Posts made by backtrader

@xyshell said in cerebro.resample() introduces data in the future:

Note: 6219.15 is the close price of the 1h data at 2020-03-20 01:00:00, while 6219.15 is also the open price of the 1min data

No. There is no such thing as the closing price of the 1-hour data, because that data doesn't exist. Naming things properly does help.

Additionally the information you provide is wrong, which is confirmed by looking at your data.

Note for the other readers: for whatever the reason, this data defies all established standards and has the following format: CHLOVTimestamp

Your data indicates that

• 6219.15 is the closing price of the 1-min bar at 00:59:00, hence the last bar to go into the resampling for 1-hour between 00:00:00 and 00:59:00 (60 bars -if all are present-) which is first delivered to you as a resampled bar at 01:00:00
• 6219.15 is the opening price of the 1-min bar at 01:00:00

At 01:00:00 you have two bits of information available:

• The current 1-min bar for 01:00:00
• The 1-hour resampled data for the period 00:00:00 to 00:59:00

As expected.

@benjomeyer said in Timers:

        timername='selltimer',

Given how you pass it ... you could probably try this better

def notify_timer(self, timer, when, timername):
    if timername=='buytimer':
        self.buy_stocks()
    elif timername=='selltimer':
        self.sell_stocks()

given that you have no additional *args or **kwargs. But a more general approach I would suggest this

def notify_timer(self, timer, when, **kwargs):
    timername = kwargs.get('timername', None)
    if timername=='buytimer':
        self.buy_stocks()
    elif timername=='selltimer':
        self.sell_stocks()

where you can also decide if a default logic for None is needed.

Note: notice the usage of elif

@Kevin-Fu said in Building Sentiment Indicator class: TypeError: must be real number, not LineBuffer:

    def next(self):
        self.date = self.data.datetime
        date = bt.num2date(self.date[0]).date()
        prev_sentiment = self.sentiment
        if date in date_sentiment:
            self.sentiment = date_sentiment[date]
        self.lines.sentiment[0] = self.sentiment

In any case that's probably where the error happens. There is no definition of self.sentiment and the indicator understands you are looking for the line named sentiment. The problems

• Lack of variable initialization
• Conflicting naming

@Kevin-Fu said in Building Sentiment Indicator class: TypeError: must be real number, not LineBuffer:

Any input would be greatly appreciated

But you provide no input with regards to the error. Only

TypeError: must be real number, not LineBuffer

Python exceptions provided a stacktrace which points to the different line numbers in the stack ... allowing to trace the error to the origin point (and showing any potential intermediate conflict)

The error is only telling us that you are passing an object (a LineBuffer) there where a float should have been passed by you. But not where you are doing it ... which is in the stacktrace ...

@Kevin-Galkov said in DataCls does autoregister:

If I run the mementum/backtrader/samples/oandatest/oandatest.py , it complains:

In any, it does complay because you don't have the proper package installed, you have something for v20.

You have to use this:

• https://github.com/ftomassetti/backtrader-oandav20

The built-in module is only compatible with the old Oanda API.

And oandatest.py works with the old built-in module.

Timers go to notify_timer which receives the timer, when it is happening and any extra *args and **kwargs you may created the timer with. You can use any of the latter to differentiate timers.

Or you can simply use the timer id to associate different logic to each timer.

Docs - Timers

• https://pypi.org/project/bta-lib/
• https://btalib.backtrader.com/
• https://github.com/mementum/bta-lib

Covering all non-candlestick indicators offered by ta-lib implemented in Python and being able to work as a drop-in replacement of ta-lib if compatibility is activated.

The fundamental problem is that a position open by a bracket should not have additional logic to close the position independently.

But if you insist in shooting yourself in the foot, the only thing you have to do is to keep a reference to any of the bracketing orders (stop-loss side or take-profit side) and cancel it whenever you close the position with the bracket-foreign logic.

if logic_orders_to_close_the_position:
   self.cancel(take_profit_order)

To do that, look at the definition of how lines are plotted in the BuySell observer.

https://github.com/mementum/backtrader/blob/700380c8bc41ab9da8295cfc7b64091bab75866b/backtrader/observers/buysell.py#L47-L54

If the data cannot be downloaded, there will an attempt to open a cached version from a file with the same name as the ticker, which obviously does not exist. This is probably a Yahoo problem.

