General Discussion

• 

  Can Backtrader use an existing python algo generating daily buy/sell/hold signals across multiple equities and timespans?
  • run-out  

  4
  0
  Votes
  4
  Posts
  60
  Views

  

  Any ideas of where I might find sample implementation of factors for French-Fama model?
• J

  Assigning independent value for multi assets
  • jabbarabdullah  

  4
  0
  Votes
  4
  Posts
  24
  Views

  B

  Sincerely: you need to make up your mind as to what you want. What you post makes no sense. You probably have an idea about the direction in which you want to go. If your take-profit/stop-loss is based on the value increase/decrease of an asset you have acquired, record the initial value and exit when the value has gone beyond your limits.
• J

  Multi Assets Multi Order Executions Problem
  • jabbarabdullah  

  6
  0
  Votes
  6
  Posts
  63
  Views

  J

  @backtrader said in Multi Assets Multi Order Executions Problem: @jabbarabdullah said in Multi Assets Multi Order Executions Problem: Appreciate if you could explain further on what do you mean by 'NO'. You say the orders are getting executed twice and the answer is: NO @ab_trader said in Multi Assets Multi Order Executions Problem: You have two data feeds, you cycle thru them twice and print same info twice. Check if data name is same as order data name, then print. On the other side you don't need to go thru all data feeds. Just print order info. I quoted the code in notify_order where you do exactly what @ab_trader is telling you. You print the execution twice (which is not the same as two executions), because you loop through the datas in notify_order I see, got it. Removed the loop in notify_order and it worked. Thanks @backtrader and @ab_trader ! Just one more thing. when I tried executing the code with different date which the take profit and stop loss in my code should work, the numbers went crazy and it did not really stop the trade as expected. It worked great for a single asset data feed code but when I change it to multi assets data feeds, it went haywire. What went wrong? The single data feed code for comparison as follows: from __future__ import (absolute_import, division, print_function, unicode_literals) import backtrader as bt import datetime as dt import pytz import math cn1='UUUU' cn2='VVVV' csh=100000 stdt=dt.datetime(2019,06,28,9,30) enddt=dt.datetime(2019,06,28,16,00) SL=0.01 TP=2*SL SU=0.005 SUpct=SU*100 prop=1 #proportion of portfolio batch=10 #rounding entrytime = dt.time(9,45) exittime = dt.time(15,55) stakesize=10 lwbnd=40 upbnd=85 commis=0.05 TPpct=TP*100 SLpct=SL*100 def rounddown(x): return int(math.floor(x / 1)) * 1 class IBCommission(bt.CommInfoBase): """A :class:`IBCommision` charges the way interactive brokers does. """ params = (('stocklike', True), ('commtype', bt.CommInfoBase.COMM_FIXED),) def _getcommission(self, size, price, pseudoexec): return self.p.commission class PropSizer(bt.Sizer): """A position sizer that will buy as many stocks as necessary for a certain proportion of the portfolio to be committed to the position, while allowing stocks to be bought in batches (say, 100)""" params = {"prop": prop, "batch": batch} def _getsizing(self, comminfo, cash, data, isbuy): """Returns the proper sizing""" if isbuy: # Buying target = csh * self.params.prop # Ideal total value of the position price = data.close[0] shares_ideal = target / price # How many shares are needed to get target batches = int(shares_ideal / self.params.batch) # How many batches is this trade? shares = batches * self.params.batch # The actual number of shares bought if shares * price > cash: return 0 # Not enough money for this trade else: return shares else: # Selling return self.broker.getposition(data).size # Clear the position class TestStrategy(bt.Strategy): data_live = False def notify_data(self, data, status, *args, **kwargs): print('*' * 5, 'DATA NOTIF:', data._getstatusname(status), *args) if status == data.LIVE: self.data_live = True def log(self, txt, dt=None, vlm=None): dt = dt or self.datas[0].datetime.datetime(0) vlm = vlm or self.data.volume[0] print('%s, %s, %s, Volume, %s' % (len(self), dt.isoformat(), txt, vlm)) def __init__(self): self.dataclose = self.datas[0].close self.order = None self.buyprice = None self.buycomm = None self.sma0 = bt.indicators.SimpleMovingAverage(self.datas[0], period=20) self.sma = bt.indicators.SimpleMovingAverage(self.datas[0], period=54) self.rsi = bt.indicators.RelativeStrengthIndex(period=14,safediv=True, upperband=upbnd,lowerband=lwbnd) def notify_order(self, order): if order.status in [order.Submitted, order.Accepted]: return self.last_executed_price = order.executed.price if order.status in [order.Completed]: if order.isbuy(): self.log( 'BUY EXECUTED, Price: %.2f, Cost: %.2f, Comm %.2f' % (order.executed.price, order.executed.value, order.executed.comm)) self.buyprice = order.executed.price self.buycomm = order.executed.comm self.last_executed_price = order.executed.price else: # Sell self.log('SELL EXECUTED, Price: %.2f, Cost: %.2f, Comm %.2f' % (order.executed.price, order.executed.value, order.executed.comm)) self.last_executed_price = order.executed.price self.bar_executed = len(self) elif order.status in [order.Canceled, order.Margin, order.Rejected]: self.log('Order Canceled/Margin/Rejected') # Write down: no pending order self.order = None def notify_trade(self, trade): if not trade.isclosed: return self.log('OPERATION PROFIT, GROSS %.2f, NET %.2f' % (trade.pnl, trade.pnlcomm)) def next(self): pv = self.broker.get_value() self.stakes = abs(rounddown(pv/self.dataclose[0])) target = csh * prop # Ideal total value of the position price = self.data.close[0] shares_ideal = target / price # How many shares are needed to get target batches = int(shares_ideal / batch) # How many batches is this trade? shares = batches * batch if self.order: return if not self.position: if (pv >= csh*(1+TP)): return if (pv <= csh*(1-SL)): return if (self.data.datetime.time(0) >= exittime): return if ((self.rsi[0] <= lwbnd)and(pv > csh*(1-SL))and(self.data.datetime.time(0) >= entrytime)): print('rsi:', self.rsi[0]) print('pv: ', pv) self.log('BUY CREATE, %.2f, VLM BOUGHT: %.2f' % (self.dataclose[0], shares)) self.order = self.buy() else: if (self.dataclose[0] >= (self.last_executed_price*(1+SU))): print('rsi:', self.rsi[0]) print('pv: ', pv) self.log('SELL CREATE (>%.2fpct), %.2f, VLM BOUGHT: %.2f' % (SUpct, self.dataclose[0], self.position.size)) self.order = self.sell(exectype=bt.Order.Stop) else: if (self.rsi[0] >= upbnd): print('rsi:', self.rsi[0]) print('pv: ', pv) self.log('SELL CREATE (RSI>Upbnd), %.2f, VLM BOUGHT: %.2f' % (self.dataclose[0], self.position.size)) self.order = self.close() else: if (pv >= csh*(1+TP)): print('rsi:', self.rsi[0]) print('pv: ', pv) self.log('TAKE PROFIT VAL CREATE (>%.2fpct), %.2f, VLM BOUGHT: %.2f' % (TPpct, self.dataclose[0], self.position.size)) self.order = self.sell(exectype=bt.Order.StopLimit, price=self.dataclose[0]) else: if (pv <= csh*(1-SL)): print('rsi:', self.rsi[0]) print('pv: ', pv) self.log('STOPLOSS VAL CREATE (<%.2fpct), %.2f, VLM BOUGHT: %.2f' % (SLpct, self.dataclose[0], self.position.size)) self.order = self.sell(exectype=bt.Order.StopLimit, price=self.dataclose[0]) return else: if (self.data.datetime.time(0) >= exittime): print('rsi:', self.rsi[0]) print('pv: ', pv) self.log('EOD STOP, %.2f, VLM BOUGHT: %.2f' % (self.dataclose[0], self.position.size)) self.order = self.close(exectype=bt.Order.Stop) def runstrat(): # Get a pandas dataframe cerebro = bt.Cerebro() ibstore = bt.stores.IBStore(port=7497, host='127.0.0.1', clientId=12345) data0 = ibstore.getdata(dataname=cn1,fromdate=stdt, historical =True, useRTH=True, tz = pytz.timezone('US/Eastern'), todate=enddt, timeframe=bt.TimeFrame.Seconds, compression=15) cerebro.adddata(data0, name=cn1) data1 = ibstore.getdata(dataname=cn2,fromdate=stdt, historical =True, useRTH=True, tz = pytz.timezone('US/Eastern'), todate=enddt, timeframe=bt.TimeFrame.Seconds, compression=15) cerebro.adddata(data1, name=cn2) data1.plotinfo.plotmaster = data0 cerebro.broker.setcash(csh) comminfo = IBCommission(commission=commis) cerebro.broker.addcommissioninfo(comminfo) cerebro.addwriter(bt.WriterFile, csv=True, rounding=2, out="C:\\Users\\User\\Desktop\\Backtest Library\\TestResults.csv") start_value = cerebro.broker.getvalue() cerebro.addstrategy(TestStrategy) cerebro.addsizer(PropSizer) # Run over everything cerebro.run() # Plot the result cerebro.plot(volume=False) # Print out the starting conditions print(' ') print('--',cn1,'--') print('Start capital: %.2f' % start_value) # Print out the final result print('Final Portfolio Value: %.2f' % cerebro.broker.getvalue()) print('Final PnL Value: %.2f' % (cerebro.broker.getvalue()-start_value)) if __name__ == '__main__': runstrat() Appreciate your comment on this. Thanks!
• 

