x.name-datetime.timedelta(days=90)].iloc[-1] 
        before6 = df_VAA[x.name-datetime.timedelta(days=185):x.name-datetime.timedelta(days=180)].iloc[-1] 
        before12 = df_VAA[x.name-datetime.timedelta(days=370):x.name-datetime.timedelta(days=365)].iloc[-1] 
           
        momentum = 12 * (x / before1 - 1) + 4 * (x / before3 - 1) + 2 * (x / before6 - 1) + (x / before12 - 1) 
    except: 
        pass 
    
    return momentum

def main():
    pd.options.display.float_format = '{:.2f}'.format 

    start_day = datetime.date(2009,1,2) # 시작일 
    end_day = datetime.date(2021,11,13) # 종료일 

    # 공격자산 
    SPY = pdr.get_data_yahoo('SPY', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    VEA = pdr.get_data_yahoo('VEA', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    EEM = pdr.get_data_yahoo('EEM', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    AGG = pdr.get_data_yahoo('AGG', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 

    # 수비자산 
    LQD = pdr.get_data_yahoo('LQD', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    SHY = pdr.get_data_yahoo('SHY', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    IEF = pdr.get_data_yahoo('IEF', start_day - datetime.timedelta(days=365), end_day)['Adj Close']

    df_VAA = pd.concat([SPY,VEA,EEM,AGG,LQD,SHY,IEF],axis=1) 
    df_VAA.columns = ['SPY','VEA','EEM','AGG','LQD','SHY','IEF'] 
    #df_VAA.head(5)

    # 각 자산별 모멘텀 지수 계산 
    df_VAA[['SPY_M','VEA_M','EEM_M','AGG_M','LQD_M','SHY_M','IEF_M']] = df_VAA.apply(lambda x: get_momentum(df_VAA, x), axis=1) 
    #df_VAA.tail(10)

    # 백테스트할 기간 데이터 추출 
    df_VAA = df_VAA[start_day:end_day] 
    
    # 매월 말일 데이터만 추출(리밸런싱에 사용) 
    df_VAA = df_VAA.resample(rule='M').apply(lambda x: x[-1]) 
    df_VAA.head(10)

    #print(df_VAA) # print to Console
    df_VAA.to_csv('./outPut.csv', sep=',', na_rep="NaN") # printf to File


if __name__ == "__main__":
    main()

월별 구매자산 지정


import pandas_datareader as pdr 
import pandas as pd
import datetime, timedelta 
import matplotlib.pyplot as plt 
import seaborn as sns 

# 모멘텀 지수 계산 함수 
def get_momentum(df_VAA, x): 
    momentum = pd.Series([0,0,0,0,0,0,0], index=['SPY','VEA','EEM','AGG','LQD','SHY','IEF']) 

    try: 
        # df_VAA[x.name-datetime.timedelta(days=35):x.name-datetime.timedelta(days=30)].iloc[-1] 하는 이유?
        # X달 전 데이터를 잘못 가져올 경우를 대비 5일치를 가져와서 마지막 행만 읽어준다
        before1 = df_VAA[x.name-datetime.timedelta(days=35):x.name-datetime.timedelta(days=30)].iloc[-1] 
        before3 = df_VAA[x.name-datetime.timedelta(days=95):x.name-datetime.timedelta(days=90)].iloc[-1] 
        before6 = df_VAA[x.name-datetime.timedelta(days=185):x.name-datetime.timedelta(days=180)].iloc[-1] 
        before12 = df_VAA[x.name-datetime.timedelta(days=370):x.name-datetime.timedelta(days=365)].iloc[-1] 
           
        momentum = 12 * (x / before1 - 1) + 4 * (x / before3 - 1) + 2 * (x / before6 - 1) + (x / before12 - 1) 
    except: 
        pass 
    
    return momentum

# VAA 전략 기준에 맞춰 자산 선택 
def select_asset(x): 
    asset = pd.Series([0,0], index=['ASSET','PRICE']) 
    
    # 공격 자산이 모두 0이상이면, 공격 자산 중 최고 모멘텀 자산 선정 
    if x['SPY_M'] > 0 and x['VEA_M'] > 0 and x['EEM_M'] > 0 and x['AGG_M'] > 0: 
        max_momentum = max(x['SPY_M'],x['VEA_M'],x['EEM_M'],x['AGG_M']) 
        
