def RunTradingModelQLearn(ticker: str,
startDate: str,
durationInYears: int,
totalFunds: int,
verbose: bool = False,
saveHistoryToFile: bool = True,
returndailyValues: bool = False):
Actions = [
'Hold', 'BuyMarket', 'BuyAgressiveLeve10', 'BuyAgressiveLeve11',
'BuyAgressiveLeve12', 'SellMarket', 'SellAgressiveLeve10',
'SellAgressiveLeve11', 'SellAgressiveLeve12'
]
modelName = 'QLearn' + '_' + ticker
tm = TradingModel(modelName, ticker, startDate, durationInYears,
totalFunds, verbose)
if not tm.modelReady:
print('Unable to initialize price history for model for ' +
str(startDate))
if returndailyValues: return pandas.DataFrame()
else: return totalFunds
else:
while not tm.ModelCompleted():
tm.ProcessDay()
currentPrices = tm.GetPriceSnapshot()
if not currentPrices == None:
pass
#do stuff
if returndailyValues:
tm.CloseModel(verbose, saveHistoryToFile)
return tm.GetdailyValue() #return daily value
else:
return tm.CloseModel(verbose,
saveHistoryToFile) #return closing value
def RunTradingModelBuyHold(ticker: str,
startDate: str,
durationInYears: int,
totalFunds: int,
verbose: bool = False,
saveHistoryToFile: bool = True,
returndailyValues: bool = False):
modelName = 'BuyHold' + '_' + ticker
tm = TradingModel(modelName, ticker, startDate, durationInYears,
totalFunds, verbose)
if not tm.modelReady:
print('Unable to initialize price history for model for ' +
str(startDate))
if returndailyValues: return pandas.DataFrame()
else: return totalFunds
else:
while not tm.ModelCompleted():
tm.ProcessDay()
currentPrices = tm.GetPriceSnapshot()
if not currentPrices == None:
if tm.TraunchesAvailable(
) and tm.FundsAvailable() > currentPrices.high:
tm.PlaceBuy(ticker, currentPrices.low, True)
if tm.AccountingError(): break
if returndailyValues:
tm.CloseModel(verbose, saveHistoryToFile)
return tm.GetdailyValue() #return daily value
else:
return tm.CloseModel(verbose,
saveHistoryToFile) #return closing value
def RunModel(modelName: str,
modelFunction,
ticker: str,
startDate: str,
durationInYears: int,
totalFunds: int,
saveHistoryToFile: bool = True,
returndailyValues: bool = False,
verbose: bool = False):
modelName = modelName + '_' + ticker
tm = TradingModel(modelName=modelName,
startingTicker=ticker,
startDate=startDate,
durationInYears=durationInYears,
totalFunds=totalFunds,
verbose=verbose)
if not tm.modelReady:
print('Unable to initialize price history for model for ' +
str(startDate))
if returndailyValues: return pandas.DataFrame()
else: return totalFunds
else:
while not tm.ModelCompleted():
tm.ProcessDay()
modelFunction(tm, ticker)
if tm.AccountingError():
print(
'Accounting error. The numbers do not add up correctly.')
break
if returndailyValues:
tm.CloseModel(verbose, saveHistoryToFile)
return tm.GetDailyValue() #return daily value
else:
return tm.CloseModel(verbose,
saveHistoryToFile) #return closing value
def RunModel(modelName: str,
modelFunction,
ticker: str,
startDate: str,
durationInYears: int,
portfolioSize: int,
saveHistoryToFile: bool = True,
returndailyValues: bool = False,
verbose: bool = False):
#Performs the logic of the given model over a period of time to evaluate the performance
modelName = modelName + '_' + ticker
print('Running model ' + modelName)
tm = TradingModel(modelName=modelName,
startingTicker=ticker,
startDate=startDate,
durationInYears=durationInYears,
totalFunds=portfolioSize,
tranchSize=round(portfolioSize / 10),
verbose=verbose)
if not tm.modelReady:
print('Unable to initialize price history for model for ' +
str(startDate))
if returndailyValues: return pd.DataFrame()
else: return portfolioSize
else:
while not tm.ModelCompleted():
modelFunction(tm, ticker)
tm.ProcessDay()
if tm.AccountingError():
print(
'Accounting error. The numbers do not add up correctly. Terminating model run.',
tm.currentDate)
tm.PositionSummary()
#tm.PrintPositions()
break
cash, asset = tm.Value()
#print('Ending Value: ', cash + asset, '(Cash', cash, ', Asset', asset, ')')
tradeCount = len(tm.tradeHistory)
RecordPerformance(ModelName=modelName,
StartDate=startDate,
EndDate=tm.currentDate,
StartValue=portfolioSize,
EndValue=(cash + asset),
TradeCount=tradeCount)
if returndailyValues:
tm.CloseModel(verbose, saveHistoryToFile)
return tm.GetDailyValue(
) #return daily value for model comparisons
else:
return tm.CloseModel(
verbose, saveHistoryToFile
) #return simple closing value to view net effect
def RunPriceMomentum(tickerList:list, startDate:str='1/1/1982', durationInYears:int=36, stockCount:int=9, ReEvaluationInterval:int=20, filterOption:int=3, longHistory:int=365, shortHistory:int=90, minPercentGain=0.05, maxVolatility=.12, portfolioSize:int=30000, returndailyValues:bool=False, verbose:bool=False):
#Choose stockCount stocks with the greatest long term (longHistory days) price appreciation, using different filter options defined in the StockPicker class
#shortHistory is a shorter time frame (like 90 days) used differently by different filters
#ReEvaluationInterval is how often to re-evaluate our choices, ideally this should be very short and not matter, otherwise the date selection is biased.