  Standard MACD vs Talib MACD - Different Values
  • Rodrigo Brito  

  4
  0
  Votes
  4
  Posts
  70
  Views

  B

  Indeed. You seem to be extremely new to indicators, timeframes, periods and resampling. ta-lib has a problem with the implementation of indicators with an "unstable" period, because of a cumbersome implementation, which is easily solved (industry de-facto standard) by using a seed value. An "Exponential Moving Average" is an example, because the definition of that average is that of a first order infinite impulse response filter. All values are always taken into account which leads to a recursive definition for updating values in real-time. Please read: https://en.wikipedia.org/wiki/Moving_average#Exponential_moving_average You can see that the problem is a ta-lib well-known issue: https://github.com/mrjbq7/ta-lib/issues/114 Back to square one: a good reason not to use ta-lib
• R

  error in dispatch
  • rastegarr  

  10
  0
  Votes
  10
  Posts
  2092
  Views

  B

  You can also shoot yourself in the knee and then see if it hurts. The problem is the data provider and the amount of symbols requested from it. Blocking a thread 15 seconds to try to overcome a structural problem is not the way. Feel free to do it yourself.
• 

  Reading PANDA CVS
  • Mariano Bengeldorf  

  3
  0
  Votes
  3
  Posts
  30
  Views

  

  Thanks @momentum!. You were right, I've activated the safediv and now it works!! Many thanks!
• 

  How to get MACD of 1 minute and 5 minute
  • mr-m0nst3r  

  5
  0
  Votes
  5
  Posts
  90
  Views

  

  @backtrader said in How to get MACD of 1 minute and 5 minute: The period is what the indicators use as loo Hi @backtrader , Thank you for your answer. First of all, this is my understanding of timeframe and period: timeframe is an attribute of the data feed; period is the "bar". If the data's timeframe is 1 minute, then 1 period is 1 minute's data (OCHLV); if data's timeframe is 5 minutes, then 1 bar is 5 minutes' data. I use bt.timeframe.Minutes all the time, so why bt.indicators.SMA(data, period=5) is not giving me the 5 minutes SMA value? I directly use the data, not a resampled one. I guess to get SMA of 5 minutes, I have to use 5 minutes timeframe data, seems clearer to me now. Thank you. And I used the Minutes data to resample to a 5 minutes timeframe using cerebro.resampledata(data, timeframe=bt.TimeFrame.Minutes, compression=5). Am I doing it right? And you're right, I'm confused when it comes to MACD. MACD=12-Period EMA − 26-Period EMA So, if the data's timeframe is 5 minutes, we'll get EMA(12) and EMA(26) of 5 minutes' timeframe, after calculation, it's the 5m MACD value, is that right? From the MACD equation, I see only period calculations, whether this period stands for 1 minute or 5 minutes, it's decided by the data's timeframe, which lead me to the conclusion that to get macd of 5m, the only way is to resample original 1m timeframe data to 5m timeframe data using compression=5. is that right? I'm stuck here, and forgive me if my expression confuses you. Thank you very much for your time.
• A

  Got nan value when resampling my own lines in data
  • Alicc  

  8
  0
  Votes
  8
  Posts
  56
  Views

  A

  @backtrader I think it's more convenient for me to put the lines into data, because what I am testing is option. I have several option hot contracts of current month and next month with different strikes from 2015-2019 in 1 minute freq. It's time consuming to compute greeks and iv every minute when testing, especially when I need to test different strategy parameters. And there are other statistics like strike price and put/call type. So I first compute all the statistics. Anyway, Can I directly put the statistics already computed into indicators. If not , how could I resample these statistic, just keep the last one . the followings are my lines, thanks
• B

