201 Gann rules produced zero testable results. We kept one construct and threw the rest away.

August 14, 2026

We extracted 201 Gann rules from trading books. None ever reached twenty trades and 116 were skipped as unevaluable. Gann angles need an arbitrary scaling constant, which makes them unfalsifiable. We shipped only the Square-of-9 price ladder, and only as confluence.

W.D. Gann is the most divisive name in technical analysis. His methods are also the clearest case we have found of a framework that a backtest cannot examine — not because it fails, but because most of it cannot be stated as a testable claim.

The number


Gann rules extracted from books	201
Ever scored by our backtest mill	14
Reached 20 trades	0
Skipped as structurally unevaluable	116

Two hundred and one rules in. Nothing came out.

For comparison, in the same library RSI produced 5 testable rules from 652, and Wyckoff 2 from 81. Gann produced none from 201.

Why angles cannot be falsified

The core of Gann's charting is the angle — the 1x1 rising one unit of price per unit of time, and its fan of 2x1, 1x2 and so on.

To draw one you must first fix a price-per-bar scaling constant: how many dollars equal one bar of time on this chart. That constant is not derived from the market. It is chosen per symbol, per timeframe, by the analyst.

The consequence is decisive. Change the scaling and the angle that "held" becomes the angle that broke. Any outcome can be made consistent with some choice of constant, which means no outcome can contradict the method.

A claim that cannot be contradicted is not a weak claim. It is not a claim.

The same applies to the time-cycle and astrological material, which we exclude outright. Peer-reviewed work has looked for predictive power in Gann methods and found none beyond chance, and our extraction reflects the reason: 116 of the 201 rules could not even be written as a condition a machine could evaluate.

The one construct that survives

There is exactly one piece of Gann's apparatus that is clean, deterministic and scale-invariant: the Square of 9 price ladder.

`` level(k) = ( sqrt(anchor) ± k · step )² step = 0.125 (45°) ``

Anchored on a significant pivot, it generates a ladder of support and resistance prices. It needs no chart scaling, no bars-per-dollar constant, no choice of visual proportion. Two analysts with the same anchor get the same levels.

That is the only Gann construct we implemented. Our engine produces the ladder and nothing else.

And we ship it with no opinion attached

The engine carries no bias, no recommendation, no direction. It does not say buy or sell. It says: here are the Square-of-9 levels from this anchor.

In our rule vocabulary it is preferred, never required — a confluence overlay. It can add weight to a level another engine already identified. It can never be the reason for a decision, and no signal can depend on it.

We also say plainly in our own internals that it is the weakest expected edge of the five theory engines we run, and it is deliberately the smallest.

Why build it at all

A fair question. Three reasons.

Round numbers and pivot-derived levels do attract attention, and a scale-invariant ladder anchored on a real swing is a reasonable way to enumerate candidate levels. As one input among several, it costs nothing to check.

It is honest about what it is. A "Gann levels" annotation that produces only the falsifiable construct, carries no direction, and cannot drive a signal is a research annotation. A Gann product that draws fans and sells forecasts is selling the unfalsifiable part.

And the absence is itself a finding. We can now say precisely what we do and do not know: we have no evidence about Gann angles or cycles in either direction, because we could not construct a test. That is a more useful statement than either "Gann works" or "Gann is nonsense," and it is the only one our data supports.

The general rule

Before asking whether a method works, ask whether any outcome could show that

it does not.

If the answer requires a parameter the analyst chooses after seeing the chart, you are not looking at a strategy. You are looking at a vocabulary for describing whatever already happened — and it will feel accurate for exactly that reason.