startDate = ToDate(startDate)
endDate = AddDays(startDate, 365 * durationInYears)
picker = StockPicker(AddDays(startDate, -730), endDate) #Include earlier dates for statistics
for t in tickerList:
picker.AddTicker(t)
tm = TradingModel(modelName='PriceMomentumShort_longHistory_' + str(longHistory) +'_shortHistory_' + str(shortHistory) + '_reeval_' + str(ReEvaluationInterval) + '_stockcount_' + str(stockCount) + '_filter' + str(filterOption) + '_' + str(minPercentGain) + str(maxVolatility), startingTicker='^SPX', startDate=startDate, durationInYears=durationInYears, totalFunds=portfolioSize, tranchSize=portfolioSize/stockCount, verbose=verbose)
dayCounter = 0
if not tm.modelReady:
print('Unable to initialize price history for PriceMomentum date ' + str(startDate))
return 0
else:
while not tm.ModelCompleted():
currentDate = tm.currentDate
if dayCounter ==0:
print('\n')
print(currentDate)
c, a = tm.Value()
print(tm.modelName, int(c), int(a), int(c+a))
print('available/buy/sell/long',tm.PositionSummary())
candidates = picker.GetHighestPriceMomentum(currentDate, longHistoryDays=longHistory, shortHistoryDays=shortHistory, stocksToReturn=stockCount, filterOption=filterOption, minPercentGain=minPercentGain, maxVolatility=maxVolatility)
AlignPositions(tm=tm, targetPositions=candidates, stockCount=stockCount, allocateByPointValue=False)
tm.ProcessDay()
dayCounter+=1
if dayCounter >= ReEvaluationInterval: dayCounter=0
cv1 = tm.CloseModel(plotResults=False, saveHistoryToFile=((durationInYears>1) or verbose))
if returndailyValues:
return tm.GetDailyValue()
else:
return cv1
def RunModel(modelName: str,
modelFunction,
ticker: str,
startDate: str,
durationInYears: int,
portfolioSize: int,
saveHistoryToFile: bool = True,
returndailyValues: bool = False,
verbose: bool = False):
modelName = modelName + '_' + ticker
tm = TradingModel(modelName=modelName,
startingTicker=ticker,
startDate=startDate,
durationInYears=durationInYears,
totalFunds=portfolioSize,
verbose=verbose)
if not tm.modelReady:
print('Unable to initialize price history for model for ' +
str(startDate))
if returndailyValues: return pd.DataFrame()
else: return portfolioSize
else:
while not tm.ModelCompleted():
tm.ProcessDay()
modelFunction(tm, ticker)
if tm.AccountingError():
print(
'Accounting error. The numbers do not add up correctly. Terminating model run.'