The sample code is incomplete but the presence of

@Benjamin-Kesby said in Optstrategy and PivotPoint indicator pickle error:

if __name__ == "__main__":
    pass

does clearly point in the direction indicated by @vladisld and the scoping problems that pickle has. The solution would be the same as the one outlined in the referenced post.

The indicators are tightly coupled to support things like managing different timeframes, remain synchronized to the data, play well with replay and resample.

Using them outside of a strategy is not possible. You can follow the path recommended by @lampalork or wait a bit until things get settled with bta-lib: https://btalib.backtrader.com

@xyshell said in Crossover indicator doesn't plot as expected:

    self.adx_crossover = bt.ind.CrossOver(
        self.adx_fast, self.adx_slow, plot=True, subplot=True)

Indicators are plotted subordinated to its master. The crossover happens on two subindicators of an indicator, so the data ceased to be the master a couple of levels before the current point.

Use plotmaster=self.data

See: Docs - Plotting - https://www.backtrader.com/docu/plotting/plotting/

@xyshell said in Crossover indicator doesn't plot as expected:

but adx_crossover indicator doesn't show up as expected.

Understanding how something doesn't show up as expected without knowing what's expected and what's actually happening is a challenging problem. A real conundrum.

@xyshell said in Crossover indicator doesn't plot as expected:

What's going on here?

A picture would probably let us experience what's actually going on.

The resampler filter wouldn't know what to do with extra lines (add, max, min, sub, multiply, take the 3rd one, take the last ...) and works only with the fields for which it has a rule, the standard ones.

You can create a 2nd data feed, assign your desired fields to open (keep 1st), high/low (keep max/min), close (keep last), volume (add) and use that resampled to match the timeframe of your master data feed.

Or create your own resampling filter.

@Mark-Weaver said in strategy.getposition returning different values for same ticker on different timeframes:

I expected

You want to do "algotrading", remove then "expect" from your vocabulary. As soon as possible.

@Mark-Weaver said in strategy.getposition returning different values for same ticker on different timeframes:

if I have a DataFeed for FCEL on a 5m aggregation, and I send a buy order through the broker API, what is it actually buying, that's different than, say, the underlying instrument behind the FCEL datafeed on a 1M aggregration?

You know nothing is different. But the code treats each single data feed as an independent data feed. Hence accounting is different.

Why should the code be making efforts to try to identify things as being the same thing? That will fail in an epic manner when the data feed provider provides no identification means.

Your AI is still believing that there are many other AI entities in place.

@Mark-Weaver said in strategy.getposition returning different values for same ticker on different timeframes:

Is that how it would work in live trading?

I don't know, I didn't write the alpaca broker, you will have to ask alpaca, but I guess it will be so.

@Mark-Weaver said in strategy.getposition returning different values for same ticker on different timeframes:

I don't see it as a parameter in the buy function

The buy method has a clear parameter data which is the target of your acquisition. Track anything you want on 5min and buy the 1min feed always (and check its position)

@Mark-Weaver said in strategy.getposition returning different values for same ticker on different timeframes:

So why would the broker interface return different position values for FCEL 5M vs FCEL 1M?

Because they are different data feeds.

@exu said in Order modification feature request:

To simulate that in BT, one would have to cancel the OCO orders and issue new ones with the recent limit price levels

Which effectively is what happens in the market. You make a change and your order goes to the back of the queue (as if it had been cancelled and then re-issued. If brokers didn't do that, they would be giving you an unfair advantage)

@exu said in Order modification feature request:

can you do that with only daily data0?

The timeframe is not relevant and is not looked into.

Since your other compression is 60*24 and 24 > 20, the answer is 0 (and depending on the stage of the resampling even partially 1, which is actually the future)

Your (A)I is far more advanced that the rest of the AIs in algotrading (non-existent AIs in fact) which makes you think that things are understood as you understand them.

@Mark-Weaver said in strategy.getposition returning different values for same ticker on different timeframes:

the posSize value is retrieved via

for i, d in enumerate(self.datas):
            posSize = self.getposition(d).size

Even if you know that they are the same "ticker" because you have personally added them, the code cannot know it. They are different data feeds, full stop. If they are the same or not is something you have to know and manage.

And for each data feed a separate position accounting is kept, hence your result.