Stop loss on the same day the position was opened

[DAtek]

Expected Behavior

Strategy closes the position at 720

Actual Behavior

Strategy closes the position on next candle at open price, 705.58

Steps to Reproduce

Run the following code. You can see the wrong behaviour on 2012-10-18

from backtesting import Backtest, Strategy
from backtesting.test import GOOG


class DummyStrategy(Strategy):
    def init(self):
        pass

    def next(self):
        if f"{self.data.index[-1]}" == "2012-10-17 00:00:00":
            self.buy(sl=720)


if __name__ == "__main__":
    data = GOOG.tail(100)
    backtest = Backtest(data, DummyStrategy, cash=10000)
    result = backtest.run()
    backtest.plot()

Additional info

• Backtesting version: 0.2.0

[kernc]

Strategy.next() is assumed to be called between bars. Strategy.buy() just places the buying order, but orders are executed and turned into trades, as mentioned in the quick start manual, the soonest on the next available bar.

Documentation improvements welcome.

[kernc]

Nevermind, I misunderstood.

I guess the reason SL isn't hit on the same bar is that we iterate over a copy of orders state:

backtesting.py/backtesting/backtesting.py

Line 774 in d146b43

for order in list(self.orders): # type: Order

so once we turn an order into a trade:

backtesting.py/backtesting/backtesting.py

Lines 881 to 883 in d146b43

	# Open a new trade
	if need_size:
	self._open_trade(adjusted_price, need_size, order.sl, order.tp, time_index)

the subsequent SL order that was placed in the queue isn't picked up in the same broker iteration.

[kernc]

But imagine an order on that day placed with self.buy(stop=758, sl=720), i.e. the order is only entered once the stop price is hit (which is above open price ($755)). This then makes it ambiguous whether the stop price was hit before the SL price (in which case we enter and shortly after exit a losing trade) or after (in which case we just enter the trade and leave it open, along with its SL order, waiting for another bar). We can draw assumptions from the chart in this particular case, but not the general case — we don't have intra-candle data.

Due to this ambiguity, and the resolving complexity, I'm kinda leaning to wontfix.

Have a better idea?

[DAtek]

I don't think the strategy should allow stop orders for exactly that reason. I would limit the entry point of a position to open and close, and the users would have still enough freedom to implement their logic to decide whether the strategy should open a position in the next iteration or not. In the previous versions it worked perfectly.

[kernc]

In the previous versions it worked perfectly.

Glad you think so. 😊

The following commit should restore the previous behavior for plain market orders. Thanks.

[DAtek]

Thank you for the quick fix!!! 😊

[rokups]

Actually i think we can reasonably improve same candle entries and exits as we can reliably deduce some intra-candle movements.

Consider this image. We always know that wicks guarantee movement up and down. We also know candle body has at least one direction of price movement, in direction of the candle.

Given all that, we can always do this:

• If a candle touches stop-loss price - always fill that in, ignoring candle movements.
• If trade entry and take-profit prices are within wicks - close trade in-profit.
• If trade entry and take-profit are within same candle - close trade in-profit only when candle matches trade direction (green candle for longs, red candle for shorts).
• If neither of these conditions are satisfied, then current behavior should be maintained and take-profit order should be filled on a next eligible candle. This includes:
  • Long trade where limit entry is below open price of red candle and take-profit is in upper wick, because price may go up from the open, reach TP and only then fill an order and proceed down.
  • Short trade where limit entry is above open of green candle and take-profit is in lower wick, because price may go down from the open, reach TP and only then fill an order and proceed up.

I am new to the codebase so im still not sure how to correctly wrangle this in. Still, thought ill post it to get some feedback, while i am familiarizing myself with the code.

[zeshansari]

guys i am having trouble with following code, i set buying order with a condition so it should take when close price hit the given level and it does but take buy position at the wrong level which is not given anywhere but one thing is there that it open order before my tp and close position at earliest with profit of 0.5 points only. Although as per my understanding it should open position at close price when the giving condition is true i.e. 3903 but it is open position at 3911 and closing it 3911.5. can anyone help how can i resolve this problem?

def trade_level(self):
if self.data.close <= 3903 +1 and self.data.close >= 3903 - 1:
self.buy(sl= 3903 - 2, tp= 3903 + 9)