Hello,

I'm switching my strategy to paper trading on interactivebrokers, the strangest thing is happening, this is my error as logged:

--------------------

---> pavement.run
Traceback (most recent call last):
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/ib/ext/EClientSocket.py", line 958, in placeOrder
    self.send(order.m_tif)
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/ib/lib/overloading.py", line 82, in __call__
    return func(*args)
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/ib/ext/EClientSocket.py", line 1754, in send_1
    self.m_dos.write(val)
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/ib/lib/__init__.py", line 152, in write
    send(pack('!c', char))
BrokenPipeError: [Errno 32] Broken pipe

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/paver/tasks.py", line 201, in _run_task
    return do_task()
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/paver/tasks.py", line 198, in do_task
    return func(**kw)
  File "pavement.py", line 133, in run
    raise SystemExit(main([CODE_DIRECTORY] + args))
  File "/Users/me/Documents/me/trading/trader/trader/main.py", line 79, in main
    raise error
  File "/Users/me/Documents/me/trading/trader/trader/main.py", line 76, in main
    action(options)
  File "/Users/me/Documents/me/trading/trader/trader/main.py", line 40, in _live
    live.run()
  File "/Users/me/Documents/me/trading/trader/trader/live.py", line 59, in run
    self.stats = self.cerebro.run()
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/backtrader/cerebro.py", line 1127, in run
    runstrat = self.runstrategies(iterstrat)
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/backtrader/cerebro.py", line 1298, in runstrategies
    self._runnext(runstrats)
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/backtrader/cerebro.py", line 1630, in _runnext
    strat._next()
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/backtrader/strategy.py", line 347, in _next
    super(Strategy, self)._next()
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/backtrader/lineiterator.py", line 271, in _next
    self.next()
  File "/Users/me/Documents/me/trading/trader/trader/strategies/BankTradingStrategy.py", line 190, in next
    self.short_trade = self.place_short()
  File "/Users/me/Documents/me/trading/trader/trader/strategies/BankTradingStrategy.py", line 115, in place_short
    limitprice=price - limitdistance)
  File "/Users/me/Documents/me/trading/trader/trader/SupportStrategy.py", line 43, in sell_bracket
    orders = super(SupportStrategy, self).sell_bracket(**kargs)
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/backtrader/strategy.py", line 1217, in sell_bracket
    o = self.sell(**kargs)
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/backtrader/strategy.py", line 969, in sell
    **kwargs)
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/backtrader/brokers/ibbroker.py", line 391, in sell
    return self.submit(order)
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/backtrader/brokers/ibbroker.py", line 337, in submit
    self.ib.placeOrder(order.m_orderId, order.data.tradecontract, order)
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/backtrader/stores/ibstore.py", line 1286, in placeOrder
    self.conn.placeOrder(orderid, contract, order)
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/ib/lib/__init__.py", line 60, in inner
    return func(*args, **kwds)
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/ib/ext/EClientSocket.py", line 1155, in placeOrder
    self.error_0(id, EClientErrors.FAIL_SEND_ORDER, str(e))
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/ib/lib/__init__.py", line 60, in inner
    return func(*args, **kwds)
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/ib/ext/EClientSocket.py", line 1709, in error_0
    self.m_anyWrapper.error(id, errorCode, errorMsg)
  File "/Users/me/Documents/me/trading/trader/trader_venv/lib/python3.7/site-packages/ib/lib/overloading.py", line 82, in __call__
    return func(*args)
TypeError: error() takes 2 positional arguments but 4 were given

I can reproduce this with:

        (Pdb) self.sell_bracket(price= 1.1350936200039081, exectype= 4, stopprice= 1.1373796835411358, limitprice= 1.1298980210556635)
        *** TypeError: error() takes 2 positional arguments but 4 were given
        (Pdb) self.sell_bracket(price= 1.1350936200039081, exectype= 4, stopprice= 1.1373796835411358, limitprice= 1.1298980210556635)
        [<backtrader.brokers.ibbroker.IBOrder object at 0x12b0b9668>, <backtrader.brokers.ibbroker.IBOrder object at 0x12b0b9908>, <backtrader.brokers.ibbroker.IBOrder object at 0x12b0b9ac8>]
        (Pdb)

basically sell_bracket fails only at the first call. the following ones see no problems

any ideas?