  IB - Interactive Historical Data Download Bug report
  • barton05  

  4
  0
  Votes
  4
  Posts
  97
  Views

  B

  @barton05 said in IB - Interactive Historical Data Download Bug report: This code will have no data downloaded instead download data from 18 Jan - 1 Jun 2019. Don't download data before the start of the actual trading start.
• J

  question about futures margin
  • jlz  

  2
  0
  Votes
  2
  Posts
  89
  Views

  B

  The multiplier affects your winnings and losses and not the needed margin. See Docs - Commission Schemes
• C

  Anyone use backtrader to do live trading on Bitcoin exchange?
  crypto pairs trading • • CptanPanic  

  312
  4
  Votes
  312
  Posts
  98073
  Views

  

  @mariano-bengeldorf said in Anyone use backtrader to do live trading on Bitcoin exchange?: Hi again... Does anybody has a code to download minute data from BTC/USD since 2017/01/01 to a CVS. I am using the following, but it only downloads 14000 data points: import ccxt from datetime import datetime, timedelta, timezone import math import argparse import pandas as pd import ciso8601 import time def getUNIX(input_date): UNIX = int(time.mktime(ciso8601.parse_datetime(input_date).timetuple()) * 1000) return UNIX def parse_args(): parser = argparse.ArgumentParser(description='CCXT Market Data Downloader') parser.add_argument('-s','--symbol', type=str, required=True, help='The Symbol of the Instrument/Currency Pair To Download') parser.add_argument('-e','--exchange', type=str, required=True, help='The exchange to download from') parser.add_argument('-t','--timeframe', type=str, default='1d', choices=['1m', '5m','15m', '30m','1h', '2h', '3h', '4h', '6h', '12h', '1d', '1M', '1y'], help='The timeframe to download') parser.add_argument('--debug', action ='store_true', help=('Print Sizer Debugs')) return parser.parse_args() # Get our arguments args = parse_args() # Get our Exchange try: exchange = getattr (ccxt, args.exchange) () except AttributeError: print('-'*36,' ERROR ','-'*35) print('Exchange "{}" not found. Please check the exchange is supported.'.format(args.exchange)) print('-'*80) quit() # Check if fetching of OHLC Data is supported if exchange.has["fetchOHLCV"] != True: print('-'*36,' ERROR ','-'*35) print('{} does not support fetching OHLC data. Please use another exchange'.format(args.exchange)) print('-'*80) quit() # Check requested timeframe is available. If not return a helpful error. if (not hasattr(exchange, 'timeframes')) or (args.timeframe not in exchange.timeframes): print('-'*36,' ERROR ','-'*35) print('The requested timeframe ({}) is not available from {}\n'.format(args.timeframe,args.exchange)) print('Available timeframes are:') for key in exchange.timeframes.keys(): print(' - ' + key) print('-'*80) quit() # Check if the symbol is available on the Exchange exchange.load_markets() if args.symbol not in exchange.symbols: print('-'*36,' ERROR ','-'*35) print('The requested symbol ({}) is not available from {}\n'.format(args.symbol,args.exchange)) print('Available symbols are:') for key in exchange.symbols: print(' - ' + key) print('-'*80) quit() # Get data since = getUNIX('2017-01-01') data = exchange.fetch_ohlcv(args.symbol, args.timeframe, since) header = ['Timestamp', 'Open', 'High', 'Low', 'Close', 'Volume'] df = pd.DataFrame(data, columns=header).set_index('Timestamp') # Save it symbol_out = args.symbol.replace("/","") filename = '{}-{}-{}.csv'.format(args.exchange, symbol_out,args.timeframe) df.to_csv(filename) Thanks I was able to solve this one, here is the code for anyone who is interested in. import ccxt from datetime import datetime, timedelta, timezone import math import argparse import pandas as pd import time import os import sys def parse_args(): parser = argparse.ArgumentParser(description='CCXT Market Data Downloader') parser.add_argument('-s','--symbol', type=str, required=True, help='The Symbol of the Instrument/Currency Pair To Download') parser.add_argument('-e','--exchange', type=str, required=True, help='The exchange to download from') parser.add_argument('-t','--timeframe', type=str, default='1d', choices=['1m', '5m','15m', '30m','1h', '2h', '3h', '4h', '6h', '12h', '1d', '1M', '1y'], help='The timeframe to download') parser.add_argument('-d','--fromdate', type=str, default='2017-01-01 00:00:00', help='The start date to download from (yyyy-mm-dd hh:mm:ss)') parser.add_argument('--debug', action ='store_true', help=('Print Sizer Debugs')) return parser.parse_args() # Get our arguments args = parse_args() # Get our Exchange try: exchange = getattr (ccxt, args.exchange) () except AttributeError: print('-'*36,' ERROR ','-'*35) print('Exchange "{}" not found. Please check the exchange is supported.'