no handler when the backtesting finnished?

[guzuomuse]

say: i entry a lot of positions but without any closed. i want to close all when backtesting finished. so i need a handler

but i can't find it. how?

so i think we need a "stop" method like "backtrader" to handle some logic for example close positions or do some customized statistic etc...
this "stop" handler should be called after backtesting finished but before "built-in" statistices calculated so that we can append our
customized statistic items
what 's your idea? @kernc

[kernc]

Backtesting.py has a decided simple API: only init() and next() to care about. With it, you could close all trades with something like:

class Example(Strategy):
    def init(self):
        ...
        self.data_length = len(self.data)

    def next(self):
        if len(self.data) == self.data_length:
            ...  # end of backtest

The reason it currently won't work is bug #158. (Expecting a fix shortly. You can do a -1 if you need it immediately.)

so that we can append our customized statistic items

What do you mean? Have an example? There's a pretty extendable trade data API you can use to compute extra stats ...

[guzuomuse]

woo...minus 1 . i indeed coded like that before but could not get it work .but now i got it

so that we can append our customized statistic items

just for custom statstic demand. for example we want to calculate the "mae" "mfe". but as you say , we can calculate with the trade data api.
thank you.

[guzuomuse]

if len(self.data) == self.data_length - 1:
     self.position.close()

give the warning info:
c:\........pandas\core\series.py:726: RuntimeWarning: invalid value encountered in log

the is because result = getattr(ufunc, method)(*inputs, **kwargs) the inputs have negtive value (the ufunc is log and the method is __call__)

[kernc]

What's your backtesting.__version__? Something like this had been fixed a while ago, and the only place we call log() ensures the values are positive:

backtesting.py/backtesting/backtesting.py

Lines 1395 to 1396 in 028f02d

	def geometric_mean(x):
	return np.exp(np.log(1 + x).sum() / (len(x) or np.nan)) - 1

[guzuomuse]

version:0.2.4
easy to reproduce
in the "get start" code just replace the
bt = Backtest(GOOG, SmaCross, commission=.002,exclusive_orders=True)
with
bt = Backtest(GOOG, SmaCross,commission=.002, exclusive_orders=False, hedging=True, cash=100000000)
the warning info will appear

[kernc]

This doesn't raise any warnings for me:

from backtesting import Backtest
from backtesting.test import GOOG
from backtesting.test._test import SmaCross
bt = Backtest(GOOG, SmaCross,commission=.002, exclusive_orders=False, hedging=True, cash=100000000)
bt.run()

Put a breakpoint in c:\...\pandas\core\series.py, line 726. Typing where in the subsequent debugger will give the full stack trace, so you'll be able to see how the process got there.

[guzuomuse]

your code above has no problem. bellow is "mine"(copied from the "get start exmaple") full version code,please check it out.

from backtesting import Backtest, Strategy
from backtesting.lib import crossover

from backtesting.test import SMA, GOOG


class SmaCross(Strategy):
    n1 = 10
    n2 = 20

    def init(self):
        self.sma1 = self.I(SMA, self.data.Close, self.n1)
        self.sma2 = self.I(SMA, self.data.Close, self.n2)

    def next(self):
        if crossover(self.sma1, self.sma2):
            self.buy()
        elif crossover(self.sma2, self.sma1):
            self.sell()


# bt = Backtest(GOOG, SmaCross, cash=10000,
#               commission=.002, exclusive_orders=True)
bt = Backtest(GOOG, SmaCross,
              commission=.002, exclusive_orders=False, hedging=True, cash=100000000)
output = bt.run()
# bt.plot()