A movement ruler with a clock

If a distance is measured as the same fraction of daily ATR, does it represent the same amount of coming movement at every hour?

This study treated ATR as a ruler. The job of the ruler was to make near-term range comparable across instruments, entry hours, and months. It did not try to predict which direction price would move.

For each hourly entry, we divided the range over the following holding window by an estimate available at entry. We compared the middle half of the resulting log ratios. A narrower spread meant the unit made those future ranges more consistent.

Candidates included different daily estimators, day boundaries, lookbacks, and shorter chart ATRs. The time-aware version combined a completed-day level with the usual range for the entry hour and holding horizon, then adjusted for recent activity relative to its own usual pace.

The historical construction used only completed information. A causality check changed later bars and verified that earlier estimates stayed unchanged. We compared instruments individually, split the period into two halves, and examined scheduled release windows and shock days separately.

Changing the estimator or lookback made relatively small differences. Accounting for the hour made a larger difference, and recent activity improved the comparison further.

At a four-hour hold, the median instrument's log-ratio spread fell from about 0.65 with broker daily ATR to about 0.51 with the time-aware unit. The latter was narrower on all 35 instruments in both sample halves, and the ordering also held at the one- and eight-hour holds tested.

Barrier-touch frequencies became more consistent across hours too. That is useful evidence that equal distances in the new unit represented more comparable amounts of near-term movement.

Scheduled US release windows and shock days still produced more movement than the estimate anticipated. A recent-range adjustment reacts after new movement becomes observable; it cannot know an unobserved surprise in advance.

The report's simulated constant-volatility random-walk comparison is a model reference. It is not a universal lower bound that no real process or estimator could beat.

The sample covers about twenty months. Wider testing would help establish how much of the gain persists across brokers, calendar conventions, regimes, and new periods. The two illustrated entries were selected to show the hour effect; the aggregate comparisons carry more evidential weight than those examples.

The result concerns consistency of range measurement. It does not establish a profitable stop, target, position size, or higher win rate. A range unit and a unit for adverse movement against a particular judgment also need not be the same thing.