def stop(self):
# 防止除以0
if len(self.pnlWins) > 0 and len(self.pnlLosses) > 0:
avgWins = average(self.pnlWins) # 计算平均盈利
avgLosses = abs(average(self.pnlLosses)) # 计算平均亏损（取绝对值）
winLossRatio = avgWins / avgLosses # 盈亏比
if winLossRatio == 0:
kellyPercent = None
else:
numberOfWins = len(self.pnlWins) # 获胜次数
numberOfLosses = len(self.pnlLosses) # 亏损次数
numberOfTrades = numberOfWins + numberOfLosses # 总交易次数
winProb = numberOfWins / numberOfTrades # 计算胜率
inverse_winProb = 1 - winProb
# 计算凯利比率
# 即每次交易投入资金占总资金的最优比率
kellyPercent = winProb - (inverse_winProb / winLossRatio)
else:
kellyPercent = None # 信息不足
self.rets.kellyRatio = kellyPercent # 例如 0.215
self.rets.kellyPercent = kellyPercent * 100 # 例如 21.5
def stop(self):
# There must be at least one winning trade and one losing trade to
# Calculate Kelly percent. Else get a division by zero error.
if len(self.pnlWins) > 0 and len(self.pnlLosses) > 0:
# Calculate average wins and losses
avgWins = average(self.pnlWins)
avgLosses = abs(average(self.pnlLosses)) # Remove the -ve sign
winLossRatio = avgWins / avgLosses
# Check winLoss ratio not 0 else division by zero later because
# otherwise a rare bug can occur if all winners have value of 0.
# (Since BT convention is to class trades with profit >=0 as a
# winner)
if winLossRatio == 0:
kellyPercent = None # Because average of winners were 0.
else:
# Calculate probability of winning from our data.
# Number of wins divide by number of trades.
numberOfWins = len(self.pnlWins)
numberOfLosses = len(self.pnlLosses)
numberOfTrades = numberOfWins + numberOfLosses
winProb = numberOfWins / numberOfTrades
inverse_winProb = 1 - winProb
# Now calculate Kelly percentage
# i.e. optimal percent of account to risk on each trade.
kellyPercent = winProb - (inverse_winProb / winLossRatio)
else:
kellyPercent = None # Not enough information to calculate.
self.rets.kellyRatio = kellyPercent # e.g. 0.215
self.rets.kellyPercent = kellyPercent * 100 # e.g. 21.5
def stop(self):
super(SharpeRatio, self).stop()
if self.p.legacyannual:
retfree = [self.p.riskfreerate] * len(self.anret.rets)
retavg = average(list(map(operator.sub, self.anret.rets, retfree)))
retdev = standarddev(self.anret.rets)
self.ratio = retavg / retdev
else:
rate = self.p.riskfreerate
factor = None
if self.p.timeframe in self.RATEFACTORS:
# rate provided on an annual basis ... downgrade it
factor = self.RATEFACTORS[self.p.timeframe]
elif self.p.timeframe == TimeFrame.Days:
factor = self.p.daysfactor
if factor is not None:
rate = math.pow(1.0 + rate, 1.0 / factor) - 1.0
returns = list(itervalues(self.timereturn.get_analysis()))
retfree = itertools.repeat(rate)
ret_free = map(operator.sub, returns, retfree)
ret_free_avg = average(list(ret_free))
retdev = standarddev(returns)
if retdev != 0:
self.ratio = ret_free_avg / retdev if self.p.convertrate else (factor ** .5) * ret_free_avg / retdev
else:
self.ratio = None
self.rets['sharperatio'] = self.ratio
def stop(self):
if self.p.legacyannual:
retfree = [self.p.riskfreerate] * len(self.anret.rets)
retavg = average(list(map(operator.sub, self.anret.rets, retfree)))
retdev = standarddev(self.anret.rets)
self.ratio = retavg / retdev
else:
rate = self.p.riskfreerate
if self.p.convertrate:
factor = None
if self.p.timeframe in self.RATEFACTORS:
# rate provided on an annual basis ... downgrade it
factor = self.RATEFACTORS[self.p.timeframe]
elif self.p.timeframe == TimeFrame.Days:
factor = self.p.daysfactor
if factor is not None:
rate = math.pow(1.0 + rate, 1.0 / factor) - 1.0
returns = list(itervalues(self.timereturn.get_analysis()))
retfree = itertools.repeat(rate)
ret_free = map(operator.sub, returns, retfree)
ret_free_avg = average(list(ret_free))
retdev = standarddev(returns)
self.ratio = ret_free_avg / retdev
def stop(self):
