I try to answer myself. It seems first one is trade status, possible values are ['Created', 'Open', 'Closed']. Latter one is order status, possible values are ['Created', 'Submitted', 'Accepted', 'Partial', 'Completed', 'Canceled', 'Expired', 'Margin', 'Rejected'].

Both pnl and pnlcomm are 0 for newly opened trade in notify_trade. However, should pnlcomm be 0, since it's net profit, and should include commission?

In a test, TradeHistory's status.status shows value of 1, but event.order.status has value of 4.
What's the difference of the two status value?

Hope the question is not too vague.

In strategy, we have notify_trade to be called when a trade is happened (opening/updating/closing). Based on the reference of this trade, what is the recommended way to get the underlying orders for this trade?

That's why a multi-asset version of quickstart would be very helpful for starters to clarify things, as it would touch upon things that single-asset wouldn't.

So you mean backtrader's multiple asset assume all assets have the same length (therefore, no survivorship bias present). It doesn't support the multiple assets with different length.

Indeed, I'm still following through quick start and exploring from there. Will read document links at further steps.

Though prenext is probably a good idea to put in the start, as it's natural concept if multi datafeed is considered. Or even better, it'll be great to have a multi-asset version of quickstart by itself.

@backtrader said in Multiple asset with missing data:

This statement is simply wrong.

From the above example, first 'next' call for A is on 2018-12-05, even though A has data 2018-11-30. From method calls perspective, A's 'next' calls gets truncated.

I don't mean by actual data, actual data can't be truncated, because it still gets called in 'prenext'.

Initially I thought 'next' for stock A should get called with all the days, but instead it gets truncated by the lesser stock B.

@backtrader said in Multiple asset with missing data:

No.

Thanks, seems working for me now. Turns out prenext gets called, and next doesn't. That means the logic should probably be put in both prenext and next.

For information, I tried
stock A: 2018-11-30 to 2018-12-10.
stock B: 2018-12-05 to 2018-12-10.

The result is:

prenext. A 2018-11-30
prenext. B 2018-11-30
prenext. A 2018-12-03
prenext. B 2018-12-03
prenext. A 2018-12-04
prenext. B 2018-12-04
next. A 2018-12-05
next. B 2018-12-05
next. A 2018-12-06
next. B 2018-12-06
next. A 2018-12-07
next. B 2018-12-07
next. A 2018-12-10
next. B 2018-12-10

The strategy code is:

class LogStrategy(bt.Strategy):

    def prenext(self):
        for data in self.datas:
            curr_date = self.datas[0].datetime.date().strftime('%Y-%m-%d')
            logger.info(f'prenext. {data.AssetId[0]:.0f}, {curr_date}')

    def next(self):
        for data in self.datas:
            curr_date = self.datas[0].datetime.date().strftime('%Y-%m-%d')
            logger.info(f'next. {data.AssetId[0]:.0f}, {curr_date}')

I'm trying to load multiple assets, however only last day's data is coming out. I assume when there're missing data in one asset on a day, the system will skip that day for all assets. It's quite likely on a particular day that some stock is not trading, so collectively a large number of days are being skipped.

Would it make sense to change this behavior? I guess if the asset doesn't trade on one day, it's fine to skip for that asset, but other assets with data should still run.