Hello,

When plotting the results of a backtest the broker cash and value in the top portion are on the same scale. like in the screenshot

With forex broker cash varies on a way greater scale than the value, so the value is usually shown as a semi flat line, making it difficult to appreciate its variations, I think it would be nice to have the two lines on separate scales, so that both can be appreciated visually.

What do you think?

@backtrader thanks for the link,

this:

• Issue a Market order
• Upon confirmation of execution you have the actual execution price and the actual market price (close)
• You can then issue two OCO orders which will bracket your existing positions. You calculate the stop-loss and take-profit prices from the execution price or from the market price

this was actually my initial implementation, I'm trying to move away from it because while it would work fine with real time trading, it introduces a bias in backtesting.
I explain: I understand that, in backtesting, events happen only at the end of full candles, it might happen that the bracket orders would have been executed in the same candle of the entry,
but obviously won't, because in backtrading they will be submitted only at the end of the candle.

maybe I could work it around (actually just diminish the bias) by backtesting with minute candles and resampling it to hour?

@backtrader said in problem with bracket orders:

I would even ask myself if a StopLimit order for the mainside is possible in a real broker, because that order type is in many occasions simulated.

fair point, stoplimit might be simulated, but for all intents I don't think it would make a massive difference in strategies that use it, as long as one knows what's going on.

Though I get the same issue if I change to use simple stop orders, rather than stoplimit:

self.long_trade = self.buy_bracket(
                price=entry_long,
                exectype=bt.Order.Stop,
                stopprice=stop_long,
                limitprice=limit_long
                )

            self.short_trade = self.sell_bracket(
                price=entry_short,
                exectype=bt.Order.Stop,
                stopprice=stop_short,
                limitprice=limit_short,
                oco=self.long_trade[0]
                )

It doesn't look like the issue comes from the execution type.

hello,

thanks for your reply,

I don't understand what you mean: do you reckon that it wouldn't be possible to place two orders like these with a broker?

basically this structure:

Hello,

I found something that I don't understand on how bracket orders work, I'll explain:

my strategy tries to catch runaways using the bollinger bands, so what I do is to place:

two stoplimit bracket orders: buy above current price, and a sell below current price,

These should open respectively a long and short position.
They are also OCO, so that if a long is open then the short will not, and vice versa.

This is my code to place such orders:

            self.long_trade = self.buy_bracket(
                price=entry_long,
                exectype=bt.Order.StopLimit,
                plimit=entry_long,
                stopprice=stop_long,
                limitprice=limit_long
                )

            self.short_trade = self.sell_bracket(
                price=entry_short,
                exectype=bt.Order.StopLimit,
                plimit=entry_short,
                stopprice=stop_short,
                limitprice=limit_short,
                oco=self.long_trade[0]
                )

With this I expect that both stoplimit orders get accepted, if price allows one of them to be opened (say the long/buy one), the other order (the short/sell and its respective bracked orders ) gets canceled, at the same time the two bracket orders (at stopprice and limitprice) become active.

With these orders my strategy can open as many long positions as it wants and everything works as expected. As soon as the price hits a short/sell order, something weird happens, here's my logs to show it:

ORDER 151,              main status:  Accepted -  Buy,  price: 1.04395
ORDER 152, (parent 151) stop status:  Accepted -  Sell, price: 1.03978
ORDER 153, (parent 151) limit status: Accepted -  Sell, price: 1.04645
ORDER 154,              main status:  Accepted -  Sell, price: 1.03895
ORDER 155, (parent 154) stop status:  Accepted -  Buy,  price: 1.04312
ORDER 156, (parent 154) limit status: Accepted -  Buy,  price: 1.03645
ORDER 154,              main status:  Completed - Sell, price: 1.03895 actual price: 1.03895
ORDER 151,              main status:  Canceled -  Buy,  price: 1.04395
ORDER 156, (parent 154) limit status: Canceled -  Buy,  price: 1.03645
ORDER 155, (parent 154) stop status:  Canceled -  Buy,  price: 1.04312

