Feat/order size

[kongyujian]

Added order sizing feature. Usage is as per discussion in #3

Opted to use absolute position size instead of a percentage based off holding cash.

Ran with the code on the example, with a tweak on sizing:

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

from backtesting.test import SMA, GOOG


class SmaCross(Strategy):
    n1 = 10
    n2 = 30

    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(size=1000)
        elif crossover(self.sma2, self.sma1):
            self.sell(size=1000)


bt = Backtest(GOOG, SmaCross, cash=10000, commission=.002)

output = bt.run()
bt.plot()

On a quick eyeball, it seems to work.

Did write one simple test. Let me know if there needs be further coverage.

[kernc]

That's a daring endeavor, or at least that was the impression I got the last time I was looking at it. Hope it pans out.

Position size likely is the best first step towards #10. The next would probably be order pyramiding so one can do:

buy 2
buy 3
buy 5
sell 10

Have an idea how to go about that?

[kernc]

Is this position computation correct? You effectively replace former cash with size, but size actually corresponds to cash / price?

I think when size is given, position might better be float(size) if enough_cash else error?

[kongyujian]

The intention for the size variable being computed as such is because:

1. If size is not specified when performing an order, it will default to the default behaviour of using entire holding cash amount.
2. I did not notice there was a margin option to define a leveraged trade. (Only found out about it after reading: How to use backtesting.py for forex trading #10 ) Was initially under the assumption that all trades were on a 1:1 leverage. Under my assumption, i believed that specifying the margin option wasn't needed in order to define a "leveraged trade". In fact, by specifying an order size that is larger than your holding cash amount would technically mean you enter a trade on cross margin. (which was why I did not implement a check for if enough_cash). Reason for this assumption is largely influenced by Trading View's pine script, and how "leveraged trades" would be entered under their backtesting tools.

So my thoughts are as follows, and would need some clarification:

1. Does the library actually calculate liquidations. If it does, the margin option would come in handy for that. A quick look into the code base shows how margin is used to calculate leverage amount. But margin and leverage are simply being used in reporting functions and likely no where else?
2. Would it make sense to have margin be calculated dynamically if order size was specified, and larger than than holding cash. In that case, we could dynamically calculate a cross margin on init instead.

Would love to take a stab at pyramiding too, there are indeed many use cases for that. Let me know your thoughts. I might be working on a path different than the philosophy of this library. :)

[kernc]

Does the library actually calculate liquidations.

It simulates them.

backtesting.py/backtesting/backtesting.py

Lines 561 to 570 in b3cfaff

	# Log account equity for the equity curve
	equity = self.equity
	self.log.equity[i] = equity
	# Hovever, if negative, set all to 0 and stop the simulation
	if equity < 0:
	self._close_position()
	self._cash = 0
	self.log.equity[i:] = 0
	raise _OutOfMoneyError

margin= therefore represents initial and maintenance margin of the trades. It's used only in stats and plots so the interested user gets a rough idea on how leveraged their strategy is allowed to be to still remain afloat.

[kernc]

In our case, position is actually equal to size when the user requests it. Thus, we only need to deduct the commission and then confirm we have (leveraged) cash to cover it all.

[kernc]

What do you think? 😅

[javamon1174]

An error message is displayed during this operation.

self.sell(tp=7200)

An error occurs when only a target price is specified for a sell order.

Conditional error in broker class

Please solve.

[drajtrade]

Does this change work properly?

[P-a-v-e]

@kernc I am just starting out with backtesting and it seems like the order sizing isn't in place. I am running backtesting version 0.1.7 however the changes above seem no to be commited.

[kernc]

@P-a-v-e PR #47 hasn't been merged yet. Welcome to try it out!

[zachrdlapetr]

@kernc when do you expect to release the 0.2.0 version with this feature?