Indicators/Strategies/Analyzers

• M

  Multi-Data Indicator implementation help.
  • Mango Loco  

  8
  0
  Votes
  8
  Posts
  327
  Views

  B

  @mango-loco said in Multi-Data Indicator implementation help.: Is the above code the most efficient/cleanest/best way to go about handling the operations I'm doing? The fact is that backtrader allows you to do it in a declarative way or in an iterative way (and to mix both). There seems to a fixation out thee in many cases in using numpy (imported as np) for whatever reason which I fail to understand, even if it is meant to only calculate a single value. But if it suits you, there is no reason to do it otherwise. @mango-loco said in Multi-Data Indicator implementation help.: Is there a way/hack to eliminate the "instantiation lag" that occurs with indicators in BT? Is it possible to pass in data from before the start of the backtest so that any indicators are already fully instantiated at t=0? There is NO lag. Your reasoning and understanding is wrong. If you want a 20-period moving average you need 20 data points to calculate it. Full stop. If you want your indicators at YOUR DEFINED t=0 to be available, work with a data set which has an earlier start datetime.
• M

  Can BT resample the data in the user defined Indicator?
  • mothee  

  6
  0
  Votes
  6
  Posts
  329
  Views

  M

  @backtrader @ab_trader I finally implement the Indicator as below code, it seems work. but two small questions, when the ATR is coupled? and how do I set the minperiod when data with different time frames, I worked around by using len(self.l.art)>0, else it would raise errors. import numpy as np import backtrader as bt from ext.ult.ult import * from ext.Ind.OCHLInd import OCHLInd # https://www.tradingview.com/script/mtWhbCRf-RS-MTF-Volatility-Stop-V0/ class MtfVSInd(bt.Indicator): """ """ lines = ('trend', 'vs', 'max_close', 'min_close', 'atr') params = ( ('period', 5), ('factor', 1.618), ('mode', 0) # debug modes ) plotinfo = dict(subplot=False) plotlines = dict(vs=dict(_plotskip=False), trend=dict(_plotskip=True, _method='bar', alpha=0.50, width=1.0), max_close=dict(_plotskip=True, color='black'), min_close=dict(_plotskip=True, color='red'), atr=dict(_plotskip=True), ) def _plotinit(self): if self.p.mode == 1: self.plotinfo.subplot = True self.plotlines.vs._plotskip = True self.plotlines.trend._plotskip = True self.plotlines.max_close._plotskip = True self.plotlines.min_close._plotskip = True self.plotlines.atr._plotskip = False elif self.p.mode == 2: self.plotinfo.subplot = True self.plotlines.vs._plotskip = True self.plotlines.trend._plotskip = False self.plotlines.max_close._plotskip = True self.plotlines.min_close._plotskip = True self.plotlines.atr._plotskip = True elif self.p.mode == 3: self.plotinfo.subplot = True self.plotlines.vs._plotskip = True self.plotlines.trend._plotskip = True self.plotlines.max_close._plotskip = False self.plotlines.min_close._plotskip = False self.plotlines.atr._plotskip = True def __init__(self): # the ATR is Day time frame self.l.atr = bt.ind.ATR(self.data1, period=self.p.period) # get the avg(high+low) self.hl = OCHLInd(self.data0, type='hl') super(MtfVSInd, self).__init__() def next(self): if len(self.l.atr) > 0: if np.isnan(self.l.trend[-1]): self.l.trend[0] = 1 elif self.data0.l.close[0] > self.l.vs[-1]: self.l.trend[0] = 1 elif self.data0.l.close[0] < self.l.vs[-1]: self.l.trend[0] = -1 else: self.l.trend[0] = self.l.trend[-1] chg_trend = self.l.trend[0] - self.l.trend[-1] if chg_trend != 0: self.l.max_close[0] = self.data0.l.close[0] elif self.data0.l.close[0] >= self.l.max_close[-1]: self.l.max_close[0] = self.data0.l.close[0] else: self.l.max_close[0] = self.l.max_close[-1] if chg_trend != 0: self.l.min_close[0] = self.data0.l.close[0] elif self.data0.l.close[0] <= self.l.min_close[-1]: self.l.min_close[0] = self.data0.l.close[0] else: self.l.min_close[0] = self.l.min_close[-1] if np.isnan(self.l.vs[-1]): self.l.vs[0] = self.hl[0] elif chg_trend > 0: self.l.vs[0] = self.data0.l.close[0] - self.p.factor * self.l.atr[0] elif self.l.trend[0] > 0: self.l.vs[0] = max(self.l.vs[-1], self.l.max_close[0] - self.p.factor * self.l.atr[0]) elif chg_trend < 0: self.l.vs[0] = self.data0.l.close[0] + self.p.factor * self.l.atr[0] elif self.l.trend[0] < 0: self.l.vs[0] = min(self.l.vs[-1], self.l.min_close[0] + self.p.factor * self.l.atr[0]) else: self.l.vs[0] = self.l.vs[-1]
• L