.format(args.exchange)) print('-'*80) quit() # Check if fetching of OHLC Data is supported if exchange.has["fetchOHLCV"] != True: print('-'*36,' ERROR ','-'*35) print('{} does not support fetching OHLC data. Please use another exchange'.format(args.exchange)) print('-'*80) quit() # Check requested timeframe is available. If not return a helpful error. if (not hasattr(exchange, 'timeframes')) or (args.timeframe not in exchange.timeframes): print('-'*36,' ERROR ','-'*35) print('The requested timeframe ({}) is not available from {}\n'.format(args.timeframe,args.exchange)) print('Available timeframes are:') for key in exchange.timeframes.keys(): print(' - ' + key) print('-'*80) quit() # Check if the symbol is available on the Exchange exchange.load_markets() if args.symbol not in exchange.symbols: print('-'*36,' ERROR ','-'*35) print('The requested symbol ({}) is not available from {}\n'.format(args.symbol,args.exchange)) print('Available symbols are:') for key in exchange.symbols: print(' - ' + key) print('-'*80) quit() # ----------------------------------------------------------------------------- root = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__)))) sys.path.append(root + '/python') # ----------------------------------------------------------------------------- # common constants msec = 1000 minute = 60 * msec hold = 30 timedic = { '1m': minute, '5m': 5*minute, '15m': 15*minute, '30m': 30*minute, '1h': 60*minute, '2h': 2*60*minute, '3h': 3*60*minute, '4h': 4*60*minute, '6h': 6*60*minute, '12h': 12*60*minute, '1d': 24*60*minute, '1M': 30*24*60*minute, '1y': 365*24*60*minute } # ----------------------------------------------------------------------------- exchange.enableRateLimit = True exchange.rateLimit = 10000 # ----------------------------------------------------------------------------- from_timestamp = exchange.parse8601(args.fromdate) # ----------------------------------------------------------------------------- now = exchange.milliseconds() # ----------------------------------------------------------------------------- # Get data data = [] while from_timestamp < now: try: print(exchange.milliseconds(), 'Fetching candles starting from', exchange.iso8601(from_timestamp)) ohlcvs = exchange.fetch_ohlcv('BTC/USD', args.timeframe, from_timestamp) print(exchange.milliseconds(), 'Fetched', len(ohlcvs), 'candles') if len(ohlcvs) > 0: first = ohlcvs[0][0] last = ohlcvs[-1][0] print('First candle epoch', first, exchange.iso8601(first)) print('Last candle epoch', last, exchange.iso8601(last)) from_timestamp = ohlcvs[-1][0] + timedic.get(args.timeframe) # good data += ohlcvs except (ccxt.ExchangeError, ccxt.AuthenticationError, ccxt.ExchangeNotAvailable, ccxt.RequestTimeout) as error: print('Got an error', type(error).__name__, error.args, ', retrying in', hold, 'seconds...') time.sleep(hold) # ------------------------------------------------------------------------------- #create dataframe header = ['Timestamp', 'Open', 'High', 'Low', 'Close', 'Volume'] df = pd.DataFrame(data, columns=header).set_index('Timestamp') # Save it symbol_out = args.symbol.replace("/","") filename = '{}-{}-{}-{}.csv'.format(args.exchange, symbol_out,args.timeframe,args.fromdate[:10]) df.to_csv(filename) Thanks Mariano
• ?

  how to sum total size of buy/sell action. then displayed total size.
  • A Former User  

  2
  0
  Votes
  2
  Posts
  80
  Views

  

  If I understand what you want, you could: in ___init__() for the strategy initialize the total size attribute: self.totalSize = 0 Then in notify_order() something like: if order.status in [order.Completed]: self.totalSize += abs(order.executed.size) abs() in case your going short as well
• B

