Quantitative Analysis
Computational Investing
Chris Harris
11 Nov 2014
Chris Harris
11 Nov 2014
Introduction
In this class conducted by Tucker Balch, Georgia Tech, the first initiative involves developing an investment strategy, which has the ability to survive backtesting. For the time period concerned, Jan 2008 to Dec 2009, the market underwent a major downdraft in response to the subprime mortgage crisis. However, as this data presents, a simple strategy of buying stocks within the S&P 500 index, which dipped below $ 5 per share, a critical margin level requirement, for a holding period of just 5 trading days, could have kept an investment portfolio in the black throughout the ‘Great Recession’, without a significant drawdown. If a stock repeats this price pattern, it can be purchased multiple times. For simplicity, no portfolio scaling takes place, as only 100 shares are purchased for each event, regardless of investment capital size. The main goal of this study is to compare one strategy’s success relative to another, so portfolio size should not factor into the equation. Through this exercise and subsequent examples presented throughout the class, I became indoctrinated into the hedge fund manager’s circle. Program development procedes according to the following scheme:
-
event.py identifies times when a stock drops below $ 5 / share within the S&P 500 index
-
marketsim.py generates buy and sell orders to align with a 5 trading day holding period
-
analyze.py determines profits and losses based on buy / sell orders and porfolio capital funding
- event.py identifies times when a stock drops below $ 5 / share within the S&P 500 index
- marketsim.py generates buy and sell orders to align with a 5 trading day holding period
- analyze.py determines profits and losses based on buy / sell orders and porfolio capital funding
Python Code
event.py:
import os
import shutil
import math
import matplotlib as mpl
import matplotlib.pyplot as plt
# obtain market data
def acquire(sym,interval):
if not os.path.isfile("../data/%s.csv" %sym):
shutil.copy("../QSTK-0.2.8/QSTK/QSData/Yahoo/%s.csv" %sym, "../data")
data = open("../data/%s.csv" %sym,'r')
stream = data.read()
data.close()
field = 7
entry = ""
val = []
matrix = []
for ch in stream:
entry = entry + ch
if ch==',' or ch=='\n':
if len(entry)>1:
val.append(entry[:-1])
entry = ""
if len(val)>=field:
if interval[1] >= val[0] >= interval[0]:
matrix.append(val)
val = []
matrix.sort()
return matrix
def plot(label,mean,std,delta):
X = []
Y = []
xl = [-delta-1,delta+1]
mpl.rc('text', color='#C8A078')
mpl.rc('figure', facecolor='black', edgecolor='black')
mpl.rc('axes', edgecolor='#C6BDBA', labelcolor='#C8A078', facecolor='black', linewidth=2)
mpl.rc('xtick', color='#C8A078')
mpl.rc('ytick', color='#C8A078')
plt.tick_params(bottom=False, top=False, left=False, right=False)
plt.figure(1)
for i in range(0,2*delta+1):
x = []
y = []
x.append(float(i-delta))
x.append(x[0])
y.append(mean[i] - std[i])
y.append(mean[i] + std[i])
if x[0] > 0.0:
plt.plot(x, y, color='#F5A078', linewidth=1)
X.append(x[0])
Y.append(mean[i])
plt.plot(X, Y, color='#004040', linewidth=2)
plt.xlim(xl)
plt.title("%s Mean Return Relative to S&P500 Index: 2008 to 2009" %label, fontsize=25)
plt.xticks(fontsize=20)
plt.yticks(fontsize=20)