  Accessing analysers in a strategy
  • livecat  

  4
  0
  Votes
  4
  Posts
  271
  Views

  B

  Because Strategy.stop() is a place where the strategy could still (in theory) do something which affects what an analyzer sees. For analyzers which calculate things during next it's a good point, for analyzers which wait until everything is complete, you will have to fetch it afterwards.
• K

  HADelta code correction?
  • kicking_potato88  

  2
  0
  Votes
  2
  Posts
  185
  Views

  B

  Indeed. Will be soon corrected.
• 

  Let strategies communicate with each other
  strategy multiple data • • Vincere  

  4
  0
  Votes
  4
  Posts
  285
  Views

  B

  The broker is a single account broker. All Strategies see the same value for a given asset. If that's not of your liking, your best bet will be to use a common parameter which can be used as a communication channel, like a dict or a list which will be written to by strategy A and read by B.
• A

  backtrader.linebuffer.LineBuffer object at
  • abyaby  

  4
  0
  Votes
  4
  Posts
  287
  Views

  A

  @backtrader Thanks that you mentioned the Limit Order. I thought, it doesnt matter, if I add cerebro.broker.setcommission(commission=0.002) Regarding unkown data I have backtest self.data.close and self.data.close[0] again. PnL with Limit order are different PnL with Market order are the same. What could it mean? I was especting data.close with market order a random number but surprisingly its the same. data.close[0] vs. data.close order.market -478.3 order.market -478.3 order.limit -634.26 order.limit 2370.7 And if some one know why data.close and data.close[0] have defferent limit order PnL ?
• 

  An indicator that knows if I have a position or not
  • Kristofer Monisit  

  4
  0
  Votes
  4
  Posts
  308
  Views

  M

  I even like better the _nextforce = True pointed out to me by @backtrader some time ago. class CustomIndicator(bt.Indicator): _nextforce = True ...
• 

  Pairs Trading - my research and experiments
  • deandree  

  1
  0
  Votes
  1
  Posts
  280
  Views

  No one has replied

• 

  Convincing strategy to buy at close of current bar in backtest
  • RandyT  

  8
  0
  Votes
  8
  Posts
  3898
  Views

  

  @backtrader said in Convincing strategy to buy at close of current bar in backtest: The observer in charge of plotting only sees the notification in the next cycle. Check the order.executed.dt field and you will see the order was properly executed on the right bar (even if the observer doesn't know it) @ab_trader said in Convincing strategy to buy at close of current bar in backtest: The order is executed as required, it is just the plotting and it confuses a bit. I know the plotting is not the main goal of the backtrader, but sometimes visual check of the signals/executions is really helpful. Hi. I faced the same issue. Does anybody know how to force the plotting to draw sell/buy signals in right place (with cheat-on-close mode enabled) ?
• 

  Plotting debug lines
  • Kas David  

  4
  0
  Votes
  4
  Posts
  197
  Views

  B

  There is no "next simplest" thing. Docs - Plotting - Section Plotted Elements You either plot a data feed, an indicator or an observer and they all happen to be classes. backtrader is a framework which is completely OO-based/oriented.
• B

  Zigzag Indicator
  • bigdavediode  

  10
  0
  Votes
  10
  Posts
  1714
  Views

  B

  I think you are confused in general. backtrader doesn't work with that notation. It uses the pivot point 0 as the current moment of time and negative indices point to the last (hence "past") values. If you want to get a complete array of past values, you will get a standard Python array. - Docs - Platform Concepts
• 

