L-M戦略検証

README.md

L-M戦略検証

Yahoo Finance のヒストリカルデータを SP500_daily.csv という名前でダウンロードしてお使いください

$ ls 
lm.py SP500_daily.csv
$ python lm.py

lm.py

import queue
import pandas as pd
import matplotlib.pyplot as plt


def yahoo_finance_open_close(filename):
    df = pd.read_csv(filename)
    series = []
    for _, item in df.iterrows():
        series.append({'open': item['Open'], 'close': item['Close']})
    return series


class Tsumitate:
    def __init__(self, per_day):
        self.per_day = per_day

    def run(self, chart, ignore_fee=False):
        fee_rate = 0.00495  # SBI証券の手数料率 0.495% (消費税込)
        if ignore_fee:
            fee_rate = 0.0
        fee_max = 22  # SBI証券の上限手数料 $22 (消費税込)

        value_series = []
        n_stocks = 0.0
        cash = 0.0
        for day in chart:
            cash += self.per_day
            # 買付時手数料を考慮
            exec_amount = cash - min(cash * fee_rate, fee_max)
            fee = min(exec_amount * fee_rate, fee_max)

            n_stocks += exec_amount / day['open']
            cash -= exec_amount + fee

            value_series.append(n_stocks * day['close'] + cash)
        return value_series


class LM:
    def __init__(self, per_day, ma_len):
        self.per_day = per_day  # 一日に積み立てる額

        self.n_stocks = 0.0  # 口数 (端数計算面倒なので float にしてしまう)
        self.avg_cost = 0.0  # 平均取得金額
        self.cash = 0  # 現金

        self.ma = queue.Queue(maxsize=ma_len)  # ma_len は移動平均線の長さ

    # レバレッジ基準 (以下ベース) となるチャートと、レバレッジをかけたチャートを与えて、
    # 資産額の推移を返す
    def run(self, base_chart, leveraged_chart):
        value_series = []
        order = 'buy'  # buy | sell. 最初は buy (積み立て含む)
        for base, leveraged in zip(base_chart, leveraged_chart):
            # 最初に現金に積み立て額を追加
            self.cash += self.per_day
            # レバ始値で売買
            self.exec_order(leveraged['open'], order)
            # 次の売買はベース終値で判断
            order = self.next_order(base['close'])
            # 現在の資産額はレバ終値で判断
            value_series.append(self.value(leveraged['close']))

        return value_series

    # 次回の売買をどうするかを返す
    def next_order(self, price):
        if self.ma.full():
            self.ma.get()
            self.ma.put(price)
            # 移動平均線を上回れば買い、下回れば売り
            if price >= sum(list(self.ma.queue)) / self.ma.maxsize:
                return 'buy'
            else:
                return 'sell'
        else:
            # とりあえず日数がたまるまでは積み立て
            self.ma.put(price)
            return 'buy'

    # 注文を実行
    # order='buy'  : 全力買い
    # order='sell' : 全力売り
    def exec_order(self, price, order):
        tax_rate = 0.20315  # ドル計算だけどとりあえず日本の譲渡益税率 20.315%
        fee_rate = 0.00495  # SBI証券の手数料率 0.495% (消費税込)
        fee_max = 22  # SBI証券の上限手数料 $22 (消費税込)
        if order == 'buy':
            # 手数料を考慮して買う (絶対現金がマイナスにならない程度の雑な計算)
            exec_amount = self.cash - min(self.cash * fee_rate, fee_max)
            # 端数処理面倒なので口数は float。誤差は気にしない
            quantity = exec_amount / price
            # 必ず self.cash > exec_amount になるので現金は足りる
            fee = min(exec_amount * fee_rate, fee_max)
            # 平均取得金額の更新
            self.avg_cost = (self.avg_cost * self.n_stocks +
                             price * quantity) / (self.n_stocks + quantity)
            self.n_stocks += quantity
            self.cash -= exec_amount + fee
        elif order == 'sell':
            # 売却益の計算
            exec_amount = self.n_stocks * price  # 約定金額
            tax = max((exec_amount - self.n_stocks * self.avg_cost)
                      * tax_rate, 0)  # 利益が出ていなかったら税金は0
            fee = min(exec_amount * fee_rate, fee_max)
            self.cash += exec_amount - tax - fee
            self.n_stocks = 0  # 全売却
            self.avg_cost = 0  # 全売却
        else:
            return

    def value(self, unit):
        return self.n_stocks * unit + self.cash


def to_leveraged(base_chart, x=3):
    chart = []
    prev = base_chart[0]  # 数を合わせるため、レバレッジ版の初日は1倍にしてしまう
    for base in base_chart:
        # 前日の終値からの騰落率の x 倍 (レバレッジの計算方法これであってる？)
        open = base['open'] * (1 + (base['open'] / prev['close'] - 1) * x)
        close = base['close'] * (1 + (base['close'] / prev['close'] - 1) * x)
        chart.append({'open': open, 'close': close})
    return chart


base_series = yahoo_finance_open_close('./SP500_daily.csv')[-200 * 10:]
leveraged_series = to_leveraged(base_series)
cash_series = [{'open': 1, 'close': 1} for _ in range(len(base_series))]

# 毎日10ドル積み立て
lm50_chart = LM(per_day=10, ma_len=50).run(base_series, leveraged_series)
lm100_chart = LM(per_day=10, ma_len=100).run(base_series, leveraged_series)
lm200_chart = LM(per_day=10, ma_len=200).run(base_series, leveraged_series)
lm400_chart = LM(per_day=10, ma_len=400).run(base_series, leveraged_series)
cash_chart = Tsumitate(per_day=10).run(cash_series, ignore_fee=True)
base_chart = Tsumitate(per_day=10).run(base_series)
lv_chart = Tsumitate(per_day=10).run(leveraged_series)

plt.plot(lm50_chart, label='L-M (50)')
plt.plot(lm100_chart, label='L-M (100)')
plt.plot(lm200_chart, label='L-M (200)')
plt.plot(lm400_chart, label='L-M (400)')
plt.plot(cash_chart, label='Cash')
plt.plot(base_chart, label='Buy and Hold (base)')
plt.plot(lv_chart, label='Buy and Hold (leveraged)')
plt.legend()
plt.show()