Cross-Domain Validation of a Moving-Average Divergence Framework — the plain-English companion

the paper (DOI) · code & data repository

Companion to the research paper of the same title. This is education, not investment advice. Nothing here tells you what to buy, sell, or predict. It explains what the paper found, how the checking worked, and what the results do and do not mean.

The claim in one sentence

Math tools invented for the stock market — where nobody can ever check the answer key — were taken to four fields of science where the answer key exists, and they passed, failed one test and said so, and then bet their reputation on this winter’s flu season, in public, with rules anyone can score.

Start with a report card

Imagine tracking a student’s grades two ways at once. One number is the average of just the last month — call it the fast average, because it changes quickly when something changes in the student’s life. The other is the average of the whole past year — the slow average, with a long memory, hard to budge.

Most of the time the two numbers sit close together. But when something real shifts — a new job eating homework time, a spark of motivation, trouble at home — the fast average moves first, because recent weeks dominate it, while the slow average barely notices. A gap opens between them. And that gap is information: it whispers that the situation is changing before the yearly average admits anything is different.

That gap — the fast average minus the slow average — is the entire tool this research line studies. Traders call their version MACD. Economists have a version for spotting recessions. This series of papers treats it as one mathematical object, proves theorems about it, and asks what it can and cannot detect.

This particular paper asks the scariest question in the series: is any of it actually real?

The problem: a tool that grew up where cheating is undetectable

Every previous paper in this line tested the gap-tool on financial markets. There’s a hidden weakness in that, and it’s worth sitting with: nobody knows the answer key for markets. No one can say, from independent physics, exactly how a stock price is “supposed” to behave. So when a tool “works” on market data, there are two possible explanations — it captured something true about the world, or it accidentally memorized the quirks of past data, the way a student can ace a practice test by memorizing the answers without understanding anything. From inside the market data, those two look identical. Forever.

Nature doesn’t have this problem. A river is a system with a known answer key: rain falls, the river swells, the water drains away, and the flow settles back toward its normal level. Physicists can tell you how strongly it settles back, from hydrology alone, without ever seeing the paper’s math. Daily temperatures at a weather station wander almost randomly from day to day — meteorologists know this independently. Sunspots follow the sun’s famous eleven-year cycle. And influenza is the extreme case: every single summer, flu cases collapse back to nearly zero, like a trampoline that always returns to flat. Epidemiologists have known this for a century.

So the paper takes the framework completely unchanged — same formulas, same settings, not one dial re-tuned — and points it at weather, rivers, sunspots, and the flu. If the tools are real, each field should score exactly where its known physics says it should. If the tools were secretly memorizing market quirks, four fields of science are about to expose them.

Test one: can the tool rank nature correctly?

The framework computes, for any system, a single number describing how strongly that system snaps back to its own normal — think of it as measuring the strength of the rubber band pulling the system home. A skateboard on a flat floor has no rubber band: give it a push and it just rolls, staying wherever it ends up. A trampoline has a powerful one: however hard you land, it flings you back toward flat.

Here’s the beautiful part: science already knows, independently, which of the four systems is a skateboard and which is a trampoline. Daily temperature wanderings — nearly a skateboard. Rivers, once you remove the seasons — a moderate rubber band. Influenza — the strongest trampoline in the panel, because summer always wins.

The tool was shown the raw data and nothing else. It ranked all four systems in exactly the order the physics demands — weather scoring in the near-skateboard zone, rivers in the middle, and flu at the top with the most trampoline-like score in the entire study. As a trap, the authors also fed it a synthetic skateboard — computer-generated data mathematically guaranteed to have no rubber band at all — and the tool scored it precisely where the math says a pure skateboard must land. It didn’t just “work.” It sorted four sciences by the strength of their physics without being told what any of them were.

Test two: calling the direction of the needle — in advance

The gap-tool’s deeper theory contains a theorem with an almost spooky promise. It says you can predict, before looking at any outcomes, whether a widening gap means “more of this coming” or “the opposite is coming” — and the prediction comes from one measurable property of the data: its persistence.

Persistence is a simple idea dressed in a scary word. Some things in life have momentum: a student on a hot streak tends to stay hot; a rainy week tends to be followed by more rain. Other things snap back: eat an enormous dinner tonight and tomorrow’s appetite is probably smaller, not bigger. Momentum systems and snap-back systems respond to a widening gap in opposite directions — and the theorem tells you which kind you’re holding, just from the data’s own habits.

The paper turned this into the hardest kind of test: a called shot. Before computing a single result, it wrote down twenty-five predictions — five natural systems crossed with five time horizons, each with the direction of the answer declared in advance, like a pool player naming the pocket before the shot. Then it looked. Of the thirteen predictions where the data was strong enough to give a clear answer, all thirteen came out in the predicted direction — and held that direction on the older half of the data and the newer half, separately.

Buried in the twenty-five were two predictions of a special, riskier kind: places where the theorem said the direction should flip — the Colorado River at the one-year horizon, and influenza at twenty-six weeks, where the flu’s seasonal rhythm inverts the logic. Predicting “more of the same” is one thing. Predicting this is exactly where the pattern reverses — and being right, twice — is the kind of shot that’s very hard to make by luck.