    # 공격 자산 중 하나라도 0이하라면, 방어 자산 중 최고 모멘텀 자산 선정 
    else : 
        max_momentum = max(x['LQD_M'],x['SHY_M'],x['IEF_M']) 
        
    asset['ASSET'] = x[x == max_momentum].index[0][:3] 
    asset['PRICE'] = x[asset['ASSET']] 
    
    return asset


def main():
    pd.options.display.float_format = '{:.2f}'.format 

    start_day = datetime.date(2009,1,2) # 시작일 
    end_day = datetime.date(2021,11,13) # 종료일 

    # 공격자산 
    SPY = pdr.get_data_yahoo('SPY', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    VEA = pdr.get_data_yahoo('VEA', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    EEM = pdr.get_data_yahoo('EEM', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    AGG = pdr.get_data_yahoo('AGG', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 

    # 수비자산 
    LQD = pdr.get_data_yahoo('LQD', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    SHY = pdr.get_data_yahoo('SHY', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    IEF = pdr.get_data_yahoo('IEF', start_day - datetime.timedelta(days=365), end_day)['Adj Close']

    df_VAA = pd.concat([SPY,VEA,EEM,AGG,LQD,SHY,IEF],axis=1) 
    df_VAA.columns = ['SPY','VEA','EEM','AGG','LQD','SHY','IEF'] 
    #df_VAA.head(5)

    # 각 자산별 모멘텀 지수 계산 
    df_VAA[['SPY_M','VEA_M','EEM_M','AGG_M','LQD_M','SHY_M','IEF_M']] = df_VAA.apply(lambda x: get_momentum(df_VAA, x), axis=1) 
    #df_VAA.tail(10)

    # 백테스트할 기간 데이터 추출 
    df_VAA = df_VAA[start_day:end_day] 
    
    # 매월 말일 데이터만 추출(리밸런싱에 사용) 
    df_VAA = df_VAA.resample(rule='M').apply(lambda x: x[-1]) 
    df_VAA.head(10)

    # 매월 선택할 자산과 가격 
    df_VAA[['ASSET','PRICE']] = df_VAA.apply(lambda x: select_asset(x), axis=1) 
    df_VAA.head(10)

    print(df_VAA) # print to Console
    df_VAA.to_csv('./outPut.csv', sep=',', na_rep="NaN") # printf to File


if __name__ == "__main__":
    main()

누적수익률 및 누적수익 그래프 그리기


import pandas_datareader as pdr 
import pandas as pd
import datetime, timedelta 
import matplotlib.pyplot as plt 
import seaborn as sns 

# 모멘텀 지수 계산 함수 
def get_momentum(df_VAA, x): 
    momentum = pd.Series([0,0,0,0,0,0,0], index=['SPY','VEA','EEM','AGG','LQD','SHY','IEF']) 

    try: 
        # df_VAA[x.name-datetime.timedelta(days=35):x.name-datetime.timedelta(days=30)].iloc[-1] 하는 이유?
        # X달 전 데이터를 잘못 가져올 경우를 대비 5일치를 가져와서 마지막 행만 읽어준다
        before1 = df_VAA[x.name-datetime.timedelta(days=35):x.name-datetime.timedelta(days=30)].iloc[-1] 
        before3 = df_VAA[x.name-datetime.timedelta(days=95):x.name-datetime.timedelta(days=90)].iloc[-1] 
        before6 = df_VAA[x.name-datetime.timedelta(days=185):x.name-datetime.timedelta(days=180)].iloc[-1] 
        before12 = df_VAA[x.name-datetime.timedelta(days=370):x.name-datetime.timedelta(days=365)].iloc[-1] 
           
        momentum = 12 * (x / before1 - 1) + 4 * (x / before3 - 1) + 2 * (x / before6 - 1) + (x / before12 - 1) 
    except: 
        pass 
    
    return momentum

# VAA 전략 기준에 맞춰 자산 선택 
def select_asset(x): 
    asset = pd.Series([0,0], index=['ASSET','PRICE']) 
    
    # 공격 자산이 모두 0이상이면, 공격 자산 중 최고 모멘텀 자산 선정 
    if x['SPY_M'] > 0 and x['VEA_M'] > 0 and x['EEM_M'] > 0 and x['AGG_M'] > 0: 
        max_momentum = max(x['SPY_M'],x['VEA_M'],x['EEM_M'],x['AGG_M']) 
        
    # 공격 자산 중 하나라도 0이하라면, 방어 자산 중 최고 모멘텀 자산 선정 
    else : 
        max_momentum = max(x['LQD_M'],x['SHY_M'],x['IEF_M']) 
        
    asset['ASSET'] = x[x == max_momentum].index[0][:3] 
    asset['PRICE'] = x[asset['ASSET']] 
    
    return asset


def main():
    pd.options.display.float_format = '{:.2f}'.format 

    start_day = datetime.date(2010,1,1) # 시작일 
    end_day = datetime.date(2021,9,22) # 종료일 

