Test oscillator events before tuning rules

Ask about later prices first

Oscillator-style tools can be described as first-order high-pass filters that remove longer price swings, while moving averages can be described as smoothers that remove high-frequency variation.

A system-design sequence can begin by asking whether a coded, identifiable event has a statistical association with later prices, before a full entry-exit procedure is written. An identifiable event is any condition that can be written as true or false on a historical bar and then scored by the subsequent percent change in price.

Mark a stochastic crossing as an event

A stochastic oscillator is a unit-scaled placement of the close between the highest and lowest closes over a lookback, used here to mark a coded event when the reading crosses below a low threshold. Built that way from the close relative to the highest and lowest closes over the lookback, it can mark a long-side event at that downward cross.

Score the later close-to-close change

Lookback-shifted scoring records the event on a past bar and measures the percent close-to-close change from that bar to the present so the subsequent path is already observed. One evaluation design records the event on a bar 10 bars in the past and measures the percent close-to-close change from that bar to the current bar.

Subsequent-return bins are a fixed grid of clipped percent changes used to accumulate an empirical distribution of outcomes after events. In that procedure, the subsequent percent change is clipped to a minus-10 to plus-10 percent band, rescaled onto a 0 to 100 axis, and counted into 100 bins to form an empirical outcome distribution.

A center-of-gravity calculation on those bin counts is used as a compact location summary of typical subsequent percent change. Center of gravity here means a first-moment location of the binned outcome counts, recentered so a balanced outline sits near a no-change reading.

Trim the procedure only after the event is scored

After an event is treated as informative, the same oscillator can be rewritten as one procedure whose lookback, threshold, holding period, and percent loss limit are adjustable. System optimization is that later step: adjusting lookback, threshold, holding period, and a percent loss limit only after an event has already been scored by the distribution of subsequent prices.

The same event-testing bed can be reused across symbols and at different bar intervals, including intraday or equal-tick sampling. Robustness testing is repeating the same event-scoring procedure on other symbols and other bar intervals to see whether the event remains informative.