• orders [151, 152, 153] are respectively the main, stop, limit orders for the buy_bracket,
• orders [154, 155, 156] are respectively the main, stop, limit orders for the sell_bracket,
• order 154 is completed, so the main buy order is canceled as expected,
• orders [152, 153] should be canceled too but they are not.
• orders [155, 156] should NOT be canceled, but they are

this leaves my strategy with an open position (which prevents from creating further orders) that has no orders to close it, effectively hanging the whole strategy.

I tried reverting the order creation in the code, first creating the short_trade, and lastly creating the long_trade:

            self.short_trade = self.sell_bracket(
                price=self.sma - distance,
                exectype=bt.Order.StopLimit,
                plimit=self.sma - distance,
                stopprice=self.sma + stopdistance,
                limitprice=self.sma - limitdistance)

            self.long_trade = self.buy_bracket(
                price=self.sma + distance,
                exectype=bt.Order.StopLimit,
                plimit=self.sma + distance,
                stopprice=self.sma - stopdistance,
                limitprice=self.sma + limitdistance,
                oco=self.short_trade[0])

and exactly the reverse happens, I can open and close short trades no problems, but as the first long/buy trade is opened its bracket orders are canceled and the whole strategy stops working.

I'm not sure what I'm doing wrong here, any ideas?

thanks, now it makes more sense.

I'm a bit confused with the leverage,

I ran my strategy twice, first time setting leverage 400, second time leverage 800, here's how I set the leverage:

cerebro.broker.setcommission(commission=0.0, leverage=400)

then pprinted the bt.analyzers.TradeAnalyzer with:

     print('TradeAnalyzer :')
     stats[0].analyzers.TradeAnalyzer.pprint()

In both cases I get a bunch of stats about my analyzer, but what is strange is that the analysis I get is absolutely the same, numbers are an exact match for both leverages.

Also my broker cash and broker value are the same.

hello,

I'm not sure what I'm doing wrong, I'm implementing an indicator and loading data from a pandas dataframe,

here's how I load the data feed:

data = bt.feeds.PandasData(dataname=pandas, **fxcm.candles_to_backtrader(pandas))

    # Add the Data Feed to Cerebro
    cerebro.adddata(data)

    # Run over everything
    cerebro.run()

where fxcm.candles_to_backtrader(pandas) is defined as:

  def candles_to_backtrader(self, candles, ask=True):
        aob = "ask"
        if not ask:
            aob = "bid"

        return {"name": aob,
                "open": "%sopen" % aob,
                "high": "%shigh" % aob,
                "low": "%slow" % aob,
                "close": "%sclose" % aob,
                "volume": "tickqty"}

my indicator is:

import backtrader as bt


class Density(bt.Indicator):
    lines = ('my_indicator',)
    params = ()

    def next(self):
        print("new iteration")
        for i in self.datas[0].close:
            print("available data %f" % i)

next gets called many times, and everytime the for loop prints all the close values as they were produced by candles_to_backtrader method.

I'm quite confused, I expected that when backtesting I were to find in datas[0] only an incremental fraction of the input data at each call to next.

So first call I'd see the oldest values, the second call I'd see the oldest and the following ones, and get the full dataset only at the latest call.

What am I missing here?

thanks.

Re: Using backtrader data to make a Pandas dataframe from

I have a similar issue, I'm creating a new indicator and I find it would be really useful to have the datas[0] (or to be able to convert it) as a pandas dataframe.

There is a quantity of packages available in python to do advanced math using dataframes, additionally handling dataframes is so much easier and common practice that it just makes sense to have a as_dataframe method to get a pandas representation of it.