  It can be done?
  • Vasu Dev  

  2
  0
  Votes
  2
  Posts
  187
  Views

  B

  Again, you may want to read the Docs - Quickstart Guide
• 

  Building SMA on top of RSI
  • Vasu Dev  

  2
  0
  Votes
  2
  Posts
  223
  Views

  B

  That sounds almost like sacrilege. Who would dare doing such a thing ... Fo sure not those who dare reading the Docs - Quickstart Guide
• M

  indicator with dynamically changing data
  • momentum  

  5
  0
  Votes
  5
  Posts
  811
  Views

  M

  I am trying to generate a dynamically changing data feed based on the RollOver class but it does not work always correctly when trying to update the history at the same time (self.params.update_history_upon_roll set to True). For example when moving to a data feed with longer history it gives wrong results. Any ideas? Thanks def _roll(self): if self._ds: clock1 = self.lines[0].lencount self._dexp = self._d self._d = self._ds.pop(0) self._dts.pop(0) clock3 = self._d.lines[0].lencount if self.params.update_history_upon_roll: if clock3 < clock1: self._d.lines.rewind() for i in range(len(self.lines.lines)): self.lines[i].array = self._d.lines[i].array[:] self.lines[i].lencount = self._d.lines[i].lencount self.lines[i].idx = self._d.lines[i].idx if clock3 < clock1: self.lines.advance() return True else: return False def _load(self): while self._d is not None: if self._doroll: self._roll() self._doroll = False _next = self._d.next() if _next is None: # no values yet, more will come continue if _next is False: # no values from current data src if self._ds: self._d = self._ds.pop(0) self._dts.pop(0) else: self._d = None continue dt0 = self._d.datetime.datetime() # current dt for active data # Synchronize other datas using dt0 for i, d_dt in enumerate(zip(self._ds, self._dts)): d, dt = d_dt while dt < dt0: if d.next() is None: continue self._dts[i] = dt = d.datetime.datetime() # Move expired future as much as needed while self._dexp is not None: if not self._dexp.next(): self._dexp = None break if self._dexp.datetime.datetime() < dt0: continue if self._dexp is None and self._checkdate(dt0, self._d): # rule has been met ... check other factors only if 2 datas # still there if self._ds and self._checkcondition(self._d, self._ds[0]): if self.params.upon_condition: self._roll() else: self._doroll = True # Fill the line and tell we die for i in range(len(self.lines.lines)): self.lines[i][0] = self._d.lines[i][0] return True # Out of the loop -> self._d is None, no data feed to return from return False
• S

  Multiple Data Feeds (Instruments) on a single strategy
  • satheesh7134  

  5
  0
  Votes
  5
  Posts
  1716
  Views

  B

  @satheesh7134 said in Multiple Data Feeds (Instruments) on a single strategy: print('BUY Signal:, %.2f' % self.getdatabyname(datafeed).close[0]) self.order = self.buy() He wasn't telling the system what to buy (or sell). See Docs - Strategy and the reference for buy / sell or as pointed out by @Divyanshu-Bagga @divyanshu-bagga said in Multiple Data Feeds (Instruments) on a single strategy: @satheesh7134 Go through the multi data feeds example. @param-jeet said in Multiple Data Feeds (Instruments) on a single strategy: In that case, wouldn't cerebro.run( ) execute twice? If I run twice then I would get two different datasets. What if, I want to execute this strategy on both data sets and generate single report as one single portfolio. You enumerate within the strategy.
• 

  Stochastic RSI indicator.
  • Nikolai  

  10
  3
  Votes
  10
  Posts
  1482
  Views

  

  @Tomas Hi. I wrote an example strategy for you. You can access the indicator lines through the name in __init__ method. class CryptoStochRSI(bt.Strategy): params = ( ('stoch_k_period', 3), ('stoch_d_period', 3), ('stoch_rsi_period', 14), ('stoch_period', 14), ('stoch_upperband', 80.0), ('stoch_lowerband', 20.0), ('take_profit', 0.04), ('stop_loss', 0.01), ('size', 20), ('debug', False), ) def __init__(self): self.stoch = StochasticRSI(k_period=self.p.stoch_k_period, d_period=self.p.stoch_d_period, rsi_period=self.p.stoch_rsi_period, stoch_period=self.p.stoch_period, upperband=self.p.stoch_upperband, lowerband=self.p.stoch_lowerband) def next(self): if not self.position: close = self.data.close[0] price1 = close price2 = price1 - self.p.stop_loss * close price3 = price1 + self.p.take_profit * close if self.stoch.l.fastk[0] < 20: self.buy_bracket(price=price1, stopprice=price2, limitprice=price3, exectype=bt.Order.Limit) if self.stoch.l.fastk[0] > 80: self.sell_bracket(price=price1, stopprice=price3, limitprice=price2, exectype=bt.Order.Limit)
• R

  self.getposition(d) not equal to self.getpositionbyname(d._name)
  • Rahul Savani  

