This browser could not start WebGL, so the moving charts are off. Every figure is still in the text, and the full table is on the verdicts page.

144 books · 21,142 pages · a century of prices

The market is written. We checked.

Gann, Baumring, Jenkins, Sepharial and dozens of others said price and time obey laws. Oppollo takes every method they stated, traces it to the page it came from, turns it into code, and tests it against chance.

240methods recorded, each tied to its page
86tested across up to a century of prices
0beat chance once you count how many were tried

Scroll. Every bright point is a method.

The shelf

240 methods from 61 schools

Each point is one rule a book stated, grouped by the school it belongs to. Every one links back to the page and the quote it came from.

The colour shows how closely the record keeps to the author's own words.

Testable or not

127 became code. 113 are still only words.

A method can only be tested once it makes a dated forecast that could turn out wrong. The dim points on the right are recorded but not built yet.

The replay

86 forecasting machines, run from 1929 to 2025

Each machine made one call at a time on 22 markets and waited for that call to resolve before making the next, so no two calls overlap. That independence is what makes the evidence count.

66,699independent observations now
51in the whole record before
The test

Did any of them see the future?

Further right means the method was right more often than chance. Higher means more evidence. To count as an edge, a point had to land beyond the glowing line. That is the bar for testing 240 methods at once, and it narrows as evidence grows.

None did. The strongest result reached z 2.52. The bar is z 3.53.

Hover or tap a point for the rule and what its book said.

What is now settled

23 answers are final

For the bright points, any edge bigger than 2.6–5.0 percentage points would have shown up. None did. These questions are closed as coded.

58 more still cannot tell. They need more markets and more history, not a different test.

Checking the yardstick

If nothing works, the results should pile into a bell curve

Each point is one method's score. If there is no edge and the test is fair, the scores should stack into the shape of chance. They do, and a little more tightly.

Centre −0.04 (ideal 0). Spread 0.83 (ideal 1). A biased test would have pushed the whole pile to one side. A spread under 1 means the correction for overlapping evidence is slightly strict, so every “settled” bar is cautious rather than generous.

Where it goes next

The search is not over

The galaxy now knows which of its stars were tested. 13,344 of 21,142 pages have been read word for word. The other 7,798 are mostly scanned charts and tables that still need a careful visual read.

Every method added raises the bar for all the others. That is why the bar is priced in from the start.

The one that looked like gold

A p-value of 10−65, and a placebo that beat it

Before the replay, one rule survived correction: woods.campaign_statistics, right 23% of the time against a 13% null. The audit then tested it against a properly matched null and three controls with no doctrine in them.

+14.14ppthe original scoreboard
−0.6ppagainst a properly matched null
−0.8ppa coin flip with the same claim geometry
+17.5ppa placebo, “trailing 5-year mean return is extreme”, with no doctrine at all

The placebo beat the discovery. The edge came from the chance model inside the grader. The rule's “campaign” never reset: it used one origin for the whole century, so both of its doctrinal gates were dead code.

A tiny p-value did not make it true. The only protection is a null you honestly tried to kill your own result with.

Promises against measurements

What the books said, and what it measured to

These are the claims that stated their own hit rate, rebuilt and measured as written. A hit rate with no matched null isn't a weak result. It isn't a result at all, and almost none of this literature supplies one.

stated in the bookmeasured as writtenbars show published ranges · right column is the gap

The shelf, counted

Word for word, page by page

Every quote resolves to a page, a hash and a layer. Pages are counted from the page ledger, not from memory. The unread share is mostly scanned figures, which is where Gann kept the Square of Nine and his master charts.

Go and look

Nothing here is a summary you cannot inspect

Every surface reads the same files this page counts from.

The VerdictsEvery replayed rule, its lift against its own null, and the smallest edge it could have seen.86 rules · 23 settled The Memory GalaxyEvery method, source, quote and page as a body in space. Search it and open any claim.1,862 bodies The Gann InstrumentReal bars with the geometry drawn by the project's own code, both angle doctrines at once.bars back to 1927 The MethodsEvery verified passage, per book and per page, with its locator.the deep record The AuditsThe refutations and findings, including this project's own mistakes.where the failures are kept