  Monte Carlos and CAR25 Simulation by Backtrader (Credit to Howard Bandy)
  • barton05  

  1
  1
  Votes
  1
  Posts
  66
  Views

  No one has replied

• 

  How to query data from InfluxDB into Backtrader?
  • Bozz srinikorn  

  1
  0
  Votes
  1
  Posts
  56
  Views

  No one has replied

• C

  Bug using Pandas hdf?
  • Chapachan  

  4
  0
  Votes
  4
  Posts
  1948
  Views

  

  @thatblokedave thanks, that's work for me.
• A

  Problem on multi datafeeds with different time range
  • Alicc  

  5
  0
  Votes
  5
  Posts
  144
  Views

  B

  Your problem it's called "survivorship bias". And the reason for not getting into next is that next is a method in which all buffers are guaranteed to deliver data. Docs - Operating the Platform There are several entries in the community discussing this, but there is a recent article which touches, amongst others, the topic https://www.backtrader.com/blog/2019-05-20-momentum-strategy/momentum-strategy/
• 

  Walk Forward Analysis Demonstration
  • Curtis Miller  

  7
  5
  Votes
  7
  Posts
  2192
  Views

  B

  @Mar-Co This is what I have done for the WFA. Its hard but fun. :) def createWFAReport(simParams,simulations): #Create simulation report format reportColumns=['grossProfit','grossAverageProfit','maxProfit', 'grossLoss','grossAverageLoss','maxLoss', 'netProfit','averageNetProfit','NAR', 'recoveryFactor','MDDLength','MDD', 'wonTrade','lossTrade','tradingTime', 'averageTradeTime','TradeNumber','maxValue', 'minValue','totalCommission'] simulationReport=pd.DataFrame(columns=reportColumns) #Loop Simulations to create summary for simulation in simulations: '''some Calculation is done here''' return simReport def WFASplit(self,trainBy='12m',testBy='3m', loopBy='m',overlap=True): startDate=self.index[0] endDate=self.index[-1] if trainBy[-1] is 'm': trainTime=relativedelta(months=int(trainBy[:-1])) else: raise ValueError if testBy[-1] is 'm': testTime=relativedelta(months=int(testBy[:-1])) else: raise ValueError assert ((relativedelta(endDate,startDate)-trainTime).days) > 0 if loopBy is 'm': test_starts = zip(rrule(MONTHLY, dtstart=startDate, until=endDate-trainTime,interval=int(testBy[:-1]))) else: raise ValueError for i in test_starts: startD=i[0] endD=i[0]+trainTime yield (self[(self.index>=startD)&(self.index<endD)], self[(self.index>=endD)&(self.index<endD+testTime)]) return None def runTrain(trainTestGenerator,_ind,stockName): WFATrainResult=[] for train,test in trainTestGenerator: logger.debug('{} Training Data:{} to {}'.format(stockName,pd.DatetimeIndex.strftime(train.head(1).index,'%Y-%m-%d'),pd.DatetimeIndex.strftime(train.tail(1).index,'%Y-%m-%d'))) #Generate Indicator ResultSet trainer = bt.Cerebro(cheat_on_open=True, stdstats=False, optreturn=False) trainer.broker.set_cash(10000) #Add Commission IB = params['commission'](commission=0.0) trainer.broker.addcommissioninfo(IB) #Below Analyzer are used to calculate the Recovery Ratio trainer.addanalyzer(btanalyzers.TradeAnalyzer, _name='TradeAn') trainer.addanalyzer(recoveryAnalyzer,timeframe=params['analysisTimeframe'], _name='recoveryFac') trainer.addanalyzer(WFAAn,_name='WFAAna') trainer.addanalyzer(btanalyzers.TimeReturn,timeframe=bt.TimeFrame.Months,_name='TR') #SetBroker trainer.broker.set_checksubmit(False) #Copy for tester tester = deepcopy(trainer) #Optimize Strategy trainingFile='{}/WFA' trainer.optstrategy(trainingIdea, inOrOut=(params['inOrOut'],), selfLog=(params['selfLog'],), indName=(row.indicator,), indFormula=(_ind['formula'],), entryExitPara=(_ind['entryExitParameters'],), indOutName=(_ind['indValue'],), nonOptParams=(None,), resultLocation=(params['resultLocation'],), timeString=(params['timeString'],), market=(row.market,), **optt) trainData = bt.feeds.PandasData(dataname=train) # Add a subset