-
Notifications
You must be signed in to change notification settings - Fork 36
/
Buy on the Dips.py
244 lines (185 loc) · 7.91 KB
/
Buy on the Dips.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Sun Jun 9 17:16:11 2019
@author: downey
"""
import pyfolio
import numpy as np
import pandas as pd
from pylab import mpl, plt
from IPython.display import set_matplotlib_formats
set_matplotlib_formats('png', 'pdf')
pd.set_option('display.max_columns',110)
pd.set_option('display.max_rows',1000)
plt.style.use('seaborn')
mpl.rcParams['font.family'] = 'serif'
raw = pd.read_csv('/FF_daily.csv', index_col = 0, parse_dates = True)
#Fama French Market US Equity Returns 1926 to mid 2019
raw.tail()
raw.head()
raw.columns = ['Market.Close']
raw.info()
symbol = 'Market.Close'
data = (
pd.DataFrame(raw[symbol])
.dropna()
)
'''
Backtest a strategy where you buy on the dips and sell the next day
1. Buy when the market is down vs. previous close
5. Sell on the next day
'''
results = pd.DataFrame()
data = pd.DataFrame(raw[symbol])
data.dropna(inplace = True)
data['Returns'] = data[symbol].pct_change()
data.dropna(inplace = True)
data.head()
#you can section of certain date frames because with this strategy it goes quickly
#to zero.
data_1990son = data['1990-01-01':'2019-01-01']
data_1990son['Position'] = np.where(data_1990son['Market.Close'] < data_1990son['Market.Close'].shift(1), 1, 0)
data_1990son['Strategy'] = data_1990son['Position'].shift(1) * data_1990son['Returns']
data_1990son.dropna(inplace = True)
D = len(data_1990son)
perf = data_1990son[['Returns', 'Strategy']].add(1).prod() ** (252 / D) - 1
#1. Compute the buy and hold returns
print("Buy and hold returns annualized " + str(round(perf[0], 4)))
#2. Compute the strategy returns and compare it with the buy and hold returns
print("strategy returns annualized " + str(round(perf[1], 4)) + \
" resulting in outperformance of " + str(round(perf[1]-perf[0], 4)))
#3. Plot buy and hold returns and strategy returns in a single chart
portfolio_index = (1 + data_1990son[['Strategy','Returns']]).cumprod()
portfolio_index['1990-01-01':'2006-01-01'].plot()
#4. Compute the Sharpe ratio
std = data_1990son[['Returns', 'Strategy']].std() * 252 ** 0.5
Sharpe = perf/std
print("The Strategy Sharpe Ratio is " + str(round(Sharpe[1], 4)))
#5. Compute and plot the drawdown of the strategy
def Max_Drawdown_Chart(x):
# We are going to use a trailing 252 trading day window
Roll_Max = x.expanding().max()
Daily_Drawdown = x/Roll_Max - 1.0
# Next we calculate the minimum (negative) daily drawdown in that window.
# Again, use min_periods=1 if you want to allow the expanding window
Max_Daily_Drawdown = Daily_Drawdown.expanding().min()
# Plot the results
Daily_Drawdown.plot()
Max_Daily_Drawdown.plot()
def Max_Drawdown(x):
Roll_Max = x.expanding().max()
Daily_Drawdown = x/Roll_Max - 1.0
return Daily_Drawdown.min()
Max_Drawdown_Chart(portfolio_index['Strategy'])
print("Max Drawdown is " + str(Max_Drawdown(portfolio_index['Strategy'])))
#7. Compute the following:
#a. Number of positive trades
portfolio_index.columns = ['Strategy Index', 'Buy and Hold Index']
data = data.join(portfolio_index, how = "left")
data['Signal'] = np.where((data['Position'] == -1), 'Sell', \
np.where((data['Position'] == 1), 'Buy', ''))
data['Signal Price'] = np.where((data['Signal'] == 'Sell') | (data['Signal'] == 'Buy'), \
data['Strategy Index'], 0)
data['Status'] = np.where((data['Signal'].shift(1) == 'Sell'), 'Short', \
np.where((data['Signal'].shift(1) == 'Buy'), 'Long', 'Flat'))
data['Entry Price'] = np.nan
#fill in first Entry Price, which is needed for the next for loop
data.iloc[0, 12] = data.iloc[0, 10]
#Entry Price Logic
#If the current status is the same as previous status, carry forward Entry price
#otherwise if Short/Long use current Signal Price
for i in range(1, len(data)):
data.iloc[i, 12] = np.where((data.iloc[i-1,11] == data.iloc[i,11]), data.iloc[i-1,12], \
