Posts made by backtrader

@momentum said in indicator with dynamically changing data:

Do you mean instead of changing the data of the indicator, change the data feed itself?

No, I said that you create a data feed which dynamically changes the data it serves.

You fail to understand that the self.datas array in each object is a different one. This is obvious if you consider that you can do this

ma_on_ma = bt.ind.SMA(bt.ind.SMA(self.data, period=5), period=5)

The outer SMA has, obviously, only 1 data feed which happens to be another SMA (the inner one)

Futhermore, you expect things to use the array self.datas and you manipulate it with that expectation. As noted above, this is not supported, given that other notations and references are used in the internals.

@zblp said in Problem with plotting buy sell arrows:

Buy sell arrows don't seem in the correct place

They are

@zblp said in Problem with plotting buy sell arrows:

There is some kind of scaling issue

No, there isn't. Your data doesn't have enough price difference from the highest high to the lowest low for the default configuration. See:

• Docs - Observer Reference

Disable the standard observers in cerebro using stdstats and add the observers you may wish with the configuration of your choice. See:

• Docs - Cerebro
• Docs - Observer and Statistics

My very sincere advice: if you don't know how to keep a reference in a class to a value, you need to really work on the very basics of programming. And work a lot.

It may seem very tempting to get a straight answer and go and develop a winning strategy: you won't develop any winning strategy, because so far you cannot program (which doesn't mean you cannot learn and be very good at it) and your code will be like a minefield.

But if you wish you can google it ... or even go to Stackoverflow where someone had the same problem (which isn't really a problem)

• https://stackoverflow.com/questions/43164468/anyone-uses-backtrader-how-can-i-keep-a-reference-in-a-method

Choose your own path.

@momentum said in Multiple data feeds - Strategy starting later:

I thought minimum period is something used for indicators. Is there some option I am missing?

Yes. The minimum period applies to anything to all objects with lines, because the buffers (i.e.: when you access [0]) have to produce something. That's guaranteed in next.

If some data feeds start later, they wouldn't be able to produce data when the earlier data feeds do it, hence breaking the contract which you have in next.

@alain said in plot size question:

hight = 30

Things are in the matplotlib arena, but that could have something to do with it.

The messages from the API are completely different, same as the keys. Your best bet is to check with the creators of the v20 broker module.

Thank you ... you overtook me!

@d-virant said in Oanda data feed:

but I assume that installing OandaPy avoids the issue of V20 not being supported?

How?

A user wrote a wrapper for the v2: https://github.com/ftomassetti/backtrader-oandav20

@sylvain said in Question on PyFolio:

Is there any way I could format the trading result and feed to Pyfolio and make it works?

Your use case is not covered by the PyFolioAnalyzer which is the analyzer wrapping the calls to PyFolio.

You would have to use the existing code as basis for a new analyzer which decides which data feeds to consider.

@baktra said in Strategy Optimizer Result Print Order:

there is something I am missing.

What do you think you are missing?

@baktra said in Strategy Optimizer Result Print Order:

I would appreciate if you help

Exactly with what?

When you get an order notified in notify_order you can assign them to a dict (probably appending it to a list) when the order is still alive (method of the order) and removing it when not alive.

As key you can directly use the data attribute from the order or its, for example _name which is set with the name you give to cerebro.adddata. See

• Docs - Order
• Docs - Cerebro

To cancel them, just go over the list for an entry an issue cancel in the strategy.

• See Docs - Strategy

@jackg101 said in Adding datafeeds mid-simulation for OTC trading:

is there a way to add datafeeds on-the-fly as the strategy decides to trade something?

No it doesn't. This is something which has been even discussed privately with option traders, but it's not something in the core of backtrader

@tizenhet said in Position size changes without buy or sell order:

                        self.buy(data=d, size=int_)
                        self.order = self.buy()

@angles922 said in backtesting with manual trades:

I am trying to add a stoploss and takeprofit onto my manually imported order history.

You cannot. Because the order is executed right at once and has no binding to other orders (the brackets).

Set the brackets (as OCO orders) when you are notified in notify_order. See Docs - OCO Orders

@knolch said in Plot doesn't show up when running the code:

I will not use Spyder now for developing and testing my trading strategies but do you have any recommendations what I could use instead?
I mean IDLE works but it's very uncomfortable. Which Editor are you using for backtesting?

Apparently people like PyCharm for which a Free Community edition is available.

• https://www.jetbrains.com/pycharm/

Unless you are comfortable with vim or emacs and don't mind using a standard shell (which for all intent and purposes happens to be bash these days)

@knolch said in Plot doesn't show up when running the code:

[[<Figure size 640x480 with 5 Axes>]]

That's the result of this in the code

cerebro.plot()

Because plot does actually return the created matplotlib figure, should the user desire to do something with it. If you were assigning the return value to a variable, you wouldn't see that.

The problem is Spyder. The configuration you have is preventing the chart from being displayed. Your options:

• Try

  cerebro.plot(iplot=False)
  

  This disables automatic inline plotting detection in backtrader, but this is probably not going to help.

• Have a look at the charting options in Spyder for plotting. If set to automatic, change it to inline and viceversa.

It's the problem with this Python kernel hijacking shells: they think they know better.

@s1000rr said in Generic CSV data feed, indicator problems on plot:

1: Not reading first ~11 lines of input file. I used a small sample of the data file 2006-day-001.txt from github

Minimum period. The lines are all being read. Read:

• Docs - Operating the platform
• Docs - Strategy

@s1000rr said in Generic CSV data feed, indicator problems on plot:

2: Buy and Sell indicators are shifted above/below the stock price

Read the documentation for the BuySell observer. It is an observer which simply indicates when and what happened. Detailed values are given to you in order execution.

• Docs - Observer Reference

@s1000rr said in Generic CSV data feed, indicator problems on plot:

3: Inconsistent time separation
Using the same plot from above we can see the varying distance that separates the vertical date lines.

You probably want to elaborate on what you mean here.

In any case if you are expecting a charting platform, backtrader isn't. The charts are meant only as a quick visual aid for inspection and are not the core of the platform. It's just an add-on built on top of matplotlib

The code must be a joke. A bad one, obviously.

Some of your posts:

• https://community.backtrader.com/post/7664

  You want to do things in notify_trade and yet you present code now which is like a random collection of characters. How have you managed to buy/sell anything?

• https://community.backtrader.com/post/7632

  You backtest twice over the same period (with an extended period once) and show random results for the common leading period, which is impossible. But how can you backtest when your code is random?

Your turn.

You could also simulate the gap entering with a simple Stop and link both the StopLimit and Stop to make them OCO orders, which guarantees that only one of them gets executed. See

• Docs - OCO Orders

The indicators use a generic notation to refer to the data with which they operate and don't care about indexing the array.

In any case I feel you are using the wrong approach imho. Instead of changing the data feeds, you should change what the data feed delivers, which at the end of the day delivers the same result, but is compatible with any way an indicator can be written.