.webp)
A market moves. Prices rise and fall, retrace part of a move, and return to levels they have visited before. That happens whether or not anything meaningful is driving it.
So when a figure says a level holds a certain proportion of the time, there are two explanations. Either the level is doing something — the market has a memory and it shows up at that price — or any two prices a similar distance apart would have produced the same number.
A figure that has not ruled the second out is not evidence of the first. It can be arithmetically correct and still tell you nothing.
Alderlux built a test. It takes a claim about a level, indicator or pattern and sets it against a second market, constructed from the real record with its memory removed. The twin keeps the properties the test needs to hold constant — volatility, drift, the distribution of daily movement — and loses the structure the claim depends on.
The same measurement then runs against both. If the claim performs the same against a market that remembers nothing, it was never reading structure. It was reading the distance between two prices, which is a property of any market, including one made of noise.
We did not invent this. It is established practice in the statistical literature, where it is known as surrogate data testing, and we credit it as such. What we are committing to is when we run it: before publication, rather than afterwards to defend a figure already in print. Those are different activities with different incentives.
It disqualifies work we have already done. It has sent findings back for rebuilding, and it has retired readings we would have been content to publish a year ago. It makes us slower than a house that publishes the first interesting number it finds.
We accept that. A research house is worth what its worst published claim is worth, and the cheapest way to protect that is to run the test first rather than defend the claim afterwards.
Where a published finding has been through the test, it says so. Where it has not, we do not present it as tested — the figure can still be published, labelled as what it is.
Two pieces in this archive show both sides. When Spot Buys and Leverage Sells reports a measured divergence against random bars, move-matched random bars and a spot-only control, and states how many definitions were tried. Test 9: Weak Rallies Don't Last carries a figure labelled provisional in the body, the standfirst and the metadata, because the test that would validate it has not been completed.
The standing reference is The Alderlux Null Test. It sets out what the test governs, what it does not, and why building the twin is harder than running it. That page is permanent, and it is the one we expect to be held to.