plt.xlabel("Trade days", fontsize=25)
plt.ylabel("Cumulative Return", fontsize=25)
plt.figure(2)
plt.plot(X, Y, color='#004040', linewidth=2)
plt.xlim(xl)
plt.title("%s Mean Return Relative to S&P500 Index: 2008 to 2009" %label, fontsize=25)
plt.xticks(fontsize=20)
plt.yticks(fontsize=20)
plt.xlabel("Trade days", fontsize=25)
plt.ylabel("Cumulative Return", fontsize=25)
plt.show()
def process(spfile):
# read symbol file
data = open(spfile,'r')
stream = data.read()
data.close()
# process symbol list
entry = ""
symbol = []
for ch in stream:
entry = entry + ch
if ch==',' or ch=='\n':
if len(entry)>1:
symbol.append(entry[:-1])
entry = ""
return symbol
def event(spfile,interval,ofile):
hold = 5
delta = 20
shr = 100
ref = "SPY"
act = ["Sell","Buy"]
# obtain reference data
rmatrix = acquire(ref,interval)
period = len(rmatrix)
print("\nperiod:\t%s to %s\n\t%i overall trade days\n\nomit:" %(interval[0],interval[1],period),end="")
# check for event
ev = 0
stream = ""
array = []
sdt = [0.0 for p in range(2*delta+1)]
symbol = process(spfile)
for sym in symbol:
matrix = acquire(sym,interval)
if len(matrix) < period:
print("\t%s" %sym)
continue
for i in range(delta,period-delta):
if float(matrix[i-1][6]) >= 5.0 and float(matrix[i][6]) < 5.0:
date = matrix[i][0]
stream = stream + "%s,%s,%s,%s,%s,%s,\n" %(date[0:4],date[5:7],date[8:10],sym,act[1],shr)
date = matrix[i+hold][0]
stream = stream + "%s,%s,%s,%s,%s,%s,\n" %(date[0:4],date[5:7],date[8:10],sym,act[0],shr)
l = 0
array.append([0.0 for q in range(2*delta+1)])
for j in range(i-delta,i+delta+1):
dt = float(matrix[j][6])/float(matrix[i][6])-float(rmatrix[j][6])/float(rmatrix[i][6])
array[ev][l] = dt
sdt[l] = sdt[l] + dt
l = l + 1
ev = ev + 1
# evaluate event data
std = [0.0 for r in range(2*delta+1)]
mean = [0.0 for s in range(2*delta+1)]
for i in range(0,2*delta+1):
sd2 = 0.0
mean[i] = sdt[i]/float(ev)
for j in range(0,ev):
d = array[j][i] - mean[i]
sd2 = sd2 + d**2.0
std[i] = math.sqrt(sd2/float(ev))
# write data file
output = open(ofile,'w')
output.write(stream)
output.close()
# display results
print("\nEvent count = %i\n" %ev)
plot("Event",mean,std,delta)
def main():
ofile = "orders.csv"
spfile = "sp5002012.txt"
interval = ["2008-01-03","2009-12-28"]
event(spfile,interval,ofile)
main()
Python Code
marketsim.py:
import sys
import os
import shutil
# obtain market data
def acquire(sym,interval):
if not os.path.isfile("../data/%s.csv" %sym):
shutil.copy("../QSTK-0.2.8/QSTK/QSData/Yahoo/%s.csv" %sym, "../data")
data = open("../data/%s.csv" %sym,'r')
stream = data.read()
data.close()
field = 7
entry = ""
val = []
matrix = []
for ch in stream:
entry = entry + ch
if ch==',' or ch=='\n':
if len(entry)>1:
val.append(entry[:-1])
entry = ""
if len(val)>=field:
if interval[1] >= val[0] >= interval[0]:
matrix.append(val)
val = []
matrix.sort()
return matrix
def gather(orders):
# read order file
data = open(orders,'r')
stream = data.read()
data.close()
# input order data
field = 6
entry = ""
val = []
fmt = []
odata = []
for ch in stream:
entry = entry + ch
if ch==',' or ch=='\n':
if len(entry)>1:
val.append(entry[:-1])
entry = ""
if len(val)>=field:
for i in range(1,3):
if len(val[i]) < 2:
val[i] = '0' + val[i]
fmt.append("%s-%s-%s" %(val[0],val[1],val[2]))
fmt.append(val[3])
fmt.append(val[4])
