Sometimes it’s wise to revisit stock-in-trade technical indicators. RSI is a prime candidate. It is seemingly omnipresent and has been massaged in virtually every way imaginable. Its lookback period has been shortened and lengthened, its results have been smoothed (often many times over), for closing prices tinkerers have substituted highs and lows and weighted closes, and other indicators have been overlaid on it. But before we get too fancy we should understand how the basic indicator works.
John Hayden published RSI: The Complete Guide (Traders Press) in 2004. No guide, of course, is ever complete, but Hayden’s book is a good start. In dissecting the RSI indicator Hayden uses the Morris modified calculation. It is not cumulative like Wilder’s original version and hence does not have the element of exponential smoothing. It simply measures the ratio of the average increase in price over N days to the average decrease in price over N days and then normalizes that ratio so that it is bound on the upside by 100 and on the downside by 0.
The Morris modified RSI behaves logarithmically. For instance, if both the up average and the down average equal 1, RSI is 50. If the up average is 2 and the down average is 1, RSI is 66.67; if the up average is 1 and the down average is 2, it is 33.33. If the up average is 3 and the down average is 1, RSI is 75; if the averages are reversed, RSI is 25. If the up average is 4 and the down average is 1, RSI is 80; reversing the averages we get an RSI of 20. From these numbers we see that “the largest increase or decrease in the RSI value occurs when the ratio changes from 1:1 to the next whole number (2:1 or 1:2)” and that “the RSI value experiences its largest changes in value as it oscillates between the index values of 40 and 60. In other words the RSI is most sensitive to price change when the RSI is oscillating between 40 and 60,” (p. 11) or, more accurately, between 33.33 and 66.67. When Hayden is touting Fibonacci levels he uses the latter pair of numbers, when he’s using a quick and dirty system he goes with the former. For instance, he claims that “in an uptrend, the RSI finds resistance at 80 and support at 40” while “in a downtrend, the RSI finds resistance at 60 and support at 20.” (p. 63) A few pages later he describes resistance in a downtrend as “the 66.7 level or more generally the 60 level.” (p. 67)
Hayden is a harsh critic of those who use divergence as a trend reversal indicator. By the very nature of its mathematical calculation RSI will offer up many bullish divergences in downtrends and bearish divergences in uptrends. What do they signify? Hayden contends, in opposition to much of the literature, that “a bullish divergence signifies that the existing trend is down and the Bears are exhausted. We should be expecting a rally to sell into. . . . Inversely, when we see a bearish divergence, the trend is up and we should probably expect a retracement to lower prices because the Bulls are exhausted. It is time that we should be looking for a reason to buy.” (p. 67)
(to be continued)