)
tm.PositionSummary()
#tm.PrintPositions()
break
if returndailyValues:
tm.CloseModel(verbose, saveHistoryToFile)
return tm.GetDailyValue(
) #return daily value for model comparisons
else:
return tm.CloseModel(
verbose, saveHistoryToFile
) #return simple closing value to view net effect
def RunTradingModelSeasonal(ticker: str,
startDate: str,
durationInYears: int,
totalFunds: int,
verbose: bool = False,
saveHistoryToFile: bool = True,
returndailyValues: bool = False):
modelName = 'Seasonal' + '_' + ticker
tm = TradingModel(modelName, ticker, startDate, durationInYears,
totalFunds, verbose)
if not tm.modelReady:
print('Unable to initialize price history for model for ' +
str(startDate))
if returndailyValues: return pandas.DataFrame()
else: return totalFunds
else:
while not tm.ModelCompleted():
tm.ProcessDay()
currentPrices = tm.GetPriceSnapshot()
if not currentPrices == None:
low = currentPrices.low
high = currentPrices.high
m = tm.currentDate.month
available, buyPending, sellPending, longPositions = tm.GetPositionSummary(
)
if m >= 11 or m <= 4: #Buy if Nov through April, else sell
if available > 0 and tm.FundsAvailable() > high:
tm.PlaceBuy(ticker, low, True)
else:
if longPositions > 0: tm.PlaceSell(ticker, high, True)
if tm.AccountingError(): break
if returndailyValues:
tm.CloseModel(verbose, saveHistoryToFile)
return tm.GetdailyValue() #return daily value
else:
return tm.CloseModel(verbose,
saveHistoryToFile) #return closing value
def RunBuyHold(ticker: str, startDate:str, durationInYears:int, ReEvaluationInterval:int=20, portfolioSize:int=30000, verbose:bool=False):
#Baseline model to compare against. Buy on day one, hold for the duration and then sell
modelName = 'BuyHold_' + (ticker) + '_' + startDate[-4:]
tm = TradingModel(modelName=modelName, startingTicker=ticker, startDate=startDate, durationInYears=durationInYears, totalFunds=portfolioSize, tranchSize=portfolioSize/10, verbose=verbose)
if not tm.modelReady:
print('Unable to initialize price history for model BuyHold date ' + str(startDate))
return 0
else:
dayCounter =0
while not tm.ModelCompleted():
if dayCounter ==0:
i=0
while tm.TranchesAvailable() and i < 100:
tm.PlaceBuy(ticker=ticker, price=1, marketOrder=True, expireAfterDays=10, verbose=verbose)
i +=1
dayCounter+=1
if dayCounter >= ReEvaluationInterval: dayCounter=0
tm.ProcessDay()
cash, asset = tm.Value()
print('Ending Value: ', cash + asset, '(Cash', cash, ', Asset', asset, ')')
return tm.CloseModel(plotResults=False, saveHistoryToFile=verbose)
def RunTradingModelSwingTrend(ticker: str,
startDate: str,
durationInYears: int,
totalFunds: int,
verbose: bool = False,
saveHistoryToFile: bool = True,
returndailyValues: bool = False):
#Give it a date range, some money, and a stock, it will execute a strategy and return the results
#minDeviationToTrade = .025
minActionableSlope = 0.002
trendState = 'Flat' #++,--,+-,-+,Flat
prevTrendState = ''
trendDuration = 0
modelName = 'SwingTrend' + '_' + ticker
tm = TradingModel(modelName, ticker, startDate, durationInYears,
totalFunds, verbose)
if not tm.modelReady:
print('Unable to initialize price history for model for ' +
str(startDate))
if returndailyValues: return pandas.DataFrame()
else: return totalFunds
else:
while not tm.ModelCompleted():
tm.ProcessDay()
p = tm.GetPriceSnapshot()
if not p == None:
available, buyPending, sellPending, longPositions = tm.GetPositionSummary(
)
maxPositions = available + buyPending + sellPending + longPositions
targetBuy = p.nextDayTarget * (1 + p.fiveDayDeviation / 2)
targetSell = p.nextDayTarget * (1 - p.fiveDayDeviation / 2)