np.where((data.iloc[i,11] == 'Short') | (data.iloc[i,11] == 'Long'), data.iloc[i,0], 0))
data['P & L'] = np.where(((data['Status'] == 'Flat') & \
(data['Status'].shift(1) == 'Short')), \
(data['Entry Price'].shift(1) - data['Market.Close']), \
np.where(((data['Status'] == 'Flat') & \
(data['Status'].shift(1) == 'Long')), \
(data['Market.Close'] - data['Entry Price'].shift(1)), np.NaN))
#b. Number of negative trades
print('The Number of Negative Trades ' + str(len(data[data['P & L'] < 0])))
# Number of positive trades
positive_trades = len(data[data['P & L'] > 0])
print('The Number of Positive Trades ' + str(positive_trades))
#c. Total number of signals generated
#where the previous signal is different than the current
signals_generated = np.count_nonzero(data['Signal'] != data['Signal'].shift(1))
print('The number of signals generated is ' + \
str(signals_generated))
#d. Total number of signals traded
total_trades = data['P & L'].dropna().shape[0]
signals_traded = total_trades/signals_generated
print('The Percent of Signals Traded is ' + str(np.round(signals_traded,4)) + \
' and the number of signals traded (total trades) ' + str(total_trades))
#e. Average profit/loss per trade
Total_Profit_Loss = sum(data['P & L'].dropna())
Avg_PL_per_trade = Total_Profit_Loss/total_trades
print('The Average profit/loss per trade ' + str(np.round(Avg_PL_per_trade,4)) + \
', but since the index has grown so much over time this statistic is not extremely meaningful')
#f. Hit Ratio
Hit_Ratio = positive_trades / total_trades
print('The Hit Ratio for this strategy is ' + str(np.round(Hit_Ratio,4)))
#g. Highest profit & loss in a single trade
trade_data = data['P & L'].dropna()
Max_Profit = trade_data.max()
Max_Loss = trade_data.min()
print('The Highest Profit on any one trade was ' + str(np.round(Max_Profit,2)) + ' dollars')
print('The Highest Loss on any one trade was ' + str(np.round(Max_Loss,2)) + ' dollars')
pyfolio.create_simple_tear_sheet(data['Strategy'])
pyfolio.create_full_tear_sheet(data['Strategy'])
##############Optional - Optimize the Strategy#################################
import time
t0 = time.time()
#SMA
lookback = range(1, 40, 1)
results = pd.DataFrame()
for LOOKBACK in lookback:
results = pd.DataFrame()
data = pd.DataFrame(raw[symbol])
data.dropna(inplace = True)
data['Returns'] = data[symbol].pct_change()
data.dropna(inplace = True)
data['Position'] = np.where(data['Market.Close'] < data['Market.Close'].shift(LOOKBACK), 1, 0)
data['Strategy'] = data['Position'].shift(1) * data['Returns']
data.dropna(inplace = True)
D = len(data)
perf = data[['Returns', 'Strategy']].add(1).prod() ** (252 / D) - 1
std = data[['Returns', 'Strategy']].std() * 252 ** 0.5
Sharpe = perf/std
results = results.append(pd.DataFrame(
{'Lookback': LOOKBACK,
'MARKET': perf['Returns'],
'STRATEGY': perf['Strategy'],
'OUT': Sharpe['Strategy'] - Sharpe['Returns']},
index = [0]), ignore_index = True)
results.info()
results.sort_values('OUT', ascending = False).head(15)
results
t1 = time.time()
total = t1-t0
print('It took ' + str(np.round(total/60,2)) + ' minutes to run the code')
#####Testing Statistical Significance of Alpha#########
from scipy import stats
data.head()
#Sample Size
N = data.shape[0]
#Calculate the variance to get the standard deviation
#For unbiased max likelihood estimate we have to divide the var by N-1, and therefore the parameter ddof = 1
var_alpha = data['Strategy'].var(ddof=1)
## Calculate the t-statistics
t = (data['Strategy'].mean() - data['Returns'].mean()) / np.sqrt(var_alpha/N)
## Compare with the critical t-value
#Degrees of freedom
df = N-1
#p-value after comparison with the t
p = 1 - stats.t.cdf(t,df=df)
print("t = " + str(t))
print("p = " + str(p))
### You can see that after comparing the
ZEROS = [0] * N
## Cross Checking with the internal scipy function
t2, p2 = stats.ttest_ind(data['Strategy'],data['Returns'])
print("t = " + str(t2))
print("p = " + str(p2/2)) #one sided t test