super(SharpeRatio, self).stop()
if self.p.legacyannual:
rate = self.p.riskfreerate
retavg = average([r - rate for r in self.anret.rets])
retdev = standarddev(self.anret.rets)
self.ratio = retavg / retdev
else:
# Get the returns from the subanalyzer
returns = list(itervalues(self.timereturn.get_analysis()))
rate = self.p.riskfreerate #
factor = None
# Hack to identify old code
if self.p.timeframe == TimeFrame.Days and \
self.p.daysfactor is not None:
factor = self.p.daysfactor
else:
if self.p.factor is not None:
factor = self.p.factor # user specified factor
elif self.p.timeframe in self.RATEFACTORS:
# Get the conversion factor from the default table
factor = self.RATEFACTORS[self.p.timeframe]
if factor is not None:
# A factor was found
if self.p.convertrate:
# Standard: downgrade annual returns to timeframe factor
rate = pow(1.0 + rate, 1.0 / factor) - 1.0
else:
# Else upgrade returns to yearly returns
returns = [pow(1.0 + x, factor) - 1.0 for x in returns]
# Get the excess returns - arithmetic mean - original sharpe
ret_free = [r - rate for r in returns]
ret_free_avg = average(ret_free)
retdev = standarddev(ret_free,
avgx=ret_free_avg,
bessel=self.p.stddev_sample)
try:
ratio = ret_free_avg / retdev
if factor is not None and \
self.p.convertrate and self.p.annualize:
ratio = math.sqrt(factor) * ratio
except (ValueError, TypeError, ZeroDivisionError):
ratio = None
self.ratio = ratio
self.rets['sharperatio'] = self.ratio
def stop(self):
super(SharpeRatio, self).stop()
if self.p.legacyannual:
rate = self.p.riskfreerate
retavg = average([r - rate for r in self.anret.rets])
retdev = standarddev(self.anret.rets)
self.ratio = retavg / retdev
else:
# Get the returns from the subanalyzer
returns = list(itervalues(self.timereturn.get_analysis()))
rate = self.p.riskfreerate #
factor = None
# Hack to identify old code
if self.p.timeframe == TimeFrame.Days and \
self.p.daysfactor is not None:
factor = self.p.daysfactor
else:
if self.p.factor is not None:
factor = self.p.factor # user specified factor
elif self.p.timeframe in self.RATEFACTORS:
# Get the conversion factor from the default table
factor = self.RATEFACTORS[self.p.timeframe]
if factor is not None:
# A factor was found
if self.p.convertrate:
# Standard: downgrade annual returns to timeframe factor
rate = pow(1.0 + rate, 1.0 / factor) - 1.0
else:
# Else upgrade returns to yearly returns
returns = [pow(1.0 + x, factor) - 1.0 for x in returns]
# Get the excess returns - arithmetic mean - original sharpe
ret_free = [r - rate for r in returns]
ret_free_avg = average(ret_free)
retdev = standarddev(ret_free, avgx=ret_free_avg,
bessel=self.p.stddev_sample)
try:
ratio = ret_free_avg / retdev
if factor is not None and \
self.p.convertrate and self.p.annualize:
ratio = math.sqrt(factor) * ratio
except (ValueError, TypeError):
ratio = None
self.ratio = ratio
self.rets['sharperatio'] = self.ratio
def stop(self):
pnl_av = average(self.pnl)
pnl_stddev = standarddev(self.pnl)
trades_sqr = pow(len(self.pnl), 0.5)
sqn = trades_sqr * pnl_av / pnl_stddev
self.ret.sqn = sqn
self.ret.trades = self.count
def stop(self):
if self.count > 1:
pnl_av = average(self.pnl)
pnl_stddev = standarddev(self.pnl)
sqn = math.sqrt(len(self.pnl)) * pnl_av / pnl_stddev
else:
sqn = 0
self.rets.sqn = sqn
self.rets.trades = self.count
def stop(self):
pnl_av = average(self.pnl)
pnl_stddev = standarddev(self.pnl)
trades_sqr = math.sqrt(len(self.pnl))
sqn = trades_sqr * pnl_av / pnl_stddev
self.rets.sqn = sqn
self.rets.trades = self.count
def stop(self):
if self.count > 1:
pnl_av = average(self.pnl)
pnl_stddev = standarddev(self.pnl)
sqn = math.sqrt(len(self.pnl)) * pnl_av / pnl_stddev
else:
sqn = 0
self.rets.sqn = sqn
self.rets.trades = self.count
def stop(self):
pnl_av = average(self.pnl)
