Look Ahead Bias question

[alexmulo]

Dear all,
I am testing several strategies using the backtesting library. Basically, if I would do a backtest using pandas I would shift the signal of 1 day to avoid the look ahead bias issue. Does the backtesting library do it from its own or should I introduce it like in this example?

class BollingerBandsStrategy(Strategy):
    n = 40  # Period for Bollinger Bands
    k = 2   # Number of standard deviations

    def init(self):
        # Compute Bollinger Bands using talib
        self.upper_band, self.middle_band, self.lower_band = self.I(
            tb.BBANDS, self.data['Adj Close'], timeperiod=self.n, nbdevup=self.k, nbdevdn=self.k
        )
    
    def next(self):
        if len(self.data) < self.n:
            return  # Not enough data to make a decision

        close = self.data['Adj Close'][-1]
        prev_close = self.data['Adj Close'][-2]
        upper = self.upper_band[-2]
        lower = self.lower_band[-2]

        if not self.position:
            # Buy signal: previous close crosses above the previous lower Bollinger Band
            if prev_close < lower and close > lower:
                self.buy()
        else:
            # Sell signal: previous close crosses below the previous upper Bollinger Band
            if prev_close > upper and close < upper:
                self.position.close()

Thanks

[alexmulo]

any suggestions?

[crueltytanya]

The algorithm will open (self.buy()) and close (self.position.close()) the position on the next candle - this eliminates looking ahead.
But... you need to be careful with the data you transmit ['Adj Close'].
If you need the value on the current candle, then the index is -1, and the deal will already be opened on the next candle.
To get a good understanding, try studying the backtesting charts to see if they match the logic of your strategy.

orange and blue line - indicators self.I(...