Test edges against chance, not story

Description is not explanation

This editorial holds that market phenomena can be described while remaining unexplained. It also holds that small-group market simulations had not shown whether greed and fear have deeper roots.

The archive therefore keeps returning to a narrower question: whether an observed behavior differs significantly from random behavior. That discussion runs through correlation, dependency, chi-square testing, and significance levels.

Ask whether the behavior differs from chance

Price patterns, even when wrapped in trading rules or filters, are treated as statistically identified constructs that retain error and uneven significance.

The chance question is asked with several tools. Correlation is a measure of association used to ask whether two market series move together beyond what chance would suggest. Dependency describes whether successive price or indicator observations are related rather than independent. A chi-square test is a count-based check of whether observed market outcomes differ from the frequencies expected under a chance or independent model. Significance is a statistical judgment of whether a measured difference from chance is large enough to treat as more than sampling noise.

Prefer a model that cannot fight the tape

Moving averages are offered as indicators that cannot travel opposite prevailing price. A moving average is a smoothed series built from ordered prices, so its direction is constrained by those prices and cannot travel opposite the prevailing move.

An oscillator is a derived reading such as stochastics, momentum, or relative strength that can move contrary to price. Those readings routinely can travel opposite the prevailing move. Relative strength is singled out as being valued for preceding turning points rather than staying aligned with price. This editorial prefers the moving average on that alignment, not because an oscillator is unused, but because its output can fight the tape.

Keep the stop simple enough to estimate

Cataloguing every combination and computing occurrence probabilities is rejected as unhelpful complexity. Simple behaviors estimated with straightforward nonparametric distributions, illustrated by stop placement, are treated as still-open work.

A stop-loss is a precommitted exit that bounds a losing position before and during a trade, treated here as a simple behavior that can be studied with straightforward nonparametric distributions. A nonparametric distribution is a data-first description of simple trading behaviors from ranks or frequencies, without assuming a fully specified theoretical shape.

A magazine can only point

The editor reports a split audience for statistical writing, with some readers finding it tedious and others finding it not deep enough, and concludes the magazine can only point toward methods. Having treated instruction on how to use statistics as nearly exhausted, the editor plans to shift later issues toward practical examples.