of data. trainer.adddata(trainData) optTable = trainer.run() final_results_list = [] for run in optTable: for x in run: x.params['res']=x.analyzers.WFAAna.get_analysis() final_results_list.append(x.params) _bestWFA=pd.DataFrame.from_dict(final_results_list,orient='columns').sort_values('res',ascending=False).iloc[0].to_dict() bestTrainParams = {key: _bestWFA[key] for key in _bestWFA if key not in ['market','inOrOut','resultLocation','selfLog','timeString','res']} bestTrainParams=pd.DataFrame(bestTrainParams,index=[0]) bestTrainParams['trainStart']=train.iloc[0].name bestTrainParams['trainEnd']=train.iloc[-1].name bestTrainParams['testStart']=test.iloc[0].name bestTrainParams['testEnd']=test.iloc[-1].name WFATrainResult.append(bestTrainParams) WFATrainResult=pd.concat(WFATrainResult) return WFATrainResult def runTest(params, WFATrainResult,_ind,datafeed,stockName): #Generate Indicator ResultSet tester = bt.Cerebro(cheat_on_open=True) tester.broker.set_cash(10000) #Add Commission IB = params['commission'](commission=0.0) tester.broker.addcommissioninfo(IB) #SetBroker tester.broker.set_checksubmit(False) logger.debug('{} Start Testing'.format(stockName)) OneSimHandler=logging.FileHandler(filename='{}/simulation/{}_{}_test.log'.format(params['resultLocation'], str(stockName),str(row.indicator))) OneSimHandler.setLevel(logging.DEBUG) OneSimHandler.setFormatter(logging.Formatter("%(asctime)s:%(relativeCreated)d - %(message)s")) oneLogger.addHandler(OneSimHandler) tester.addstrategy(trainingIdea, inOrOut=params['inOrOut'], selfLog=params['selfLog'], indName=row.indicator, indFormula=_ind['formula'], entryExitPara=_ind['entryExitParameters'], indOutName=_ind['indValue'], nonOptParams=None, resultLocation=params['resultLocation'], timeString=params['timeString'], market=market, WFATestParams=WFATrainResult) data=bt.feeds.PandasData(dataname=datafeed) tester.adddata(data,name=stockName) #Add analyzers for Tester tester.addanalyzer(btanalyzers.DrawDown, _name='MDD') tester.addanalyzer(btanalyzers.TradeAnalyzer, _name='TradeAn') tester.addanalyzer(btanalyzers.SQN, _name='SQN') tester.addanalyzer(recoveryAnalyzer,timeframe=params['analysisTimeframe'], _name='recoveryFac') tester.addanalyzer(ITDDAnalyzer,_name='ITDD') tester.addanalyzer(simpleAn,_name='simpleAna') tester.addanalyzer(btanalyzers.TimeReturn,timeframe=bt.TimeFrame.Months,_name='TR') # Run and Return Cerebro cere=tester.run()[0] _report=cere.analyzers.simpleAna.writeAnalysis(bnhReturn) oneLogger.removeHandler(OneSimHandler) if params['plotGraph']: plotSimGraph(tester,params,stockName,row.indicator) return _report if __name__ == "__main__": session='WinsWFAProd' stockList, indicatorDict, params = jsonConfigMap(session) params['timeString']='0621_113833' params['topResult']=pd.read_csv('{}{}/consoReport.csv'.format(params['resultLocation'],params['timeString']),index_col=0) params['resultLocation']+=params['timeString']+'/WFA' simulations=[] try: #Create Folder shutil.rmtree(params['resultLocation']) except FileNotFoundError : pass except PermissionError: pass os.makedirs(params['resultLocation'],exist_ok=True) for element in ['order','trade','mr','ohlc','simulation','bestTrain','graph']: os.makedirs(params['resultLocation']+'/'+element,exist_ok=True) #Create master Log handler = logging.FileHandler(filename='{}/Master.log'.format(params['resultLocation'])) handler.setFormatter(logging.Formatter('%(asctime)s:%(name)s - %(levelname)s {}- %(message)s')) logger = logging.getLogger() logger.setLevel(logging.DEBUG) logger.addHandler(handler) #Create OptReport Log reportHandler = logging.FileHandler(filename='{}/optReport.csv'.format(params['resultLocation'])) reportHandler.setFormatter(logging.Formatter('%(message)s')) reportHandler.setLevel(logging.INFO) reportLogger=logging.getLogger('report') reportLogger.addHandler(reportHandler) simResultColumns=['stockName','market','indicator', 