if p.longEMASlope >= minActionableSlope and p.shortEMASlope >= minActionableSlope: #++ Positive trend, 70% long
trendState = '++'
if p.low > p.channelHigh: #Over Bought
if sellPending < 3 and longPositions > 7:
tm.PlaceSell(ticker, targetSell * (1.03), False,
10)
elif p.low < p.channelLow: #Still early
if buyPending < 3 and longPositions < 6:
tm.PlaceBuy(ticker, targetBuy, True)
if trendDuration > 1 and buyPending < 3:
tm.PlaceBuy(ticker, targetBuy, True)
else:
if buyPending < 3 and longPositions < 6:
tm.PlaceBuy(ticker, targetBuy, False)
if buyPending < 5 and longPositions < maxPositions:
tm.PlaceBuy(ticker, targetBuy, False)
elif p.longEMASlope >= minActionableSlope and p.shortEMASlope < minActionableSlope: #+- Correction or early downturn
trendState = '+-'
if p.low > p.channelHigh: #Over Bought, try to get out
if sellPending < 3 and longPositions > 7:
tm.PlaceSell(ticker, targetSell, False, 3)
elif p.low < p.channelLow and p.high > p.channelLow: #Deep correction
if sellPending < 3 and longPositions > 7:
tm.PlaceSell(ticker, targetSell, False, 3)
else:
pass
elif p.longEMASlope < -minActionableSlope and p.shortEMASlope < -minActionableSlope: #-- Negative trend, aim for < 30% long
trendState = '--'
if p.high < p.channelLow: #Over sold
if buyPending < 3 and longPositions < 6:
tm.PlaceBuy(ticker, targetBuy * .95, False, 2)
elif p.low < p.channelLow and p.high > p.channelLow: #Straddle Low, early down or up
pass
else:
if sellPending < 5 and longPositions > 3:
tm.PlaceSell(ticker, targetSell, True)
if trendDuration > 1:
tm.PlaceSell(ticker, targetSell, True)
if sellPending < 5 and longPositions > 3:
tm.PlaceSell(ticker, targetSell, False, 2)
tm.PlaceSell(ticker, targetSell, False, 2)
elif p.longEMASlope < (
-1 * minActionableSlope) and p.shortEMASlope < (
-1 *
minActionableSlope): #-+ Bounce or early recovery
trendState = '-+'
if p.high < p.channelLow: #Over sold
pass
elif p.low < p.channelLow and p.high > p.channelLow: #Straddle Low
if sellPending < 3 and longPositions > 3:
tm.PlaceSell(ticker, targetSell, False, 3)
else:
pass
else: #flat, aim for 70% long
trendState = 'Flat'
if p.low > p.channelHigh: #Over Bought
if sellPending < 3 and longPositions > 7:
tm.PlaceSell(ticker, targetSell * (1.03), False,
10)
elif p.high < p.channelLow: #Over sold
if buyPending < 3 and longPositions < 8:
tm.PlaceBuy(ticker, targetBuy, False, 5)
if buyPending < 4:
tm.PlaceBuy(ticker, targetBuy, False, 5)
else:
pass
if sellPending < 3 and longPositions > 7:
tm.PlaceSell(ticker, targetSell, False, 5)
if buyPending < 3 and longPositions < maxPositions:
tm.PlaceBuy(ticker, targetBuy, False, 5)
if trendState == prevTrendState:
trendDuration = trendDuration + 1
else:
trendDuration = 0
if tm.AccountingError(): break
if returndailyValues:
tm.CloseModel(verbose, saveHistoryToFile)
return tm.GetdailyValue() #return daily value
else:
return tm.CloseModel(verbose,
saveHistoryToFile) #return closing value
def RunPriceMomentum(tickerList: list,
startDate: str = '1/1/1982',
durationInYears: int = 36,
stockCount: int = 9,
ReEvaluationInterval: int = 20,
filterOption: int = 3,
longHistory: int = 365,
shortHistory: int = 90,
minPercentGain=0.05,
maxVolatility=.12,
portfolioSize: int = 30000,
returndailyValues: bool = False,
verbose: bool = False):
#Choose the stock with the greatest long term (longHistory days) price appreciation
#shortHistory is a shorter time frame (like 90 days) used differently by different filters
#ReEvaluationInterval is how often to re-evaluate our choices, ideally this should be very short and not matter, otherwise the date selection is biased.
