动量策略
%matplotlib inline
import matplotlib.pyplot as plt
plt.style.use('seaborn')
import matplotlib as mpl
mpl.rcParams['font.family'] = 'serif'
import warnings; warnings.simplefilter('ignore') #忽略警告信息;
import numpy as np
import pandas as pd
import tushare as ts
1. 数据准备 & 回测准备
data = ts.get_k_data('hs300', start = '2010-01-01', end='2017-06-30')[['date','close']]
data.rename(columns={'close': 'price'}, inplace=True)
data.set_index('date', inplace = True)
2. 策略开发思路
data['returns'] = np.log(data['price'] / data['price'].shift(1))
data['position'] = np.sign(data['returns'])
#关键语句,np.sign()很多地方用到;向量化;
data['strategy'] = data['position'].shift(1) * data['returns']
#计算Momentum策略收益;避免未来函数;
3. 策略可视化
data[['returns', 'strategy']].cumsum().apply(np.exp).plot(figsize=(10, 6))
#计算出策略的最终的累计收益;
策略的问题和思考? 过于频繁的买卖开仓;
4. 策略优化之思路——参数优化和穷举
data['position_5'] = np.sign(data['returns'].rolling(5).mean())
data['strategy_5'] = data['position_5'].shift(1) * data['returns']
data[['returns', 'strategy_5']].dropna().cumsum().apply(np.exp).plot(figsize=(10, 6))
参数寻优——使用离散Return计算方法
data['returns_dis'] = data['price'] / data['price'].shift(1)-1
#data['returns_dis'] = data['price'].pct_change()
data['returns_dis_cum'] = (data['returns_dis']+1).cumprod()
price_plot = ['returns_dis_cum']
for days in [10,20,30,60]:
# data['position_%d' % days] = np.sign(data['returns'].rolling(days).mean())
price_plot.append('sty_cumr_%dd' % days)
data['position_%dd' % days] = np.where(data['returns'].rolling(days).mean()>0,1,-1)
data['strategy_%dd' % days] = data['position_%dd' % days].shift(1) * data['returns']
data['sty_cumr_%dd' % days] = (data['strategy_%dd' % days]+1).cumprod()
data[price_plot].dropna().plot(
title='HS300 Multi Parameters Momuntum Strategy',
figsize=(10, 6), style=['--', '--', '--', '--','--'])
5. 策略优化思路之—— High Frequency Data用于Momentum策略
hs300_hf = ts.get_k_data('hs300', ktype='5')
hs300_hf.set_index('date',inplace = True)
hs300_hf.index = hs300_hf.index.to_datetime()
hs300_hf['2017-07-15':'2017-07-28'].head(5)
hs300_hf['returns'] = np.log(hs300_hf['close'] / hs300_hf['close'].shift(1))
hs300_hf['position'] = np.sign(hs300_hf['returns'].rolling(10).mean()) #10个5分钟平均;
hs300_hf['strategy'] = hs300_hf['position'].shift(1) * hs300_hf['returns']
hs300_hf[['returns', 'strategy']].dropna().cumsum().apply(np.exp).plot(figsize=(10, 6),
style=['--', '--'])