fmt.append(val[5])
odata.append(fmt)
val = []
fmt = []
odata.sort()
return odata
def execute(orders,initial,interval,values):
# retrieve market data
symbol = []
matrix = []
odata = gather(orders)
print()
for i in range(0,len(odata)):
print("%s\t%s\t%s\t%s" %(odata[i][0],odata[i][1],odata[i][2],odata[i][3]))
if odata[i][1] in symbol:
continue
symbol.append(odata[i][1])
matrix.append(acquire(odata[i][1],interval))
# process daily value
stream = ""
cash = initial
shr = [0.0 for i in range(len(symbol))]
for i in range(0,len(matrix[0])):
date = matrix[0][i][0]
for j in range(0,len(odata)):
if date==odata[j][0]:
for k,item in enumerate(symbol):
if odata[j][1]==symbol[k]:
delta=float(odata[j][3])
equity = delta*float(matrix[k][i][6])
if odata[j][2]=="Buy":
cash = cash - equity
shr[k] = shr[k] + delta
if odata[j][2]=="Sell":
cash = cash + equity
shr[k] = shr[k] - delta
val = cash
for j,item in enumerate(symbol):
val = val + shr[j]*float(matrix[j][i][6])
stream = stream + "%s,%s,%s,%.8g\n" %(date[0:4],date[5:7],date[8:10],val)
# write data file
output = open("values.csv",'w')
output.write(stream)
output.close()
print("\ninitial = %.8g\norders = %s\nvalues = %s\n" %(initial,orders,values))
def main():
initial = float(sys.argv[1])
orders = sys.argv[2]
values = sys.argv[3]
interval = ["2008-01-03","2009-12-28"]
execute(orders,initial,interval,values)
main()
Python Code
analyze.py:
import sys
import os
import shutil
import math
import matplotlib as mpl
import matplotlib.pyplot as plt
# obtain market data
def acquire(sym,interval):
if not os.path.isfile("../data/%s.csv" %sym):
shutil.copy("../QSTK-0.2.8/QSTK/QSData/Yahoo/%s.csv" %sym, "../data")
data = open("../data/%s.csv" %sym,'r')
stream = data.read()
data.close()
field = 7
entry = ""
val = []
matrix = []
for ch in stream:
entry = entry + ch
if ch==',' or ch=='\n':
if len(entry)>1:
val.append(entry[:-1])
entry = ""
if len(val)>=field:
if interval[1] >= val[0] >= interval[0]:
matrix.append(val)
val = []
matrix.sort()
return matrix
def finance(matrix):
tsh = 252.0
r = [0.0 for m in range(4)]
dr = [0.0 for n in range(len(matrix))]
# total return
r[0] = float(matrix[len(matrix)-1][1])/float(matrix[0][1]) - 1.0
# daily return
# --------------------------------------------------------------------------------------------
#
# leave first element 'dr[0]' = 0
#
# --------------------------------------------------------------------------------------------
for i in range(1,len(matrix)):
dr[i] = float(matrix[i][1])/float(matrix[i-1][1]) - 1.0
# mean daily return
sdr = 0.0
for i, item in enumerate(matrix):
sdr = sdr + dr[i]
r[1] = sdr/float(len(matrix))
# standard deviation
sd2 = 0.0
for i, item in enumerate(matrix):
d = r[1] - dr[i]
sd2 = sd2 + d**2.0
r[2] = math.sqrt(sd2/float(len(matrix)))
# sharpe ratio
r[3] = math.sqrt(tsh)*r[1]/r[2]
return r,dr
def linearregression(matrix):
s0 = 0.
sx = 0.
sx2 = 0.
sy = 0.
sxy = 0.
X = [0.0 for l in range(len(matrix))]
Y = [0.0 for m in range(len(matrix))]
p = [0.0 for n in range(3)]
for i, item in enumerate(matrix):
X[i] = float(i)
Y[i] = float(matrix[i][1])
s0 = s0 + 1.
sx = sx + X[i]
sx2 = sx2 + X[i]*X[i]
sy = sy + Y[i]
sxy = sxy + X[i]*Y[i]
meanx = sx/s0
meany = sy/s0
p[0] = (s0*sxy - sx*sy)/(s0*sx2 - sx*sx)
p[1] = meany - p[0]*meanx
ssy = 0.
ssr = 0.
