MMFI Is A Key Measure of Stock Market BreadthI recently updated my piece defining oversold and overbought conditions using MMFI or the percentage of stocks trading above their 50-day moving averages (DMAs). Here is the key text from that post:
Above the 50, AT50, is overbought above 70%
Above the 50, AT50, is oversold below 20%
The Technicals of Converting from Above the 40 to Above the 50
Converting from Above the 40 to Above the 50 is relatively straightforward because the two measures are closely correlated. TradingView provides historical data back to January 2, 2002. Worden provides historical T2108 data back to 1987. The correlation over 8 1/2 years is 0.95. Over 20 years the correlation is 0.95. Thus, the relationship looks sufficiently consistent over time.
Above the 50, AT50 (MMFI) and Above the 40, AT40 (T2108) are highly correlated.
The correlation between AT50 and AT40 are highest at the extremes and worst in the middle. The correlation is better at the lower extreme than the higher extreme. Fortunately, the extremes of market breadth provide the most information for trading (overbought and oversold). The equation shown in the chart comes from the black diagonal trend line.
The linear approximation of the above scatter plot provides the new threshold values. AT50 overbought = ( – .3954) / 1.0007 = 70%. AT50 oversold = ( – .3954) / 1.0007 = 20%. So while wide variability exists in the relationship between AT50 and AT40 in the middle of the 0% to 100% range, the relationship works well at the extremes.
The Precision of the New Overbought and Oversold Thresholds
I examined the precision of the new overbought and oversold thresholds to add a layer of reassurance for this conversion.
Over the 20 years of data, the minimum value for AT50 for any AT40 overbought period is 55%. Accordingly, capturing any and all AT40 overbought periods requires triggering the overbought trading rules with AT50 above 55%. This conservative approach to eliminating false negatives for overbought provides 100% recall but poor precision. Using 100% recall would trigger overbought trading conditions 54% of the time based on the past 20 years of history. Something that happens the majority of the time is not an extreme! Thus, for trading extreme market conditions, precision is more important than recall.
The maximum possible value for AT50 for any AT40 oversold period was 42%. Even intuitively, traders can recognize that 42% is not low enough to define meaningful oversold conditions. Indeed, AT50 has traded below 42% for 25% of the trading days in the past 20 years.
Precision in Chart Form
The charts below provide a visual of these relationships. Each bar defines a “bin” which is a range of values. For example, think of the 70% bar as representative of all percentages starting with 70, like 70.1, 70.5, 70.9, 70.99, etc… The green bar marks the threshold for overbought in the first chart and oversold in the second chart. The yellow bars provide alternative thresholds based on the relatively high odds. The height of the bar represents the “odds” that AT40 is overbought (or oversold) given the value of AT50 on the x-axis. I calculated the odds as the percentage of time that AT40 is overbought (or oversold) given the value of AT50 over 20 years of data.
These charts show that the oversold trading period is more distinct than the overbought trading period relative to the AT40 definitions. Of course, if I started my work from AT50, this fuzziness would not be an issue. Regardless, traders should not treat the overbought and oversold thresholds as numbers fixed in stone. When the stock market approaches these thresholds, I use other data to identify when trading conditions are “close enough” to oversold or overbought. For oversold conditions, the volatility index (VIX) is useful. For overbought conditions, I look for signs of buying exhaustion as the stock market falls out of or away from overbought territory.
