# Oppollo — Findings

*Updated 2026-08-17. Written for the site; every number below traces to a
reproducible report in the research repository.*

## What Oppollo is

Oppollo is a research system that takes the classical market-forecasting
literature — W.D. Gann, George Bayer, Sepharial, Bradley Cowan, Michael
Jenkins, Baumring's Gann Harmony seminars, and the harmonic, geometric and
number doctrine underneath them — and tests it the way the authors never did:
every method implemented exactly from its primary pages, every prediction
written down before the outcome, every result scored against what pure luck
would have achieved.

The library behind it: **eight source folders, ~122 books, 20,637 distinct
pages**, every page fingerprinted, every quoted sentence verified
letter-for-letter against the page it came from.

## The measurement

Every method was run across **98.6 years of market data** — the S&P from 1927,
six major stocks from 1962, gold, silver, oil and the grains — producing
**1,478,199 falsifiable predictions**, each one graded and each one priced
against a measured model of chance.

The result: **no method shows a repeatable edge.** Zero of 81 implemented
methods survive correction for the number of ways the corpus gets to try. The
system as a whole scores 0.4 points below what luck alone would score.

Three methods briefly appeared to win. All three were refuted by adversarial
audit, and the strongest apparent result in the project's history — a Woods
campaign rule at odds of 10⁶² against chance — turned out to be an error in
the measuring instrument itself, not in the market.

## The instrument mattered more than the answer

The most valuable output of the project is the ruler. The model of "what luck
would score" was found wrong **five separate times**, each time in the
direction that flattered the methods, and each error had survived inspection
because its up-side and down-side halves cancelled in any averaged check.
Documented in full in THE_RULER.md. Among the corrected errors:

- a window's high and low are not uniformly placed — lows cluster at the start
  and highs at the end, and every timing claim had been collecting up to 12
  points of free accuracy from the difference;
- after a market fall, a bullish one-year call is ~60% likely to come true by
  mean reversion alone — the model priced it at 5.6%, paying any
  "buy-the-dip" method ten times over;
- predictions repeated every few days over 90-day windows were being counted
  as independent evidence at up to 291 copies per real observation.

Every replacement model is measured on real bars, validated out of sample on
data it never saw, and kept beside its retired predecessor so the size of
every correction stays auditable.

## The harmony question

The tradition's central claim is that the methods work **together** — one
system, many doors. That was tested directly, and the finding is the most
interesting in the project:

**The harmony is real.** The methods genuinely converge — so strongly that
the 81 implemented rules collapse to about **seven independent voices**. Some
pairs agree on 100% of the occasions they speak together. Cowan's Venus
square lands on Gann's month grid 84% of the time. The convergence Baumring
taught is measurably there.

**But it is self-agreement, not market knowledge.** When several
independent-looking methods point at the same date, that cluster passes about
10 points above naive expectation — and the entire surplus reproduces under
shuffled labels and on surrogate markets where nothing is real. The best
single method inside those clusters performs exactly at chance. A detector
proven able to find a deliberately planted confluence edge found none in the
corpus. The methods echo each other; the echo is not information.

## What survives

Three things, and they are not nothing:

1. **The map.** The complete doctrine graph — which method derives from
   which, which are one construction wearing different names, where the
   schools contradict each other — extracted from primary pages with
   citations on both ends of every edge. This did not exist before.
2. **The instrument.** A grading system that caught five of its own errors
   and three of its own false positives, with the machinery (measured nulls,
   surrogate markets, overlap-priced statistics, adversarial audit) reusable
   on any forecasting claim whatsoever.
3. **The sealed year.** One final year of data has never been touched by any
   measurement. It remains available for the one thing that ever earns the
   right to spend it.

## Status

Research phase. A full-depth surrogate sweep — the last untried lens — is
running now. Nothing here is trading advice; the finding to date is the
opposite of one.