Returns Volatility Indicator

[Mariano Volpedo]

Hey, dear community !

Building volatility targeting strategies, I found that this relevant indicator is missing.

You could find more about this here:
https://quantpedia.com/an-introduction-to-volatility-targeting/#:~:text=Volatility target is the level,volatility is easier to predict.

Hope this helps. Comments?

PS: GARCH implementation for the next version but literature says it's expensive and you get similar results using faster EWMA

class StdDevReturns(bt.Indicator):

    lines = ('stddev',)
    params = dict(
                    method="simple", # or ewma_pd or ewma_np
                    annualized = True,
                    period=20,
                    alpha=0.06, # only for ewma
                    adjust=True, # only for ewma_pd
                    trading_year_days=252,
                )
    
    def __init__(self):
        self.addminperiod(self.params.period + 2)
        if self.p.annualized:
            self.annualized = self.p.trading_year_days ** (1 / 2)
        else:
            self.annualized = 1
        
    def next(self):
        returns = self.Returns(self.data0.get(size=self.p.period))
        
        if self.p.method == "simple":
            self.lines.stddev[0] = returns.std() * self.annualized
        elif self.p.method == "ewma_pd":
            data = pd.DataFrame(returns)
            self.lines.stddev[0] = data.ewm(alpha=self.p.alpha, adjust= self.p.adjust).std().iloc[-1][0] * self.annualized
        elif self.p.method == "ewma_np":            
            self.lines.stddev[0] = self.CalculateEWMAVol(returns, 1 - self.p.alpha) * self.annualized

    def Returns(self, prices):
        return np.diff(np.log(prices)) #np.diff(prices) / prices[1:]
    
    def CalculateEWMAVol(self, returns, _lambda):   
        period = len(returns)
        average_returns = returns.mean()
    
        e = np.arange(period-1, -1, -1)
        r = np.repeat(_lambda, period)
        vector_lambda = np.power(r, e)
    
        sxxewm = (np.power(returns - average_returns, 2) * vector_lambda).sum()
        Vart = sxxewm/vector_lambda.sum()
        EWMAVol = math.sqrt(Vart)
    
        return EWMAVol