MMI: Market Meanness Index

Mariano Volpedo

Hi,

To develop momentum strategies, its relevant to understand is the trend is strong or a mean reversion is coming. Tipically, Hurst Exponent claims to work, but unfortunally I dont see value on this indicator with my data (latinamerican mutual funds).

Finally I found the MMI as an alternative with better results to implement in my strategies. You can get the explanation and several backtestings on https://financial-hacker.com/the-market-meanness-index/ (ps: you can read there with 55 is important)

This is my implementation on backtrader, pls your feedback is welcome

class MMI(bt.Indicator):
    lines = ('mmi',)
    plotinfo = dict(plothlines=[55])
    params = dict(period=100,)

    def __init__(self):
        self.m = Average(self.data.close, period = self.p.period) #replace with median
        
    def next(self):
            nl = 0
            nh = 0

            for j in range(self.p.period-1):
                if j < 1:
                    continue
                i=j * -1
                if (self.data.close[i] > self.m[0]) and (self.data.close[i] > self.data.close[i-1]) :
                    nl += 1
                if (self.data.close[i] < self.m[0]) and (self.data.close[i] < self.data.close[i-1]) :
                    nh += 1
            self.l.mmi[0] = 100.0 * (nl+nh)/(self.p.period - 1)

barton05

@Mariano-Volpedo

Great strategy! Tried to replace with moving average and more on __init__ instead of next.
The results swift to 60-63 instead of ~55. Not sure is it because of any calculation error....

    def __init__(self):
        self.l.m = bt.ind.MovAv.Simple(self.data.close, period=self.p.period) #replace with median
        self.l.nl=bt.And(self.data.close>self.l.m,self.data.close>self.data.close(-1))
        self.l.nh=bt.And(self.data.close<self.l.m,self.data.close<self.data.close(-1))
        self.l.upSum = bt.ind.SumN(self.l.nl, period=self.p.period)
        self.l.downSum = bt.ind.SumN(self.l.nh, period=self.p.period)
        self.l.mmi= 100.0 * (self.l.upSum+self.l.downSum)/(self.p.period - 1)