MEASUREMENT LOG

What has been measured,
and what hasn’t

Every claim on this site has a state. Refuted counts the same as supported.

How a claim moves

Registered: 4
Collecting: 4
Under evaluation: 0
Closed: 4

Closed

ClosedRefutedFrom a tool2026-10-16 registered

The setup marker's default 1.5R target is profitable at its own hit rate.

How it is checked

Count every marked setup on the loaded history of one instrument across three timeframes. For each, record whether price reached the target or the stop first. When a bar touches both, count the stop. Compare the resulting expectancy against zero.

Sample

122 / 122

Result

122 marks, 35 reached the target, 86 the stop — a 29 percent hit rate, identical across all three timeframes. At a 1.5R target the expectancy is minus 0.275R per trade before costs. Raising the minimum to 2.5R produced 4-5 setups per chart across the entire loaded history. The tool is not published as a signal generator.

The lab note

ClosedSupportedFrom a tool2026-10-12 registered

Taking the decision on the bar after the touch, rather than on the touch itself, improves the outcome.

How it is checked

Hold every other rule fixed. Run the same setup logic twice over the same history: once deciding on the touch bar, once on the bar after it. Compare the totals.

Sample

20 months of MNQ history

Result

Over twenty months of MNQ history the same logic went from minus 1,380 to plus 396 points on this single change. The touch bar is still mid-fight; the next one shows who won it.

ClosedRefutedFrom a tool2026-10-10 registered

The candle-shape delta estimate is a usable substitute for aggressor data.

How it is checked

Run the candle-shape estimate and a real trade-by-trade aggressor delta on the same instrument over the same window. Compare the sign and the magnitude on matched bars.

Sample

a parallel session on one instrument

Result

On one matched window the estimate read plus 12,110 while the real aggressor delta was minus 20. Not merely imprecise — opposite in sign. The estimate is weakest exactly where it would matter most: at absorption, where heavy selling into a level that does not fall prints as buying. Every tool using it now says so on its own page.

ClosedSupportedFrom the terminal2026-10-05 registered

On a given day, more trades correlate with a worse result.

How it is checked

For each trading day, record the number of trades taken and the day's net result. Compute the correlation across the sample. State the expectation before looking.

Sample

the logged trading days

Result

Correlation minus 0.49 between the daily trade count and the day's result. This became a filter: on a day that keeps reversing around its own average, every extra trade makes it worse. What it does not say is which direction the causation runs.

Collecting

CollectingFrom the terminal2026-10-08 registered

Levels hold in a ranging market and fail on a trend day.

How it is checked

Classify each day as ranging or trending by a measure fixed in advance. For each level touch, record whether price reacted or went through. Compare the two populations against a control of all touches.

Collection paused; resumes 2026-10-20.

Sample

at least two weeks of logged proposals

CollectingFrom the terminal2026-10-08 registered

Multi-day Fibonacci retracement levels are negligible intraday.

How it is checked

For every proposal, record the distance to the nearest multi-day Fibonacci level. Compare the outcome of proposals near such a level against those far from one.

Collection paused; resumes 2026-10-20.

Sample

at least two weeks of logged proposals

CollectingFrom the terminal2026-10-08 registered

The previous day's high and low are usable, but the situation decides what for.

How it is checked

Record every approach to the previous day's high or low, together with the direction of approach, how many times the level has already been touched, and whether it has been broken before. Compare outcomes by those categories.

Collection paused; resumes 2026-10-20.

Sample

at least two weeks of logged proposals

CollectingFrom the terminal2026-10-08 registered

A fair value gap is usable when there is force behind the move and the micro trend agrees.

How it is checked

For every gap reaction, record the character of the move — how much delta it took to move one point, against the sample's own median — and whether the shorter timeframes point the same way. Compare against all gap reactions as the control.

Collection paused; resumes 2026-10-20.

Sample

742 so far

at least two weeks of logged proposals

Registered

RegisteredFrom a lesson2026-10-14 registered

A pierce of a level followed by a close back inside it is followed by a move away from the level more often than a pierce that holds.

How it is checked

Define pierce depth and the close-back condition in ticks. Count every instance on a fixed history, and compare against all level touches as a control.

Sample

RegisteredFrom a lesson2026-10-14 registered

Price returns into a fair value gap more often than it continues without returning.

How it is checked

Define the minimum gap size in ATR and the return condition. Count every gap on a fixed history and record whether it was returned into within a fixed number of bars.

Sample

RegisteredFrom a lesson2026-10-14 registered

A retest after a gap or a strong move is followed by continuation more often than a first touch is.

How it is checked

Define what counts as a strong move and what counts as a retest, in ATR and in bars. Compare the two populations against each other.

Sample

RegisteredFrom a lesson2026-10-14 registered

A double top or bottom is followed by a reversal more often than a single touch of the same level.

How it is checked

Define the tolerance between the two touches and the maximum number of bars between them. Count every instance and compare against all single touches of the same levels.

Sample