A trading engine built from the primary sources of W. D. Gann and Dr. Jerome Baumring — where every order replays to a page, and every claim is graded against its null.
Every order replays to its evidence, and every piece of evidence to a page.
Oppollo is not an indicator stack. It is a chain of custody. Nothing votes that cannot be traced, and nothing trades that cannot be graded.
01
Page
13,344 pages verified across 144 books — OCR'd, hash-pinned, and re-read against the scans.
02
Rule
A falsifiable sentence with a verbatim quote, a locator, and the page's SHA-256.
03
Method
Pure code implementing the rule. The model never does arithmetic.
04
Claim
A prediction with its tolerance declared before the outcome is known.
05
Verdict
Cross-lineage confluence. A school agreeing with itself is not confluence.
06
Grade
A pure-function grader rules on every claim. Sample sizes always attached.
“Divide your capital into 10 equal parts and never risk more than 10% of your capital on any one trade.”
W.D. GANN — Form Reading course, bound at Gann Harmony vol. 7, scan 0111 · sha256 1cd8… — this exact rule sizes every position the engine will ever take.
The Measurement
144 books read. Nine claims testable as written. One beat its null.
A hit rate with no matched null is not a weak result — it is not a result. Almost nothing in this literature supplies the second number. These are the claims that stated a rate, measured as written against a matched null: the same rule with the method removed.
0books read page by page
0pages verified
0claims testable as written
0beat a matched null
What the book says → what it measured
stated in the bookmeasured as written
Not plotted: Mikula — “(band, not a rate)”. Coercing a band into a percentage to fill the chart would be inventing the number.
Every pair above is from the graded ledger; bars are published intervals. Full workings, sources and page locators: /methods/.
The Memory
Every rule the engine holds, clustered by the hand that wrote it.
The doctrine ledger, fetched live and drawn as a memory map: each point is one rule record, clustered by author, colored by how close its code sits to the page. Hover any point for its claim and source. This is the 2D shadow of the World — the same ledger in three dimensions.
Every substantive page of the corpus was read — no sampling. Every extracted procedure carries its locator. And every method's voting power is a function of how close its code sits to the page. The bars below are filled at runtime from the live ledger; if the ledger cannot be reached they stay empty rather than invent a number.
—rule records (live)
—verbatim quotes
—paraphrase
—inferred
verbatim
— · weight 1.0
paraphrase
— · weight 0.7
inferred
— · weight 0.3
engineered
— · weight 0.0
The Masters
One tradition, many hands.
Baumring bound his sources into the seminar volumes themselves. The library on disk carries the whole lineage — and each school votes as its own lens, graded on its own record.
W.D. Gann
Time · Form · Number
Form Reading, the fourth test, overbalance of space and time, the Master Time Cycles, the tenth-part capital rule — implemented from his own typed courses.
Jerome Baumring
The Synthesis
The Sedona seminars: nine volumes weaving Gann into Helmholtz's resonance, Hambidge's symmetry, and the law of vibration.
Nirenstein
Waves · Vectors · Means
261 pages of 1986 lecture notes: vector counting, empirical wave sets, the ellipse whose major axis is time, the three means as retracements.
Pickell-Daniel
The Circumscribed Objective
Price and date from one compass construction — recovered from a destroyed plate by vision, verified against its own 1933 wheat example.
Sepharial & Bayer
Cycles of Recurrence
The Kabala of cycles, planetary periods, and the recurrence machinery bound through volumes 3, 6 and 9.
The Curriculum
Helmholtz · Graydon · Hambidge · Colman
The reading list Baumring assigned — much of it physically bound into the corpus, waiting to be mined for the methods nobody ever coded.
“Only when there are 3 bodies together is there an interaction of an effect. One alone is not enough.” — the notes, p. 107
The Engine
Running now. Trading nothing.
The engine climbs a ladder — RESEARCH → SHADOW → PAPER → LIVE — and each rung is earned per method on graded, out-of-sample claims. Not a switch a human flips on a good day. Its own library, graded on the same terms as the books, came in below its null — published in full at /engine/.
RESEARCHconnecting to the engine…
—claims on record
—cited rules
—graded outcomes
—quote fidelity
Full telemetry, claims and grades — with sample sizes — in the Engine. The measured record of every method is in the Findings.
The Findings
The verdict is scoped, and the scope was never 100%.
We tested 81 implemented methods against surrogate markets built to have nothing in them, and none scored outside what the best of forty-four rules scores on a market with its structure destroyed. Every family-corrected p-value was 0.529 or worse. The detector works: a planted 100% signal reads 8.5 standard deviations clear of its own surrogate, and it still sees a planted edge at 10%.
That result stands. What was wrong was this page. The finding itself said from the day it was written that it was a verdict on 16% of the corpus and not a verdict on the literature. This page dropped that sentence.
Here is why the sentence matters. Those 81 methods were mined from books we had swept, not read. We have since read the shelf end to end and verified methods against the printed page — and in at least one case the method we tested was not the method the source teaches. So the tradition-level claim comes down until the re-measurement runs against what the books actually say. The run stays published at /research/data/.
The system this one replaced reported 100% accuracy. Re-graded honestly, its record was 1 for 27 — and eight more of its claims could not fail at all, so they counted for nothing. We publish that number because an engine that grades itself gently is worthless. The methods were never wrong. They were never tested. Now they are — and the engine's own library, scored against a matched null on the same terms as the books, came in below it. That is published too.
A claim's tolerance is declared before the outcome. Duplicates collapse. A pass whose probability by pure chance exceeds one-half is discarded as non-evidence. No number is quoted without its sample size. Below n = 20, we say so in the same sentence.