startDate = datetime.datetime.strptime(startDate, '%m/%d/%Y')
endDate = startDate + datetime.timedelta(days=365 * durationInYears)
picker = StockPicker(startDate, endDate)
for t in tickerList:
picker.AddTicker(t)
tm = TradingModel(modelName='PriceMomentumShort_longHistory_' +
str(longHistory) + '_shortHistory_' + str(shortHistory) +
'_reeval_' + str(ReEvaluationInterval) + '_stockcount_' +
str(stockCount) + '_filter' + str(filterOption) + '_' +
str(minPercentGain) + str(maxVolatility),
startingTicker='^SPX',
startDate=startDate,
durationInYears=durationInYears,
totalFunds=portfolioSize,
tranchSize=portfolioSize / stockCount,
verbose=verbose)
dayCounter = 0
if not tm.modelReady:
print('Unable to initialize price history for PriceMomentum date ' +
str(startDate))
return 0
else:
while not tm.ModelCompleted():
currentDate = tm.currentDate
if dayCounter == 0:
print('\n')
print(currentDate)
c, a = tm.Value()
print(tm.modelName, int(c), int(a), int(c + a))
print('available/buy/sell/long', tm.PositionSummary())
tm.SellAllPositions(currentDate)
tm.ProcessDay()
dayCounter += 1
shortList = picker.GetHighestPriceMomentum(
currentDate,
longHistoryDays=longHistory,
shortHistoryDays=shortHistory,
stocksToReturn=stockCount,
filterOption=filterOption,
minPercentGain=minPercentGain,
maxVolatility=maxVolatility)
shortList = shortList[:stockCount]
print(shortList)
if len(shortList) > 0:
i = 0
ii = 0
while tm.TranchesAvailable(
) and i < 100: #Over long periods 100 will not be enough, 1000 would be better, but 100 also limits the impact of early gains, after a bit you are trading $300K
tm.PlaceBuy(ticker=shortList.index[ii],
price=1,
marketOrder=True,
expireAfterDays=10,
verbose=verbose)
i += 1
ii += 1
if ii >= len(shortList): ii = 0
tm.ProcessDay()
dayCounter += 1
if dayCounter >= ReEvaluationInterval: dayCounter = 0
cv1 = tm.CloseModel(plotResults=False,
saveHistoryToFile=((durationInYears > 1)
or verbose))
if returndailyValues:
return tm.GetDailyValue()
else:
return cv1
def RunPriceMomentumBlended(tickerList: list,
startDate: str = '1/1/1980',
durationInYears: int = 29,
stockCount: int = 9,
ReEvaluationInterval: int = 20,
longHistory: int = 365,
shortHistory: int = 90,
portfolioSize: int = 30000,
returndailyValues: bool = False,
verbose: bool = False):
#Uses blended option for selecting stocks using three different filters
startDate = datetime.datetime.strptime(startDate, '%m/%d/%Y')
endDate = startDate + datetime.timedelta(days=365 * durationInYears)
picker = StockPicker(startDate, endDate)
for t in tickerList:
picker.AddTicker(t)
tm = TradingModel(modelName='PriceMomentum_Blended_longHistory_' +
str(longHistory) + '_shortHistory_' + str(shortHistory) +
'_reeval_' + str(ReEvaluationInterval) + '_stockcount_' +
str(stockCount) + '_filterBlended_134',
startingTicker='^SPX',
startDate=startDate,
durationInYears=durationInYears,
totalFunds=portfolioSize,
tranchSize=portfolioSize / stockCount,
verbose=verbose)
dayCounter = 0
if not tm.modelReady:
print('Unable to initialize price history for PriceMomentum date ' +
str(startDate))
return 0
else:
while not tm.ModelCompleted():
currentDate = tm.currentDate
if dayCounter == 0:
print('\n')
print(currentDate)
c, a = tm.Value()
print(tm.modelName, int(c), int(a), int(c + a))
print('available/buy/sell/long', tm.PositionSummary())
tm.SellAllPositions(currentDate)
tm.ProcessDay()
dayCounter += 1
shortList1 = picker.GetHighestPriceMomentum(
currentDate,
longHistoryDays=longHistory,
shortHistoryDays=shortHistory,
stocksToReturn=int(stockCount / 3),
filterOption=1)
shortList2 = picker.GetHighestPriceMomentum(
currentDate,
longHistoryDays=longHistory,
shortHistoryDays=shortHistory,
stocksToReturn=int(stockCount / 3),
filterOption=2)
shortList3 = picker.GetHighestPriceMomentum(
currentDate,
longHistoryDays=longHistory,
shortHistoryDays=shortHistory,
stocksToReturn=int(stockCount / 3),
filterOption=4)
#shortList3 = picker.GetHighestPriceMomentum(currentDate, longHistoryDays=longHistory, shortHistoryDays=shortHistory, stocksToReturn=int(stockCount/3), filterOption=6, maxVolatility=.12)
shortList = pd.concat([shortList1, shortList2, shortList3])
shortList
print(shortList)
if len(shortList) > 0:
i = 0
ii = 0
while tm.TranchesAvailable() and i < 100:
tm.PlaceBuy(ticker=shortList.index[ii],
price=1,
marketOrder=True,
expireAfterDays=10,
verbose=verbose)
i += 1
ii += 1
if ii >= len(shortList): ii = 0
tm.ProcessDay()
dayCounter += 1
if dayCounter >= ReEvaluationInterval: dayCounter = 0
cv1 = tm.CloseModel(plotResults=False,
saveHistoryToFile=((durationInYears > 1)
or verbose))
if returndailyValues:
return tm.GetDailyValue()
else:
return cv1