@dasch @backtrader

Thanks for your help. The PEBKAC is resolved :-)

I guess it is obvious to you, but perhaps a default mode of "backtest bars should match live bars" would follow the principle of least surprise? In any case, here are some live resampled bars, followed by the corresponding historical resampled bars with a timeframe of 5 minutes with rightedge=False. This shows that live bars match historical bars. Now the question I have to answer myself is: "why did righedge=False not work in my system when I tried it live?"

Live

Data0, 0196,                    736793.5625, 2018-04-09T09:30:00.000000, 1.27725, 1.27794, 1.27709, 1.27773, -1.0, 0, 1.277366
Data0, 0197,              736793.5659722222, 2018-04-09T09:35:00.000000, 1.27773, 1.27882, 1.27745, 1.27874, -1.0, 0, 1.277632
Data0, 0198,              736793.5694444445, 2018-04-09T09:40:00.000000, 1.27874, 1.27876, 1.27756, 1.27769, -1.0, 0, 1.27775
Data0, 0199,              736793.5729166666, 2018-04-09T09:45:00.000000, 1.27769, 1.27791, 1.27729, 1.27751, -1.0, 0, 1.277784
Data0, 0200,              736793.5763888889, 2018-04-09T09:50:00.000000, 1.27751, 1.27751, 1.2765, 1.27675, -1.0, 0, 1.277684
Data0, 0201,              736793.5798611111, 2018-04-09T09:55:00.000000, 1.27675, 1.27726, 1.27634, 1.27681, -1.0, 0, 1.2775
Data0, 0202,              736793.5833333334, 2018-04-09T10:00:00.000000, 1.27681, 1.27716, 1.27641, 1.27712, -1.0, 0, 1.277176
Data0, 0203,              736793.5868055555, 2018-04-09T10:05:00.000000, 1.27712, 1.27748, 1.27683, 1.27736, -1.0, 0, 1.27711
Data0, 0204,              736793.5902777778, 2018-04-09T10:10:00.000000, 1.27736, 1.27757, 1.27725, 1.27736, -1.0, 0, 1.2770800000000002
Data0, 0205,                   736793.59375, 2018-04-09T10:15:00.000000, 1.27736, 1.2774, 1.27652, 1.27652, -1.0, 0, 1.277034

Historical

Data0, 0196,                    736793.5625, 2018-04-09T09:30:00.000000, 1.27725, 1.27794, 1.27709, 1.27773, -1.0, 0, 1.277366
Data0, 0197,              736793.5659722222, 2018-04-09T09:35:00.000000, 1.27773, 1.27882, 1.27745, 1.27874, -1.0, 0, 1.277632
Data0, 0198,              736793.5694444445, 2018-04-09T09:40:00.000000, 1.27874, 1.27876, 1.27756, 1.27769, -1.0, 0, 1.27775
Data0, 0199,              736793.5729166666, 2018-04-09T09:45:00.000000, 1.27769, 1.27791, 1.27729, 1.27751, -1.0, 0, 1.277784
Data0, 0200,              736793.5763888889, 2018-04-09T09:50:00.000000, 1.27751, 1.27751, 1.2765, 1.27675, -1.0, 0, 1.277684
Data0, 0201,              736793.5798611111, 2018-04-09T09:55:00.000000, 1.27675, 1.27726, 1.27634, 1.27681, -1.0, 0, 1.2775
Data0, 0202,              736793.5833333334, 2018-04-09T10:00:00.000000, 1.27681, 1.27716, 1.27641, 1.27712, -1.0, 0, 1.277176
Data0, 0203,              736793.5868055555, 2018-04-09T10:05:00.000000, 1.27712, 1.27748, 1.27683, 1.27736, -1.0, 0, 1.27711
Data0, 0204,              736793.5902777778, 2018-04-09T10:10:00.000000, 1.27736, 1.27757, 1.27725, 1.27736, -1.0, 0, 1.2770800000000002
Data0, 0205,                   736793.59375, 2018-04-09T10:15:00.000000, 1.27736, 1.2774, 1.27652, 1.27652, -1.0, 0, 1.277034

@blonc

Try:

def__init__(self):
   self.d0 = self.params.d0
   self.d1 = self.params.d1

You'd do something like this:

class MyStrategy(...):
  def next(self):
    now = self.data0.datetime.datetime(0)
    if now.time() >= datetime.time(11,31) and now.time() <= datetime.time(12,59):
       ... your logic here ...

I've had this problem as well, the only reliable solution I've found is to restart the backtrader process and manually manage any open orders. Kinda sucks to have to do this.

You'd do something like this:

class MyStrategy(...):
  def next(self):
    now = self.data0.datetime.datetime(0)
    if now.time() >= datetime.time(11,31) and now.time() <= datetime.time(12,59):
       ... your logic here ...

@tw00000 said in Trailing take profit:

It does seem in times of volatility the StopTrailLimit would be very disadvantageous - are there any advantages?

It can be problematic. I've had markets go right over my limit. The only thing I do is have a OCO stop market order further than the trail in case markets go crazy and I'm not there to do something.

Here is how I debug live trades:

1. Use notify_trade and notify_order functions and print out the status as it comes in
2. Don't use GTC orders unless your strategy cancels orders. I use time-limited orders

Additionally, when I place an order, I assign the current order to a member: self._order = ... and unset it when notify_trade marks it as closed. Then, I don't do any of the strategy logic unless self._order is None.

Looks like I should use a trading calendar

See summary. Is there any way to set the specific day of the week used to mean the beginning/end of the week? Currently it appears to default to Sunday/Monday which is OK but I'd like more control.

@backtrader Actually, I'm unable to see how to apply this.

If we have

self.buy_bracket(price=100,...)

But the actual price is $.75, the order will go to the broker at $100... I'm very confused. Maybe the right thing to do is to manually create the bracket orders.

@backtrader oh that's very interesting and makes perfect sense, thank you!

This is tough, I think because the "size" has to be different if you want to convert 100% of the amount back to your base currency.

So say my base currency is USD and I am trading USD/GBP (I know this isn't a real pair). If I convert $100 to £75, the order could look like this:

self.buy_bracket(price=.75,limitprice=.78,stopprice=0.74)

The sizer is called once, and lets say the idea is to always trade a size of $100, so we configured bt.sizers.FixedSizer,stake=100

Now say the profit limit is hit, that means we will convert £100 into $133. But we only had £75 which means we will be taking some sort of a position instead of closing it out.

I'm trying to figure out how we would size this properly, but it seems to be a limitation in my brain.

I tried to resolve this issue by walking recursively down the object tree rooted at cerebro and setting all lock types to None. Didn't work, still found a lock somewhere.