@asuralm said in How to correctly trade on data1 and data2?:

Am I using the bt.observers.BuySell._stclock=True correctly?

There is nothing right or wrong about it. It was a hack.

@asuralm said in How to correctly trade on data1 and data2?:

Yes they are indeed different data feeds but it seems the plotting is not correct when multi data feed the testing engineer no matter what.

If each example you present has different feeds, different logic (summary: everything is different) there is no way to discern where a/the problem may be.

Hi Sergio,
You can trade live using the same framework on IB, Oanda or visual chart. See:

https://www.backtrader.com/docu/live/live.html

Note that you will likely need to do some tweaks to account for the differences between live and backtesting but nothing major. Also note not everything that you can do in backtesting may be available for the live broker. It depends on their API.

Cheers

Affecting only replaying. This commit should alleviate the situation

https://github.com/mementum/backtrader/commit/26dbab587fdc360c6fbd0617d21da88e2cc9c941

The code is usually run against an old 1.5.0rc3. The situation that arises with a downgrade is that many other packages should also be downgraded to achieve the actual status that version had at that time. This was seen with pyfolio, which has changed the API. Version 0.5.1 had a compatible API, but that version had also expectations about the surrounding ecosystem ... hence the problem of downgrading.

It's not implemented. You might be adding the call to a fork of yours, until it may be looked into.

buy_bracket itself doesn't look at anything. It simply generates orders and delegates to the rest of the platform. But if the size doesn't make sense (float) things will probably break somewhere.

There is no specific provision to forbid passing a float, but there is also no effort made to check that it isn't passed.

Hi @backtrader,

Is it possible to get bid/ask lines in an Oanda datasource, similarly to what you've described in your "Escape from OHLC land" post? That would also solve the spread problem because that way calculating the spread would be super easy.
As I can see there is an option ("bidask") that controls what is retrieved from Oanda, so I believe it is already partially implemented.

Would it be possible to add the bid/ask lines to the data if the timeframe is set to bt.TimeFrame.Ticks?

Cheers,
Tamás

Although not shown in your code snippet, the error trace indicates that a moving average (directly or over another indicator which makes use of it) is in play.

Apparently the minimum data, as you already realized, was not enough for the batch calculation.

In this case running with runonce=False would have also not generated the error (each calculation would have been on a step by step basis)

See:

Thread title: How to get the (live) broker's API response data for bracket order?

This doesn't work:

valid=self.datas[0].datetime.datetime() + timedelta( hours=num_hours_traversed)

not working => the order doesn't register at all

They are for sure connected in your code, tests and line of thought, but they provide no information here becase there is no context to bind them together.

OptReturn is a simplified return instance which mimics a strategy, to reduce the overhead of message passing from worker processes to the main process.

See: Docs - Optimization Improvements

Plotting has never been tested with optimization (not an expected use case)

https://github.com/ContinuumIO/anaconda-issues/issues/905

@backtrader Thank you so much!

Good. It might in any case be helpful for others (and even for a small correction/improvement) if you share the details in case there is a workaround/solution for an issue. But the important thing is always to get it working.

Unfortunately not.

I guess there's a lot to be said for thoroughly reading the manual :)

A little patch:

if self.p.apikey is not None: urlargs.append('api_key=%s' % self.p.apikey)

There was a typo previously.

The logic in the strategy has to remember when a position was entered, probably in notify_order by checking the date from the notified order

order.executed.dt which can be translated to a datetime instance like this for example: order.data.num2date(order.executed.dt)`

And then checking when the data hits a month later self.data.datetime, where you can get instances with self.data.datetime.datetime() for example.

Thanks for the clarification.

@backtrader

or ( ). As I said, I'm learning the platform right now, so I'm thankful for the input.

@backtrader - got it, thank you.