The test that failed — and stayed in the paper

One more idea was registered in advance: that by splitting the national flu picture into regions and watching the shape of the wave spreading across the map, you could predict when the national epidemic would peak. It’s a lovely hypothesis. The paper tested it exactly as pre-registered.

It failed. No reliable early warning of the peak exists in that spatial signal.

The failure is reported in the paper, right beside the successes. And it had teeth — a third public prediction for this winter had been planned, contingent on this test passing. The test failed, so that prediction was never registered. The idea was allowed to die in public rather than being quietly buried or reworded into something weaker that could still “pass.”

The payoff: eight weeks of warning

Out of the wreckage of the peak idea comes the paper’s most practical result, about the start of flu season. Build the simplest possible detector from the gap-tool — fast average of flu activity minus slow average, no medical model, no flu-specific tuning, nothing but the same arithmetic used on rivers and sunspots — and compare it against the obvious alternative, a tripwire that fires when flu activity crosses a fixed level.

Across twenty-seven historical flu seasons, the gap detector fired on average about eight weeks earlier than the tripwire — roughly two months of extra warning — and the advantage held between five and ten weeks no matter how the tripwire’s level was tuned. As a bonus, running the same gap on each flu strain’s share of positive tests identifies which strain will dominate the season, at the same time. Two months is not an abstract number in public health: it’s vaccine campaigns, staffing, and hospital preparation.

The bet: this winter, in public, with rules

Everything above was computed on data that already existed — and the authors know exactly what that’s worth. History can be flattering. So the paper ends in the future tense: two dated predictions about the 2026–27 flu season, which hadn’t begun when the predictions were frozen into the public record with a timestamp no one can back-date.

Prediction one: the gap detector will beat the tripwire to this season’s onset by at least six weeks, on the CDC’s public flu data, using exact thresholds written down in advance. Prediction two: the first strain whose gap fires will turn out to be the season’s dominant strain. Both get scored once, on the first official CDC data release after September 1, 2027 — one frozen snapshot of the data, so later revisions can’t nudge the verdict either way. The rules for “confirmed,” “falsified,” and “doesn’t count” (a season too mild to test rolls forward to the next one) are all pre-written, and the registration includes a step-by-step recipe so that anyone — including a motivated teenager with a laptop — can download the free public data and check the result themselves. If either prediction fails, the failure goes into the paper’s public corrections log.

And the theory exposes an even bigger target: if any clearly-resolved result ever lands with the opposite sign from what the persistence theorem predicts, the central mechanism itself is declared wrong.

What the checking caught in this paper

The general machinery every paper in this series runs through — the fingerprinted data, the machine-checked ledger of numbers, the verification program, the adversarial review — is described once in the series’ shared verification note, which follows every companion on LaggingTruth.com. What belongs here is what the process caught and changed in this paper.

The hostile review’s biggest catch reshaped the paper’s headline. The original draft leaned partly on a statistical shortcut — a quick formula for judging which grid results were “significant.” The reviewer demonstrated that, the way the data had to be sampled, this shortcut graded too generously — like a teacher whose curve quietly inflates everyone’s marks. The fix wasn’t to defend the shortcut; it was to demote it. The paper’s weight moved onto the result that needs no grading curve at all — the called-shot directions holding on both halves of the data, a fact you can check by looking — and the shortcut now appears only as a clearly-labeled cross-check (where, for what it’s worth, it agrees: thirteen graded, thirteen right, zero wrong). The review also caught the draft misdescribing the mechanism behind the Colorado River’s predicted flip — the corrected explanation is what shipped. Every finding, major and minor, is answered in writing in the public repository: fixed, or rebutted with a reason.

What this cannot do

• It does not validate trading. The tools proved they measure something real in systems with rubber bands. Markets remain the place with no answer key; nothing here picks stocks, and the papers say so at every opportunity.
• The flu detector is a demonstration, not a public-health system. Twenty-seven historical seasons are encouraging; the two live predictions are the real exam, and it hasn’t been graded yet.
• The peak-timing idea is dead. The paper’s own registered test killed it. Onset warning survived; peak prediction did not.
• The direction-calling theorem needs the data’s habits measured well. Its magic ingredient — persistence — must be read carefully, especially in strongly seasonal systems like flu, where the rhythm itself is what flips the logic.

Everything remains attackable: the called-shot grid, the fingerprinted data, the verification gate, and the two open predictions with their September 2027 scoring date are all public, and a loss goes in the public log.

The takeaway

There is exactly one honest way to find out whether a tool measures reality or just memorized its homework: take it somewhere reality is independently known, and let it be graded. This paper did that four times. The same unchanged mathematics ranked weather, rivers, sunspots, and flu by the true strength of their physics; called thirteen directions in advance — including two deliberate reversals — and hit all thirteen on both halves of the data; turned itself into a flu alarm with two months of early warning; and let its one failed idea die in public. Now it waits for this winter’s flu season to grade two predictions it can no longer edit. Tools that survive where the answer key exists have earned a hearing where it doesn’t.

This companion is licensed CC BY-NC 4.0. The research paper it accompanies is licensed CC BY-NC-ND 4.0, and the analysis and verification code is MIT-licensed. Education, not advice: nothing in this document is financial advice, an investment recommendation, or a forecast.