next function in Strategy not taking steps

[jjphung]

Hello all,

For some reason my Strategy class isn't iterating through the data using the next function.

    # ... 

    def init(self):
        self.c = self.data.Close.s
        h = self.data.High.s
        l = self.data.Low.s
        self.tenkansens = self.I(self.tenkansen, h, l).s
        self.kijunsens = self.I(self.kijunsen, h, l).s
        self.spanAs = self.I(self.spanA, self.tenkansens, self.kijunsens).s
        self.spanBs = self.I(self.spanB, h, l).s
        print(self.tenkansens)

    def next(self):
        print(self.tenkansens.iloc[-1])

The print statement in next just only prints out the same number over and over again, while the print statement in init does indeed show unique values.

I also checked all 4 of these series to see if there were at least 2 elements in there that weren't NaN, so I should see at least see one iteration in the next function.

Am I missing anything here? Let me know if you need anymore details.

[kernc]

You are assigning:

self.tenkansens = self.I(self.tenkansen, h, l).s     # type: pd.Series

that is, the indicator but transformed into pd.Series, which makes it invisible to Backtest for iterative lengthening.

If you instead do it like this, it should work:

def init(self):
    self.tenkansens = self.I(self.tenkansen, h, l)   # type: _Indicator

def next(self):
    print(self.tenkansens.s.iloc[-1])

[jjphung]

Awesome! Did not know it would be invisible. Thanks.

[kernc]

Yeah, we use this to recognize the "indicators" saved on self:

backtesting.py/backtesting/backtesting.py

Lines 1137 to 1140 in 73e1534

	# Indicators used in Strategy.next()
	indicator_attrs = {attr: indicator
	for attr, indicator in strategy.__dict__.items()
	if isinstance(indicator, _Indicator)}.items()

If you have a better idea, welcome!

[jjphung]

Sounds good :)

I do have a follow up question though:

    def init(self):
        self.c = self.data.Close

    def next(self):
        if self.tenkansens.s.iloc[-1] < self.c:
            # Sample signal triggered

For the code sample above, would this be the correct way of checking if the current close price is above the tenkansen's price? Or is there a better way? Thanks again!

[kernc]

I think it would be. Also this:

    def next(self):
        if self.data.Close > self.tankansens:
#        if self.data.Close[-1] > self.tankansens[-1]:  # Same, but relatively more performant
            ...

i.e. without saving self.c separately because self.data is always available, and without costly conversion to pd.Series just to throw it away.