3. Problem with automargin / margin and expected PnL
For code/output blocks: Use ``` (aka backtick or grave accent) in a single line before and after the block. See: http://commonmark.org/help/

Problem with automargin / margin and expected PnL


  • Hi everyone,

    I found something very strange while using the automargin (margin=True) in commissionInfo during a backtest, that I cannot understand. When I sell, profits and loss are less than expected. The opposite happen when I buy, p&l are more than expected.

    To replicate this as simple as possible, I used a dataframe of size 29 * 3 (date, open, close, with open = close, starting from 1 to 29), startcash=1000. I buy (or sell) the first day and keep until the end.

    • with CommissionInfo(commission = 0, margin = True, mult = 10, automargin = 1)
      cerebro.broker.getvalue() returns 757 (sell) / 1297 (buy)

    • with CommissionInfo(commission = 0, margin = 1, mult = 10)
      cerebro.broker.getvalue() returns 730 (sell) / 1270 (buy)

    • if I set automargin=1, mult =10, I observe10% difference between automargin & margin (long +10%, short -10%)

    • automargin=2, mult=10, it's 20%

    • ... (automargin/mult) * 100 %

    Maybe I missed something, maybe I don't really know how works commissions in backtrader, but I would like to know why both previous results are different.

    Thank you in advance !

    from __future__ import (absolute_import, division, print_function, unicode_literals)
import backtrader as bt
import pandas as pd
import datetime 

class TestStrat(bt.Strategy):
    
    params = (('position', 0),)
        
    def next(self):
        if self.p.position == 0:
            self.sell(self.data)
#           self.buy(self.data)
            self.p.position = 1
            
cerebro = bt.Cerebro()

startcash = 1000.0
cerebro.broker.setcash(startcash)

df = pd.DataFrame(data = {
            'date': [datetime.datetime(2000,1,i) for i in range(1,30)],
            'open': [i for i in range(1,30)],
            'close': [i for i in range(1,30)]
        })

df = df.set_index('date').sort_index()
data = bt.feeds.PandasData(dataname = df)
cerebro.adddata( data )

cerebro.broker.addcommissioninfo(bt.CommissionInfo(commission = 0, margin = True, mult = 10, automargin=1))

# cerebro.broker.addcommissioninfo(bt.CommissionInfo(commission = 0, margin = 1, mult = 10)) 


cerebro.addstrategy(TestStrat)
cerebro.addobserver(bt.observers.Broker)

strats = cerebro.run()

print(cerebro.broker.getvalue())
    0

  • :'(

    0

  • Hi,

    It seems to be a bug.

    See this Pull Request for a fix:
    https://github.com/backtrader/backtrader/pull/387

    @backtrader could you check please?

    0
 
});