Posts made by run-out

@albertoburchi There is no admin or access to admin for this forum except by the creator of Backtrader and he is not active at the moment. Just do what you can. Good luck.

@fanpeix Perhaps PctRank will get you close to the mark? If gives you the percent position but with bucket sizes based on the perdiod, I think.

@zugba Thank you for your code. I copied and ran your code using both your data and a sample of backtrader data. I used both the standard csv loader and your custom class.

All ran smoothly with backtrader data. The only error was when I use your data with your custom class. This suggest an error in your data. Try using a different data source.

Here's the data I used. 2006-volume-day-001.txt

@eff You are using get somewhere but I cannot see where as you only included limited code.

get looks backward to get an array of size n. You must ensure there is a minimum period so that you can get n numbers. Use addminperiod

@protrasol Yes we have had some messages about this. This forum is controlled by the creator of Backtrader, @backtrader Daniel Rodriguez. If he sees this message, hopefully he will correct the issues or allow someone else access to help out.

@protrasol I use pycharm as well and can see everything inside the objects. You will like it!

@protrasol I think my best recommendation for you would be to use your debugging tool. Just put a stop right after cerebro.run() is returned to strat, and you can explore everything that is inside the strat object. From there you can plan how to extract data.

Which IDE are you using? Pycharm and VSCode are free and have great debuggers.

Coding without this technique is like driving with a blindfold on. You are constantly trying to figure out what's happening.

@sj I regret I've hardly ever used the timers. Perhaps someone else can help clear this up?

@hanaan Yes. You need to use get in next. It returns an array based on the parameters size and ago.

@hanaan said in Error with simple custom indicator:

self.emaVariation = vol.PercentVariationInPeriod(self.ema_slow, plotname='EMA Variation', subplot=True)

There a two built in indicators that make this a bit easier for you. Highest aka MaxN and Lowest aka MinN.

You can use them in your indicator like this:

class PercentVariationInPeriod(bt.Indicator):
    lines = ("Variation",)

    params = (("period", 180),)

    def __init__(self):
        high = bt.ind.MaxN(period=self.p.period)
        low = bt.ind.MinN(period=self.p.period)
        self.lines.Variation = (high - low) / low

Looking at your code I couldn't see anything wrong with it. So I copied and ran it using just tsla downloaded data from yahoo and it seems to run fine. Maybe try another data source?

Starting Portfolio Value: 1000000.00
2014-01-02, Close, 30.02
2014-01-02, BUY CREATE, 30.02
2014-01-03, BUY EXECUTED, 30.00
2014-01-03, Close, 29.91
2014-01-06, Close, 29.40
2014-01-07, Close, 29.87
2014-01-08, Close, 30.26
2014-01-09, Close, 29.51
2014-01-10, Close, 29.14
2014-01-10, SELL CREATE, 29.14
2014-01-13, SELL EXECUTED, 29.16
2014-01-13, Close, 27.87
2014-01-13, BUY CREATE, 27.87
2014-01-14, BUY EXECUTED, 28.10
2014-01-14, Close, 32.25

@sj From the docs:

Trying cheat-on-open

The sample below has a strategy with 2 different behaviors:

    If cheat-on-open is True, it will only operate from next_open

    If cheat-on-open is False, it will only operate from next

In both cases the matching price must be the same

    If not cheating, the order is issued at the end of the previous day and will be matched with the next incoming price which is the open price

    If cheating, the order is issued on the same day it is executed. Because the order is issued before the broker has evaluated orders, it will also be matched with the next incoming price, the open price.

    This second scenario, allows calculation of exact stakes for all-in strategies, because one can directly access the current open price.

@yashagw It's very hard to help you with the limited question you ask. Please include your code and a sample of your data if you can. Thanks.

@gstrader Correct. You need to have one dataframe for each symbol, add it to cerebro, and then repeat until all your desired symbols are entered and ready for the backtest.

@gstrader said in datafeeds with cross-section OHCL data:

Is there to derive a feed from existing structure to support loading data from multiple dataframes,

Yes. Please search google, the community and the docs. Use these keywords:

backtrader, pandas, feeds, multi, multiple,

@protrasol I couldn't find an example to share. But I use dictionaries when the complexity goes up. Say for example if you are trading bracket orders for many stocks. In this type of scenario you might want a dictionary with the data as key, then a dict for each data with perhaps each order type as keys.

self.tracking_dict = dict(
    data=dict(
        enter_order=`the_entry_order`, 
        stop_order=`the_stop_order`, 
        limit_order=`the_limit_order`,
    )

This method allows you to track other information associated with the bracket order if you wish. It provides an ability to track info related to the bracket and then act on that info during the backtest.

@protrasol Your answer lies in the docs. Have a look at the bracket orders sample code. Follow the use of orefs. In the sample code orefs is a list you create and manage to track the orders. You can use this list for a wide variety of purposes including printing out any current order status. Make sure you follow all uses throughout the code since you must manage the orders contained in orefs.

You can also use a dictionary if you like using keys to label the orders.

@ctrado said in Need help to run minimal simple example:

Also, when and how do I know if have to feed self.data to the indicator? Sometime you have it and sometimes not.

like in

bt.ind.SMA(period = self.params.slow_length)

you have no data....

Sometimes I see

data.close or data[0] ....

very weird.

@ctrado Please read the docs here for some explanation of how indicators are constructed.

@bmccoy said in Keep getting error module 'backtrader' has no attribute 'Timeframe':

Timeframe

Try TimeFrame

@run-out And also the fact that no shares are sold.