The claim

Traders are taught that price is drawn to liquidity, so a target at an old high or an un-swept pool should be hit more often than an ordinary level. Many also treat 'internal liquidity first, then external' as a sequence you can lean on.

How we tested it

We took 38,952 trades on NQ 3, 5 and 15-minute charts (2021–2026, training and validation periods only; the final holdout was never opened) from three structural entries: an inverted fair value gap, a swept-and-reclaimed swing and a displacement close. Trades were grouped into ten buckets by target distance, from 0.28R to 17.45R. Each bucket was compared with 1/(1+k), the chance that a driftless random walk travels k times the stop distance before it hits the stop. A bar that touched both stop and target was booked as a loss.

What we found

The observed win rate stayed within 2.3 percentage points of 1/(1+k) in every bucket. At 1.23R it was 45.0% against a chance line of 44.9%. At 2.44R it was 28.6% against 29.1%, and at 4.88R, 15.3% against 17.0%. The largest gaps were at the nearest targets, where booking same-bar touches as losses bites hardest.

An un-swept pool had no pull beyond its distance. Against levels at the same distance with nothing there, the largest lift across 36 configurations was 0.36 percentage points. 'Internal before external' turned out to be geometry: the nearest internal pool sits between price and the range extreme, so it was reached first in all 20 cells, exactly as distance alone predicts.

A sweep of internal liquidity looked like a 13.4-point boost to the chance of reaching the external level. Matched on distance, it was +0.6 points on training data and −0.1 on validation. The sweep had simply moved price closer.

Gross expectancy was −0.017 ± 0.013R over 82,364 resolved targets: zero within measurement. The one thing a structural entry changed was the stop, 1.52 × ATR against 0.77 × ATR at a random bar, which halves trading costs as a share of R (3.6% against 7.2% on NQ 5-minute).

What to do with it

Before a trade, convert the distance to your target into R and read 1/(1+k): 50% at 1R, 33.3% at 2R, 16.7% at 5R. In this test, the story behind a level did not change those odds; only its distance did.

Limits

The study could not detect an effect smaller than about 0.05–0.08R per trade, and it used one stop model, a 24-hour horizon and unfiltered mechanical entries. Gold had too few signals to confirm or contradict the NQ result.

Education only, not financial advice. These are measurements of past data, not a forecast or a signal.