Spectral peaks are mode diagnostics, not forecasts

A peak that retunes a lookback

The archive used spectral analysis to decompose ordered price, volume, or breadth series into frequency bands so low-frequency trend energy could be separated from higher-frequency cycle energy. A dominant cycle was the estimate of the currently strongest periodic component in those observations. It was used to retune lookbacks instead of assuming a fixed calendar length.

A fixed fourteen-day oscillator length was treated as arbitrary. The proposed fix was to retune the indicator to the measured dominant cycle rather than leave it on a calendar constant. Changing cycle periods and shifting market character were cited as reasons to make filter rules adaptive rather than leave them static.

Why a conventional spectrum was set aside

Conventional Fourier transforms were judged unsuitable for market cycle measurement because the series do not remain stationary long enough to support a reliable estimate. Maximum-entropy spectrum analysis was adopted because it was already used on short scientific records, a constraint treated as analogous to limited market samples.

That short-record estimator concentrates frequency resolution when the sample is too brief or nonstationary for a conventional transform. Editorial view: adopting it does not cancel the nonstationarity that made a conventional transform unsuitable.

Two modes and the lag that remains

Market behavior was split into a low-frequency trend mode and a higher-frequency cycle mode, with separate filters assigned to each rather than one all-purpose smoother. Trend mode is directional price behavior handled with heavier smoothing and therefore more lag. Cycle mode is sideways, oscillating behavior whose swings sit at higher frequencies and punish confirmation delay.

Smoothing and lag were described as inseparable. Frequency components can be reshaped and rejection sharpened, but lag cannot be designed away. That is the lag-smoothing tradeoff: any smoother delays its output. Rejection can be sharpened and lag reduced, but lag cannot be removed.

In sideways cycle mode, confirmation lag was treated as too costly, so a phase reconstruction of the measured cycle was used instead of waiting for a conventional oscillator signal. Editorial view: that substitution was a response to confirmation lag inside cycle mode. It does not tell you when trend energy has already invalidated the oscillator.

A robustness screen is still not a forecast

Before a cycle-based rule set was called robust, a historical sample on the order of thirty trades per free parameter was required, rather than a short recent window. Editorial view: that requirement tests whether the rule set was given enough historical cases. It does not turn a spectral peak into a forecast, and it does not say the oscillator is still valid once trend mode has taken over.