Custom Commission Scheme that take a timeseries for Interest

lampalork

Dear All,

Has anyone implemented implemented a custom Commission Scheme that takes a timeseries/feed as input instead of a flat fixed rate? If someone is kee to share; I'd like to take a look.

Many thanks in advance
Lamp'

Re: Access get_credit_interest in next()

run-out

Could you provide an example please? Thanks.

lampalork

Hi,

Usually, brokers will charge interests when taking SHORT position on stocks. Backtrader allows you to define a fixed interest % for this using out of the box commission scheme (https://www.backtrader.com/docu/commission-credit/). However; I'd like to run backrests over a long period of time (various interest rate regime) and therefore I would like to be able to model this interest rate as Fed Fund rate + 50bp for instance. I can easily create a data feed for FF+50 bp. What I would like to know is if someone has already created a custom Commission Scheme that allows variable interest rate over time.

Ideally, we would pass a datafeed to the Commission Scheme constructor (instead of fixed rate) and the method _get_credit_interest() would fetch the right prevailing interest rate.

Hope this is clear enough.

Kind regards,
Lamp'

run-out

OK, that helps a lot.

First thing is you need to add a signal to your data line, or independently as a separate line. I'm going to assume you know that part. If not, let us know. Let's assume you add a data line as FF to your data feed.

Now you need to create a class for the commission. The _get_credit_interest method has the data line in it, so hopefully you aren't using the method. We can call the method internally and get return the data line. From there we isolate your FF line and get the current value.

class CommissionFF(bt.CommInfoBase):
    """Custom commission for both trading and brokerage charges."""

    params = (
        ("commtype", bt.CommissionInfo.COMM_PERC),
        ("commission", 0.0004),
    )

    def _getcommission(self, size, price, pseudoexec):
        """Calculates the commission of an operation at a given price """
        
        return abs(size) * self.p.commission * data.FF[0]

    def _get_credit_interest(self, data, size, price, days, dt0, dt1):

        return data

Of course you need to add all this in to cerebro. I've added in the kwargs as an example only.

 comm_kwargs = dict(
        commission=float(.0004),
    )

    comminfo = CommissionFF(**comm_kwargs)

    cerebro.broker.addcommissioninfo(comminfo)

lampalork

@run-out said in Custom Commission Scheme that take a timeseries for Interest:

e a class for the commission. The _get_credit_interest method has the data line in it, so hopefully you aren't using the method. We can call the method internally and get return the data line. From there we isolate your FF line and get the current value.

I'll give it a try; I'll keep you posted. thanks a lot

run-out

I think I got lost in the question a bit. I just re-read the question, you can simply access the data line in the _get_credit_interest() method and modify your rate. Sadly I can't adjust my previous response.

class CommissionFF(bt.CommInfoBase):
    """Custom commission """

    params = (
        ("commtype", bt.CommissionInfo.COMM_PERC),
        ("commission", 0.0004),
    )

    def _get_credit_interest(self, data, size, price, days, dt0, dt1):

        int_rate = data.FF[0] + 0.5

       return    calculations etc...

lampalork

Hi @run-out

Thanks a lot for your answer. I have started looking at this and unfortunately; I'm hitting a very strange issue.
For now, I just want to create a custom Commission Scheme that applies interest on Short AND Long positions. The little twist is that I want to apply different interest rate for Short and for Long side. let say for argument sake 1.5% for Long position, 4.5% for Short side.

I feel what I'm doing is fairly simple and should work but for some reasons, I get an error that says "TypeError: getvaluesize() missing 1 required positional argument: 'price'"

I initially thought that my backtrader version was too out-dated; so I have updated it to 1.9.75.123).

I looked at the code on github but I really don't understand why I get this error. Any help/insights would be greatly
appreciated! Note that everything works blissfully if I remove this Commission Scheme.

Kind regards
Lamp'

class CommissionFF(bt.CommInfoBase):
    params = (
        ("commtype", bt.CommissionInfo.COMM_PERC),
        ("commission", 0.0),
    )

    def _get_credit_interest(self, data, size, price, days, dt0, dt1):
        # days * abs(size) * price * (interest / 365)
        if size > 0: # Long
            return days * abs(size) * price * (0.015 / 265)
        else: # Short
            return days * abs(size) * price * (0.045 / 265)

----
# Usage of the comminfo:

# data from TS
data0 = get_data_feed('tradestation', os.path.join(data_folder, 'TLT.csv'), 'TLT', None)

# Cerebro
cerebro = bt.Cerebro()
cerebro.adddata(data0)

# Broker
start_capital = 10000
myBroker = bt.brokers.BackBroker()
comminfo = CommissionFF
myBroker.addcommissioninfo(comminfo)
myBroker.set_cash(start_capital)
cerebro.broker = myBroker

lampalork

ok, my bad, very stupid issue. I'll now experiment with the use of a data line. I'll keep you posted.
Sorry for the mistake.

this line should be changed:

comminfo = CommissionFF
# must be changed to 
comminfo = CommissionFF()

regards,
Lamp'

lampalork

Finally, managed to spend some time on this, Here is the final code for the benefit of others

class CommissionFF(CommInfoBase):
    params = (
        ('commtype', CommInfoBase.COMM_FIXED),
        ('commission', 0.0),
        ('commtype', None),
        ('interest_long', True),
        ('spread_long', 0.0050), # FF + 50bp
        ('spread_short', -0.0050) # FF - 50bp
    )

    def _get_credit_interest(self, data, size, price, days, dt0, dt1):
        # (days / 365) * abs(size) * price * (interest + spread)
        # long: positive commission (i.e. debit)
        # short: negative commission (i.e. credit)
        if size > 0: # Long
            return (days / 365) * size * price * max(data.fed_fund[0] + self.params.spread_long, 0)
        else: # Short
            return (days / 365) * size * price * max(data.fed_fund[0] + self.params.spread_short, 0)

André Tavares

@lampalork why don't you use interest for short interest and interest_long for long interest both from class CommInfoBase.

lampalork

@André-Tavares said in Custom Commission Scheme that take a timeseries for Interest:

interest_long

to my knowledge interest_long is a boolean flag not a float

Wayne Filkins 0

Does this have anything to do with the "short interest" indicator? I'm trying to figure out how to create it in backtrader but I don't know if what this post is about is the same thing.