    # 공격자산 
    SPY = pdr.get_data_yahoo('SPY', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    VEA = pdr.get_data_yahoo('VEA', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    EEM = pdr.get_data_yahoo('EEM', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    AGG = pdr.get_data_yahoo('AGG', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 

    # 수비자산 
    LQD = pdr.get_data_yahoo('LQD', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    SHY = pdr.get_data_yahoo('SHY', start_day - datetime.timedelta(days=365), end_day)['Adj Close'] 
    IEF = pdr.get_data_yahoo('IEF', start_day - datetime.timedelta(days=365), end_day)['Adj Close']

    df_VAA = pd.concat([SPY,VEA,EEM,AGG,LQD,SHY,IEF],axis=1) 
    df_VAA.columns = ['SPY','VEA','EEM','AGG','LQD','SHY','IEF'] 
    #df_VAA.head(5)

    # 각 자산별 모멘텀 지수 계산 
    df_VAA[['SPY_M','VEA_M','EEM_M','AGG_M','LQD_M','SHY_M','IEF_M']] = df_VAA.apply(lambda x: get_momentum(df_VAA, x), axis=1) 
    #df_VAA.tail(10)

    # 백테스트할 기간 데이터 추출 
    df_VAA = df_VAA[start_day:end_day] 
    
    # 매월 말일 데이터만 추출(리밸런싱에 사용) 
    df_VAA = df_VAA.resample(rule='M').apply(lambda x: x[-1]) 
    # df_VAA.head(10)

    # 매월 선택할 자산과 가격 
    df_VAA[['ASSET','PRICE']] = df_VAA.apply(lambda x: select_asset(x), axis=1) 
    # df_VAA.head(10)

    # 매월 수익률 & 누적 수익률 계산 
    df_VAA['PROFIT'] = 0 
    df_VAA['PROFIT_ACC'] = 0 
    

    for i in range(len(df_VAA)): 
        profit = 0 
        if i != 0: 
            
            #지난달과 동일 자산 보유 
            if df_VAA.iloc[i]['ASSET'] == df_VAA.iloc[i-1]['ASSET']: 
                profit = (df_VAA.iloc[i]['PRICE'] - df_VAA.iloc[i-1]['PRICE']) / df_VAA.iloc[i-1]['PRICE'] * 100 
                
            #지난달과 동일 자산 보유 
            else : 
                profit = (df_VAA.iloc[i][df_VAA.iloc[i-1]['ASSET']] - df_VAA.iloc[i-1]['PRICE']) / df_VAA.iloc[i-1]['PRICE'] * 100 
                
            df_VAA.loc[df_VAA.index[i], 'PROFIT'] = profit 
            df_VAA.loc[df_VAA.index[i], 'PROFIT_ACC'] = ((1+df_VAA.loc[df_VAA.index[i-1], 'PROFIT_ACC']/100)*(1+profit/100)-1)*100


    df_VAA.tail(10)
    plt.figure(figsize=(15,8)) 
    sns.lineplot(data=df_VAA, x=df_VAA.index, y=df_VAA['PROFIT_ACC'])
    plt.show()

    # print(df_VAA) # print to Console
    df_VAA.to_csv('./outPut.csv', sep=',', na_rep="NaN") # printf to File


if __name__ == "__main__":
    main()


About Taehyung Kim

안녕하세요? 8년차 현업 C++ 개발자 김태형이라고 합니다. 😁 C/C++을 사랑하며 다양한 사람과의 협업을 즐깁니다. ☕ 꾸준한 자기개발을 미덕이라 생각하며 노력중이며, 제가 얻은 지식을 홈페이지에 정리 중입니다. 좀 더 상세한 제 이력서 혹은 Private 프로젝트 접근 권한을 원하신다면 메일주세요. 😎

Star
Categories
  • Git
  • Python
  • Django
  • Flutter
  • Cpp
  • Quize
  • OpenSource
  • DirectX
  • Design-pattern
  • win32
  • Unreal
  • DeugU
  • WindowsProgramming
  • Qt
  • 헷깔헷깔시리즈
  • iocp
  • Database
  • CSharp
  • Unity
  • Charp-Server
  • Go
  • Unreal-entry
    • Useful Links
  • Home
  • About
  • Blog