Round numbers and half-figures: the clustering is real, the trading rule usually is not

[Amanhasnoname]

Orders really do cluster at round levels. Stops, take profits and option-related hedges tend to sit at the figure (1.1000), the half-figure (1.1050) and sometimes the 20 and 80 levels. Anyone who has watched price stall at x.xx00 and then spike through it has seen the effect.

The problem comes when that gets turned into "buy the figure, sell the figure". Clustering means two opposing things can happen: price can bounce off a pile of resting limit orders, or slice through a pile of stops and keep going. Which one wins on a given day depends on whether there is flow behind the break, and a chart rarely says so in advance.

A more testable approach is to treat the level as a zone with timing rules. Log your entries within, say, 3 pips of a figure for a month. Split them into touches that rejected within two minutes, and touches that broke and held for five. Then check whether size was the difference, meaning the first touch after a long approach versus the third or fourth touch.

Repeated tests tend to wear levels down, so the first touch often behaves differently from later ones.

What have your own logs shown about first touch versus repeated touches at the big figures?