Evaluating weekly trend-following and mean-reversion timing rules

One-variable rules as a starting point

One-variable timing rules are framed as teaching starting points that may fit trending or range-bound regimes but are unlikely to remain effective in every market condition.

All three procedures are long-horizon timing designs, typically holding for many months or years, and were built and evaluated on weekly closes of a broad equity index. The moving average is the shared quantitative baseline. Trend-following and mean-reversion are the two timing procedures compared on that same weekly tape.

The lagged trend-following design

The trend-following design uses a 40-week simple moving average, a three-week average of that average, a five-week high, and a 10-week low of the weekly close. The moving average is the trend baseline. The three-week average of that average is a smoothed confirmation line.

A trend-following entry requires the weekly close to exceed both the smoothed 40-week average and the five-week high. An exit requires the weekly close to fall below both the smoothed 40-week average and the 10-week low. The procedure waits for price to confirm a prevailing direction before entering or exiting, and it accepts delayed signals.

Trend-following action is specified for Friday of the signal week and requires computing in advance the close that would trigger the rule. That weekday and that same-week confirmation belong to the execution calendar.

The thresholded mean-reversion design

The mean-reversion design measures a five-week average of the gap between the weekly close and a 55-week simple moving average, then acts when that oscillator recrosses stated thresholds. That smoothed gap is the overextended oscillator. Large departures from the long moving average are treated as extreme.

Mean-reversion entries occur when the smoothed oscillator moves from below -15 to above -15 or from below zero to above zero. Exits occur when it falls below -15, crosses from above zero to below zero, or crosses from above +50 to below +50.

Mean-reversion action is specified for Friday of the week after the triggering oscillator values have already printed. The extra week is part of the execution calendar.

Compare both designs on the same weekly tape

Both designs are long-horizon timing procedures on the same weekly index closes. They differ in what they wait for. Trend-following waits for the close to confirm the smoothed moving-average line and a recent five-week high or 10-week low. Mean-reversion waits for the overextended oscillator to recross fixed thresholds.

They also differ in the execution calendar. Trend-following is specified for Friday of the signal week, after the triggering close has been computed in advance. Mean-reversion is specified for Friday of the following week, after the oscillator values have already printed.

What the historical evaluation recorded

The historical evaluation spans 22 years through mid-1994 and records between 10 and 22 complete buy-and-sell cycles, with time invested ranging from 56.3% to 68.8% of the sample. Editorial reading: those figures describe how often the closed procedures completed a cycle and how much of the sample they spent invested. They are not a claim about present-day results.