pnl_stddev = standarddev(self.pnl)
trades_sqr = pow(len(self.pnl), 0.5)
sqn = trades_sqr * pnl_av / pnl_stddev
self.ret.sqn = sqn
self.ret.trades = self.count
def stop(self):
if self.count > 1:
pnl_av = average(self.pnl)
pnl_stddev = standarddev(self.pnl)
try:
sqn = math.sqrt(len(self.pnl)) * pnl_av / pnl_stddev
except ZeroDivisionError:
sqn = None
else:
sqn = 0
self.rets.sqn = sqn
self.rets.trades = self.count
def stop(self):
if self.count > 1:
pnl_av = average(self.pnl)
pnl_stddev = standarddev(self.pnl)
try:
sqn = math.sqrt(len(self.pnl)) * pnl_av / pnl_stddev
except ZeroDivisionError:
sqn = None
else:
sqn = 0
self.rets.sqn = sqn
self.rets.trades = self.count
def stop(self):
trets = self._tr.get_analysis() # dict key = date, value = ret
pos = nul = neg = 0
trets = list(itervalues(trets))
for tret in trets:
if tret > 0.0:
pos += 1
elif tret < 0.0:
neg += 1
else:
if self.p.zeroispos:
pos += tret == 0.0
else:
nul += tret == 0.0
self.rets['average'] = avg = average(trets)
self.rets['stddev'] = standarddev(trets, avg)
self.rets['positive'] = pos
self.rets['negative'] = neg
self.rets['nochange'] = nul
self.rets['best'] = max(trets)
self.rets['worst'] = min(trets)
def stop(self):
trets = self._tr.get_analysis() # dict key = date, value = ret
pos = nul = neg = 0
trets = list(itervalues(trets))
for tret in trets:
if tret > 0.0:
pos += 1
elif tret < 0.0:
neg += 1
else:
if self.p.zeroispos:
pos += tret == 0.0
else:
nul += tret == 0.0
self.rets['average'] = avg = average(trets)
self.rets['stddev'] = standarddev(trets, avg)
self.rets['positive'] = pos
self.rets['negative'] = neg
self.rets['nochange'] = nul
self.rets['best'] = max(trets)
self.rets['worst'] = min(trets)
def stop(self):
retfree = [self.p.riskfreerate] * len(self.anret.rets)
retavg = average(list(map(operator.sub, self.anret.rets, retfree)))
retdev = standarddev(self.anret.rets)
self.ratio = retavg / retdev
def Stock_Hsi_Diff_Sqn(equity, startdate, enddate=end, period=p):
index="^HSI"
df1=pd.DataFrame() # df1 is equity
df2=pd.DataFrame() # df2 is index
df3=pd.DataFrame() # df3 is the joined dataframe
df_result=pd.DataFrame(columns=['Date', 'sqn', 'Close'])
# print("\r\n\r\n ==== Loading stock", equity, " ====\r\n")
# df1=GetYahooData(equity, startdate, ed=date(2018,3,15) )
df1=GetYahooData(equity, startdate, enddate )
# print(df1)
# print("\r\n\r\n ==== Loading index", index, " ====\r\n")
# df2=GetYahooData(index, startdate, ed=date(2018,3,15))
df2=GetYahooData(index, startdate, enddate)
df1['pchg1']=(df1['Close']-df1['Close'].shift(1)) / df1['Close'] * 100
df2['pchg2']=(df2['Close']-df2['Close'].shift(1)) / df2['Close'] * 100
df1.columns=['Date', 'Open1','High1','Low1','Close1','Adj Close1','Volume1','pchg1']
df2.columns=['Date', 'Open2','High2','Low2','Close2','Adj Close2','Volume2','pchg2']
df1=df1.set_index('Date')
df2=df2.set_index('Date')
df3=pd.concat([df1, df2], axis=1, join='inner')
df3['pertchange']=df3['pchg1'] - df3['pchg2']
# df3['pertchange'].plot()
# print(df3)
# count = 0
# print('len(df3):' , len(df3))
for x in range(1, len(df3)-period):
pnl = list()
for i in df3['pertchange'][x:x+period]:
pnl.append(i)
# print("pnl", pnl)
pnl_av = average(pnl)
pnl_stddev = standarddev(pnl)
# print("pnl_av: ", pnl_av, "pnl_stddev", pnl_stddev)
try:
sqn = math.sqrt(len(pnl)) * pnl_av / pnl_stddev
except ZeroDivisionError:
sqn = None
print("ZeroDivisionError")
# print(x, df3.index[x+99],': SQN: %.2f '% sqn, 'Closing: %.2f'% df3['Close1'][x+99])
df_result=df_result.append({'Date':df3.index[x+period], 'sqn': sqn, 'Close':df3['Close1'][x+period] },
ignore_index=True)
df_result=df_result.set_index('Date')
df_result.index = pd.to_datetime(df_result.index)
print("Equity ", equity)
fig = plt.figure()
ax1 = fig.add_subplot(111)
ax1.set_title(equity, fontsize="large")
