Noise-matched rules still need trend filters and robustness tests

Three choices tested as one procedure

The archive describes a historical workflow in which a trading system is not finished when a single rule is written. Short-horizon price movement is characterized as news-driven and erratic price noise, so directional persistence is treated as a longer-horizon property.

TradersWeek editorial: the design is unfinished until three choices are tested as one procedure. Those choices are which noise class the market belongs to, whether a longer trend filter allows or blocks the short-horizon rule, and whether the same signal survives neighboring calculation periods.

Assign the rule family with an efficiency ratio

A 10-day efficiency ratio, net change divided by the sum of absolute daily changes, is used to assign trendier series to trend rules and noisier series to short-horizon fade rules. That ratio is net displacement over the sum of absolute one-period moves, used to rank a series as trendier or noisier.

Strategy family is then matched to measured noise. Noisier index markets, and sometimes gold, are assigned fade rules, the short-horizon mean reversion used where prices reverse often. Quieter interest-rate markets are assigned breakout-and-hold trend following, a directional rule that stays with a measured price trend over the system holding period.

Lengthen the calculation period instead of adding complexity

As markets become more active or more volatile, the stated adaptation for a long-horizon trend rule is to lengthen the calculation period rather than add rule complexity.

An opening-range breakout is adapted to noisier sessions by scaling the first entry near a 0.1 fraction of 10-day volatility and the profit objective near a 0.7 fraction of the same measure.

Let a longer trend filter accept or block the short rule

A trend filter is a longer-horizon directional overlay that accepts, blocks, or times a shorter-horizon entry. Short-term pattern rules, including a three-day cycle, are described as more usable when that overlay screens trades and when the chosen market reverses frequently.

Price-versus-momentum divergence is framed as technical divergence: a short-lived split between price direction and a momentum oscillator inside an existing trend. The archive workflow treats that split as a brief stall inside a larger price trend, to be taken in the trend’s direction for a few days rather than as a reversal forecast.

Keep the signal only if nearby windows still agree

Robustness testing checks that entry, exit, and abstention stay consistent across nearby lookbacks, markets, or noise classes. A daily divergence setup is treated as robust only if it appears in at least half of 11 momentum windows from 5 to 15 days, and is closed when it remains in only two windows.

For trend following, robustness is described as allocating across several calculation periods instead of choosing one, because many long-horizon windows can look acceptable after the fact.

A test of standing aside before scheduled reports and re-entering afterward is described as usually forfeiting a move that continued the existing trend.