for i, item in enumerate(matrix):
ssy = ssy + (Y[i] - meany)*(Y[i] - meany)
ssr = ssr + (Y[i] - p[1] - p[0]*X[i])*(Y[i] - p[1] - p[0]*X[i])
p[2] = math.sqrt(1.0 - ssr/ssy)
return p
def plot(label,symbol,matrix,array,p):
X = []
Y = []
xl = [-10,len(array)+10]
mpl.rc('text', color='#C8A078')
mpl.rc('figure', facecolor='black', edgecolor='black')
mpl.rc('axes', edgecolor='#C6BDBA', labelcolor='#C8A078', facecolor='black', linewidth=2)
mpl.rc('xtick', color='#C8A078')
mpl.rc('ytick', color='#C8A078')
plt.tick_params(bottom=False, top=False, left=False, right=False)
plt.figure(1)
for i in range(0,len(array)):
X.append(float(i))
Y.append(p[0]*X[i] + p[1])
plt.plot(X, Y, color='#F5A078', linewidth=2)
Y = []
for i in range(0,len(array)):
Y.append(float(array[i][1]))
plt.plot(X, Y, color='#004040', linewidth=2, label="Port")
color = ['#640000','#004000','#004040','#640064']
Y = []
for i in range(0,len(array)):
Y.append(float(matrix[i][1])/float(matrix[0][1])*float(array[0][1]))
plt.plot(X, Y, color='#640064', linewidth=2, label=symbol)
plt.xlim(xl)
plt.legend(loc=(0.8,0.56), frameon=False, fontsize=20)
fmt = "equity = %.4gt + %.4g\n corr = %.4g\n" %(p[0],p[1],p[2])
plt.annotate(fmt, xy=(25,900000), fontsize=20)
plt.title("%s Performance: %s to %s" %(label,array[0][0],array[len(array)-1][0]), fontsize=25)
plt.xticks(fontsize=20)
plt.yticks(fontsize=20)
plt.xlabel("Trade day", fontsize=25)
plt.ylabel("Value, dollar", fontsize=25)
plt.show()
def valdata(values):
# read value file
data = open(values,'r')
stream = data.read()
data.close()
# input value data
field = 4
entry = ""
val = []
fmt = []
vdata = []
for ch in stream:
entry = entry + ch
if ch==',' or ch=='\n':
if len(entry)>1:
val.append(entry[:-1])
entry = ""
if len(val)>=field:
for i in range(1,3):
if len(val[i]) < 2:
val[i] = '0' + val[i]
fmt.append("%s-%s-%s" %(val[0],val[1],val[2]))
fmt.append(val[3])
vdata.append(fmt)
val = []
fmt = []
return vdata
def refdata(vdata,symbol):
# retrieve market data
interval = [vdata[0][0],vdata[len(vdata)-1][0]]
matrix = acquire(symbol,interval)
period = len(matrix)
print()
val = []
rdata = []
for i in range(0,period):
val.append(matrix[i][0])
val.append(matrix[i][6])
rdata.append(val)
val = []
if i==0 or i==period-1:
print("%s\t%s" %(vdata[i][0],vdata[i][1]))
return rdata
def analyze(values,symbol):
vdata = valdata(values)
rdata = refdata(vdata,symbol)
# output results
r,dr = finance(vdata)
p = linearregression(vdata)
f = ("\nPortfolio:\n\ntotal return = %.4g\nmean daily return = %.4g\nstandard deviation = %.4g\nSharpe ratio = %.4g"
%(r[0],r[1],r[2],r[3]))
print(f)
r,dr = finance(rdata)
f = ("\n$SPX:\n\ntotal return = %.4g\nmean daily return = %.4g\nstandard deviation = %.4g\nSharpe ratio = %.4g\n"
%(r[0],r[1],r[2],r[3]))
print(f)
plot("Portfolio",symbol,rdata,vdata,p)
def main():
values = sys.argv[1]
symbol = sys.argv[2]
analyze(values,symbol)
main()
Results
$ python event.py
period: 2008-01-03 to 2009-12-28
501 overall trade days
omit: BWA
CFN
DPS
KMI
LO
LYB
MJN
MPC
PEG
PM
PSX
QEP
SNI
TRIP
V
WPX
XYL
Event count = 193
$ python marketsim.py 10000 orders.csv values.csv
2008-02-06 LSI Buy 100
2008-02-13 LSI Sell 100
2008-02-19 LSI Buy 100
2008-02-26 LSI Sell 100
2008-02-28 THC Buy 100
2008-03-06 THC Buy 100
2008-03-06 THC Sell 100
2008-03-07 LSI Buy 100
2008-03-13 THC Sell 100
2008-03-14 LSI Buy 100
...
2009-10-28 LSI Buy 100
2009-10-30 RF Buy 100
2009-11-03 LSI Buy 100
2009-11-03 RF Sell 100
2009-11-04 AMD Sell 100
2009-11-04 LSI Sell 100
2009-11-06 RF Sell 100
2009-11-10 LSI Sell 100
2009-11-27 THC Buy 100
2009-12-04 THC Sell 100
initial = 10000
orders = orders.csv
values = values.csv
$ python analyze.py values.csv \$SPX
2008-01-03 10000
2009-12-28 13590
Portfolio:
total return = 0.359
mean daily return = 0.0006709
standard deviation = 0.01091
Sharpe ratio = 0.9762
$SPX:
total return = -0.2207
mean daily return = -0.0002553
standard deviation = 0.022
Sharpe ratio = -0.1842
Figure 1: Event Mean Return with Standard Deviation Range Bars
Figure 1: Event Mean Return with Standard Deviation Range Bars
Figure 2: Event Mean Return
Figure 2: Event Mean Return
Figure 3: Portfolio Performance
Figure 3: Portfolio Performance