ax1.plot(df_result['sqn'], 'silver', label = 'sqn (Equity - Index)')
ax1.axhline(linewidth=1, color='k')
ax1.set_ylabel('y1', color='silver')
# for t1 in ax1.get_yticklabels():
# t1.set_color('k')
ax2 = ax1.twinx()
ax2.plot(df_result['Close'], 'r-', label = equity)
ax2.set_ylabel('y2', color='r')
# for tl in ax2.get_yticklabels():
# tl.set_color('r')
ax1.legend(loc='upper center', bbox_to_anchor=(0.35, -0.2))
ax2.legend(loc='upper center', bbox_to_anchor=(0.7, -0.2))
fig.autofmt_xdate()
plt.show()
# df_result.plot()
for x in range(1, len(df3)):
pnl = list()
for i in df3['pertchange'][1:]:
pnl.append(i)
pnl_av = average(pnl)
pnl_stddev = standarddev(pnl)
try:
sqn2 = math.sqrt(len(pnl)) * pnl_av / pnl_stddev
except ZeroDivisionError:
sqn2 = None
print("ZeroDivisionError")
print(equity, "has SQN nmber diff(stock-index) %.2f"% sqn2)
def stop(self):
retfree = [self.p.riskfreerate] * len(self.anret.rets)
retavg = average(list(map(operator.sub, self.anret.rets, retfree)))
retdev = standarddev(self.anret.rets)
self.ratio = retavg / retdev
df1['pchg1'] = (df1['Close'] -
df1['Close'].shift(1)) / df1['Close'] * 100
df1.columns = [
'Date', 'Open1', 'High1', 'Low1', 'Close1', 'Adj Close1',
'Volume1', 'pchg1'
]
df1 = df1.set_index('Date')
df3 = pd.concat([df1, df2], axis=1, join='inner')
df3['pertchange'] = df3['pchg1'] - df3['pchg2']
for x in range(1, len(df3)):
pnl = list()
for i in df3['pertchange'][1:]:
pnl.append(i)
pnl_av = average(pnl)
pnl_stddev = standarddev(pnl)
try:
sqn2 = math.sqrt(len(pnl)) * pnl_av / pnl_stddev
except ZeroDivisionError:
sqn2 = None
print("ZeroDivisionError")
print(equity, "has SQN nmber %.2f" % sqn2)
d[equity] = sqn2
except:
pass
# finnally, we print the sorted result:
for w in sorted(d, key=d.get, reverse=False):
print(w, "%.2f" % d[w])
def sqn(equity, num=default_num, list_table=False, graph=True):
#
# Define equity and start date of equity data
# equity="^HSI"
startdate = date(2017, 1, 1)
df = pd.DataFrame()
df_result = pd.DataFrame(columns=['Date', 'sqn', 'Close'])
df = GetYahooData(equity, startdate) # download from yahoo directly
# df=GetYahooDataFromFile(equity, startdate ) # read local files
if not isinstance(df, pd.DataFrame):
return None
df = df[pd.notnull(df['Close'])]
df = df.reset_index(drop=True)
df['pertchange'] = (df['Close'] - df['Close'].shift(1)) / df['Close'] * 100
last_sqn = 0
# count = 0
if list_table:
print('No of trading days:', len(df))
for x in range(1, len(df) - num - 1):
pnl = list()
for i in df['pertchange'][x:x + num]:
pnl.append(i)
# print("pnl", pnl)
pnl_av = average(pnl)
pnl_stddev = standarddev(pnl)
# print("pnl_av: ", pnl_av, "pnl_stddev", pnl_stddev)
try:
sqn = math.sqrt(len(pnl)) * pnl_av / pnl_stddev
if list_table:
print(x, df['Date'][x + num], ': SQN: %.2f ' % sqn,
'Closing: %.2f' % df['Close'][x + num])
df_result = df_result.append(
{
'Date': df['Date'][x + num],
'sqn': sqn,
'Close': df['Close'][x + num]
},
ignore_index=True)
except ZeroDivisionError:
sqn = None
print("ZeroDivisionError")
if list_table:
print(x, df['Date'][x + num],
': SQN: Nan Closing: %.2f' % df['Close'][x + num])
if x == (len(df) - num -
2): # if this is the last number, we record the sqn number
last_sqn = df_result['sqn'].iloc[-1]
if graph:
fig = plt.figure()
# df_result['Date'] = pd.to_datetime(df_result['Date'])
df_result.set_index('Date', inplace=True)
ax1 = fig.add_subplot(111)
ax1.plot(df_result['sqn'], label="SQN(99)")
ax1.set_ylabel('y1')
ax2 = ax1.twinx()
ax2.plot(df_result['Close'], 'r-', label=equity)
ax2.set_ylabel('y2', color='r')
for tl in ax2.get_yticklabels():
tl.set_color('r')
ax1.legend(loc='upper center', bbox_to_anchor=(0.35, -0.05))
ax2.legend(loc='upper center', bbox_to_anchor=(0.7, -0.05))
ax1.set_title(equity)
plt.show()
return ("%.2f" % last_sqn)
