【一括投資 vs ドルコスト平均法】pythonで過去成績を比較してみた

import numpy as np
import matplotlib.pyplot as plt
import pandas_datareader.data as web
import datetime
%matplotlib inline
import warnings
warnings.filterwarnings('ignore')

# 期間設定
start = datetime.datetime(1980,3,31)
end = datetime.datetime(2021,12,25)

# データ取得
df = web.DataReader(['VTI'],'yahoo',start,end)['Adj Close']

# 系列名変更
df.columns = ['VTI']

%matplotlib inline

fig, ax = plt.subplots(1, figsize=(12, 8))
df.plot(ax =ax, label='VTI')
plt.grid()
ax.set_ylabel(u'price')
ax.set_xlabel(u'date')

def package(year):
    price = 10000    # 初期保有資産額
    df['hold'] = price // df.iloc[:,0] # 購入株式数
    df['rest'] = price % df.iloc[:,0]  # 残り金額
    df['return'] = np.zeros(len(df))        # リターン
    for i in range(len(df)-251*year):
        df.iloc[i,3] = df.iloc[251*year + i,0] *df.iloc[i,1] + df.iloc[i,2]  # 251営業日で約一年後×10年
    df['return_p'] = df['return'] / price
    return df


def dollercost(year):
    df['hold'] = np.zeros(len(df)) 
    df['rest'] = np.zeros(len(df)) 
    df['return'] = np.zeros(len(df)) 
    price = 10000    # 初期保有資産額
    price_m = price/12/year  # 毎月投資額
    
    for i in range(len(df)-251*year):
        for j in np.arange(0, 12*year)*20:   # 1月（20営業日）ごとに購入
            hold_m = price_m / df.iloc[i+j,0]  # 購入株式数( あまりも含む)
            df.iloc[i,1] = df.iloc[i,1] + hold_m           # 保有株数の計算
        df.iloc[i,3] = df.iloc[251*year + i,0] * df.iloc[i,1]
    df['return_p'] = df['return'] / price  # 251営業日で約一年後×10年
    return df


year = 1    # 運用期間
df_p = package(year)
df_p= df_p.replace(0, np.NaN)

df_d = dollercost(year)
df_d= df_d.replace(0, np.NaN)

# グラフの描写
plt.hist(df_p['return_p'], bins=30, alpha = 0.5, label='Package', density=True)
plt.hist(df_d['return_p'], bins=30, alpha = 0.5, label='Dollar cost', density=True)
plt.grid()
plt.xlabel('Return')
plt.ylabel('Probability')
plt.legend(loc='upper right')


# 一括投資がドルコスト平均法を上回った確率の計算
ret_d =df_p['return_p'] - df_d['return_p']
count = 0
for i in range(len(ret_d)):
    if ret_d[i] >= 0:
        count=count + 1
    
p_pack = count/(len(df)-251*year)
print(p_pack)   # 一括投資がドルコスト平均法を上回った確率

# 一括購入
ret_p_mean = df_p['return_p'].mean()
ret_p_std = df_p['return_p'].std()
# ドルコスト平均法
ret_d_mean = df_d['return_p'].mean()
ret_d_std = df_d['return_p'].std()