'grossProfit','grossAverageProfit','maxProfit', 'grossLoss','grossAverageLoss','maxLoss', 'netProfit','averageNetProfit','NAR','profitFactor', 'recoveryFactor','MDDLength','MDD', 'selfMDD','winRate','tradingTimeRatio', 'averageTradingBar','tradeNumber','maxValue', 'minValue','initialv','totalCommission', 'barNumber','expectancy100','bnhReturn','bnhRatio'] reportLogger.info(str(simResultColumns).strip("[]").replace("'","").replace(" ","")) #Create Simulation Log oneLogger=logging.getLogger('oneLogger') oneLogger.propagate=False postHandler=logging.StreamHandler() postHandler.setLevel(logging.INFO) if params['selfLog']: oneLogger.setLevel(logging.DEBUG) else: oneLogger.setLevel(logging.INFO) oneLogger.addHandler(postHandler) #Record Start Time startTime=time.time() for row in params['topResult'].itertuples(): simParams=pd.DataFrame(columns=['startTime','endTime','Parameter']) indicator=indicatorDict[row.indicator] stockName=row.stockName market=row.market try: optt = eval(indicator['optParam']) except: logger.info('{}: Indicator does not have WFA parameters and skipped'.format(row.indicator)) continue datafeed=feeder(stockName, market,params) bnhReturn=round(datafeed.iloc[-1]['close'] - datafeed.iloc[0]['open'],2) #Extract Feeder from Data to save time for multi-simulation print('Start WFA for {}-{} from {} to {}'.format(stockName,row.indicator,datafeed.iloc[0].name,datafeed.iloc[-1].name)) trainTestGenerator=WFASplit(datafeed,trainBy='8m',testBy='8m',loopBy='m') _ind=indicatorDict[row.indicator] #Training WFATrainResult=runTrain(trainTestGenerator,_ind,stockName) WFATrainResult=pd.DataFrame.from_records(WFATrainResult) WFATrainResult.to_csv('{}/bestTrain/{}_{}_train.csv'.format(params['resultLocation'],stockName,row.indicator),index=False) #TESTING _report=runTest(params, WFATrainResult,_ind, datafeed,stockName) #Consolidate T if _report[0] is not None: reportLogger.info(str(_report).strip("[]").strip(" ").replace(" ","").replace("'","")) simulations.append(_report) #After simulations simulations=pd.DataFrame(simulations) reportColumns=['stockName','market','indicator', 'grossProfit','grossAverageProfit','maxProfit', 'grossLoss','grossAverageLoss','maxLoss', 'netProfit','averageNetProfit','NAR', 'profitFactor','recoveryFactor','MDDLength', 'selfMDD','winRate','tradingTimeRatio', 'averageTradingBar','tradeNumber','maxValue', 'minValue','initialv','totalCommission', 'barNumber','expectancy100','bnhReturn', 'bnhRatio'] simulations.columns=reportColumns consoResult=cr.scoring(simulations) consoResult=consoResult.sort_values(['res'],ascending=False) consoResult.sort_values('res').to_csv('{}/optReport.csv'.format(params['resultLocation']),index=False) timeRequired=time.time()-startTime print('timeRequired={:.2f}s'.format(timeRequired))
• A

  New Documentation Site Search Function
  • algoguy235  

  2
  0
  Votes
  2
  Posts
  163
  Views

  B

  Google Custom Search has been integrated. You can now use it to look for quicknotify (for example)
• S

  For IB, now there is TWS API, is IBPy still valid?
  • sdmovie  

  3
  0
  Votes
  3
  Posts
  219
  Views

  J

  Agree with you @backtrader... If it ain't broke, don't fix it
• 

  STOP TRAIL order questions.
  • ciceron  

  5
  0
  Votes
  5
  Posts
  212
  Views

  

  @ciceron hi again. I was wrong. Oanda broker has StopTrail implementation. if order.exectype == bt.Order.StopTrail: okwargs['distance'] = order.trailamount https://github.com/ftomassetti/backtrader-oandav20/blob/master/btoandav20/stores/oandav20store.py I didn't have any practice with Oanda broker, so i don't know how stoptrail works on Oanda side. But your understanding is correct for backtesting.