  8
  0
  Votes
  8
  Posts
  418
  Views

  R

  So managed to get an MWE. The problem relates to 0 prices. The problem goes away if those prices as empty in the csv and treated by NaNs by backtrader. This is really as simple as I can get it, hopefully you agree: Contents of input file S1.csv: date,OPEN,CLOSE 1996-01-01,1.0,1.0 1996-01-02,1.0,1.0 1996-01-03,1.0,1.0 Contents of input file S2.csv: date,OPEN,CLOSE 1996-01-01,0.0,0.0 1996-01-02,0.0,0.0 1996-01-03,0.0,0.0 Contents of input file S3.csv: date,OPEN,CLOSE 1996-01-01,, 1996-01-02,, 1996-01-03,, Source code for MWE: import backtrader as bt def add_data(cerebro, symbols): for symbol in symbols: fname = 'DATA_S/%s.csv' % symbol data = bt.feeds.GenericCSVData( dataname=fname, dtformat=('%Y-%m-%d'), datetime=0, open=1, close=2, high=-1, low=-1, volume=-1, openinterest=-1 ) cerebro.adddata(data) class DrStrategy(bt.Strategy): def next(self): print(self.data.datetime.date(0)) for i, d in enumerate(self.datas): if i == 0: self.buy(data=d) print("Datas %s, position size: %s" % (i, self.getposition(d).size)) if __name__ == "__main__": cerebro = bt.Cerebro(cheat_on_open=True) cerebro.addstrategy(DrStrategy) # position reported for datas 1 erroneously same as for datas 0 add_data(cerebro, symbols=['S1', 'S2']) # looks ok (missing entries in S.3 instead of 0s) # add_data(cerebro, symbols=['S1', 'S3']) results = cerebro.run() Output from running it with add_data(cerebro, symbols=['S1', 'S2']), where we get the problem: 1996-01-01 Datas 0, position size: 0 Datas 1, position size: 0 1996-01-02 Datas 0, position size: 1 Datas 1, position size: 1 1996-01-03 Datas 0, position size: 2 Datas 1, position size: 2 Output from running it with add_data(cerebro, symbols=['S1', 'S3']), where we do not get the problem: 1996-01-01 Datas 0, position size: 0 Datas 1, position size: 0 1996-01-02 Datas 0, position size: 1 Datas 1, position size: 0 1996-01-03 Datas 0, position size: 2 Datas 1, position size: 0 Any feedback, comments, suggestions welcome. Happy to investigate anything related that might be useful. Of course, I will personally be careful to avoid 0 prices. Thanks, --Rahul.
• C

  Force exit or exit open position according to risk management
  • codesutras  

  5
  0
  Votes
  5
  Posts
  328
  Views

  B

  @codesutras said in Force exit or exit open position according to risk management: Exit the the open position (or Force Exit ) when current value of portfolio drop by rules defined by RMS. Do as indicated by @ab_trader. Get value and exit when it's below your threshold. @codesutras said in Force exit or exit open position according to risk management: If Possible, perform this operation during intraday price move too. (Note: I've tried to achieve this through Cheat_on_open option. But, that's crazy configuration and it has changed performance of my strategy drastically.) @codesutras said in Force exit or exit open position according to risk management: Also, I'm working with csv feed for daily data. You want to exit intraday but you have daily data. There is physical impossibility here. Cheat on open is meant to let you calculate an entry against the opening price (which you may approximately experience during the pre-opening auction). It has nothing to do with intraday
• R

  One for the cookbook: On Balance Volume
  • Rudi  

  1
  0
  Votes
  1
  Posts
  217
  Views

  No one has replied

• 

  Cross-Sectional Mean Reversion Strategy
  strategy mean reversion universe spy • • teddykoker  

  2
  1
  Votes
  2
  Posts
  511
  Views

  B

  Thanks for the contribution and nice post.