Relative-strength-index rules under a 20-percent trend filter

One recipe and two later switches

A historical test applied a long-only relative-strength-index rule to 1816 randomly chosen New York and Nasdaq-listed stocks from 1 January 2005 through 31 August 2014. The baseline rule bought when the oscillator first crossed below an oversold-threshold of 30 and sold on the first later cross above an overbought-threshold of 70. Both sides were booked at the signal-day-close. Universe-level results were summarized with equal weight on every trade, independent of the underlying share price.

Editorial framing: the archive then changed only a greater-than-20-percent trend-filter and the oversold-threshold, so the same long-only recipe can be read as a regime-dependent filter rather than as a single 30/70 switch.

How the oscillator compresses the relative-strength-ratio

The relative-strength-index is a 0-to-100 oscillator built from the ratio of average gains to average losses over a stated lookback. The relative-strength-ratio is the unbounded average-gain divided by average-loss quantity that the oscillator compresses into that fixed range. In a smooth advance the relative-strength-ratio can keep rising after the oscillator has already flattened near 70.

Oversold prints during a long decline

On the illustrated 2007 decline, the oscillator crossed below 30 more than 10 times. It repeatedly printed oversold readings while price stayed in the downtrend.

Allowing the rule only after a 20-percent rise

A second procedure allowed the same oscillator rule only after a rise of more than 20 percent confirmed an uptrend. The trend-filter blocked entries in declines of more than 20 percent and liquidated open trades at the close when that uptrend reversed.

That state rule is the trend-following-gate: the oscillator procedure runs only inside a confirmed advance and open trades are liquidated when the advance reverses.

Raising the oversold-threshold to 50

After the uptrend gate was applied, crosses below 30 became scarce because the oscillator often hovered near 50 in advances. A third procedure raised the oversold-threshold from 30 to 50 and left the remaining rules unchanged, including the overbought-threshold of 70 and the signal-day-close convention.

Trend-duration and the summarized result

Sorting the same 1816 names by typical trend-duration, then contrasting the 500 shortest-trend names with the 500 longest-trend names, showed a negative association between trend-duration and the oscillator rule's summarized result.