Uncorrelated — QEIv18™ Energy Risk Intelligence

01 — The measure

How the correlation was measured, and how it was to be described.

Each rule produces a daily state, invested or hedged, and so does the reading. The two series were compared over the whole record with the phi coefficient — the standard correlation between two binary series. A value of 1 means the rules agree on every day; 0 means that knowing one state says nothing about the other; a negative value means they tend to disagree. Before the study ran, NeoAmorfic fixed how the result would be described: below 0.30 the reading would be called uncorrelated with the rule; between 0.30 and 0.60, partially overlapping, with the overlap stated; above 0.60, substantially the same rule under another name. The highest value measured, across every instrument and every rule, was 0.23.

0.23

highest correlation measured on this platform (heating oil, downside-volatility filter)

instrument-and-rule pairs tested across the company's energy, gold and equity-index readings; every one below 0.30

0.30

the threshold below which the reading would be called uncorrelated — fixed before the result, not after it

02 — The comparison set

The rules it was measured against.

The comparison set is the four overlay rules a systematic desk or a risk system is most likely to be running already, each computed from the same daily settlement prices as the reading, on the same instruments and the same dates.

RULE 01

Realised-volatility filter

Hedged when the instrument's recent realised volatility is high against its own history. The rule most risk systems already contain in some form.

RULE 02

Downside-volatility filter

The same construction on negative returns only — the semivariance of the down days. It is the closest conventional relative of a reading built on downside disorder, and the one most likely to overlap with it.

RULE 03

Volatility-managed rule

The rule of Moreira and Muir (2017), which cuts exposure as recent variance rises, expressed as a daily invested-or-hedged state so that it can be compared day for day.

RULE 04

Trend rule

Invested when price is above its 200-day moving average and hedged below it. The trend overlay most widely used across commodity and index books.

03 — Position correlation, instrument by instrument

Every pair came in below 0.30.

Phi coefficient between the QEIv18 reading and each rule, daily states, June 2010 to September 2026. Gold is shown as it appears on this platform: a macro risk reference on the same lens.

Instrument	Contract · lens	Realised vol	Downside vol	Vol-managed	Trend
Crude	NYMEX CL · 2nd-month · Entropy · 60-day	0.05	0.17	0.05	0.05
Heating oil	NYMEX HO · Entropy · 60-day	0.07	0.23	0.07	0.10
Gasoline	NYMEX RB · Entropy · 60-day	−0.03	0.17	−0.04	0.02
Gold reference	COMEX GC · Entropy · 60-day	−0.11	0.16	−0.09	0.04

Reading the table. Each cell is the phi coefficient between two daily state series over the whole record. Shaded cells mark the highest values found. The same measurement on the company's gold and equity-index platforms — a second gold lens and two lenses on each of three index futures, 28 further pairs — gave a highest value of 0.23 and a lowest of −0.14.

04 — The conditional test

What the reading adds to a desk that already runs one of these rules.

A low correlation shows that two rules disagree; it does not show that the disagreement is worth anything. The second test takes only the days on which a conventional rule held the position, splits them by what QEIv18 read on the same day, and compares the average return on the following day. If the reading carried no information, the two averages would be the same.

Conventional rule holding the position	Next day, QEIv18 invested (bp)	Next day, QEIv18 stood aside (bp)	Difference (bp)
Crude · NYMEX CL · 2nd-month · Entropy · 60-day
Realised-volatility filter	+3.4	−8.0	+11.4
Downside-volatility filter	+5.1	−4.9	+10.0
Volatility-managed rule	+9.3	−6.5	+15.8
Trend rule	+12.5	−13.4	+25.9
Heating oil · NYMEX HO · Entropy · 60-day
Realised-volatility filter	+7.8	−8.6	+16.4
Downside-volatility filter	+8.1	−12.5	+20.6
Volatility-managed rule	+9.7	−7.2	+16.9
Trend rule	+16.6	−12.3	+28.9
Gasoline · NYMEX RB · Entropy · 60-day
Realised-volatility filter	−2.3	−0.4	−1.9
Downside-volatility filter	+0.2	−2.3	+2.5
Volatility-managed rule	+1.7	−0.9	+2.6
Trend rule	+6.4	−4.0	+10.4
Gold · COMEX GC · Entropy · 60-day
Realised-volatility filter	+5.3	+0.1	+5.2
Downside-volatility filter	+3.6	−3.5	+7.1
Volatility-managed rule	+5.8	+1.1	+4.7
Trend rule	+6.8	+0.7	+6.1

What the figures show. On crude and heating oil, the days a conventional rule held the position while the reading had stood aside returned between ten and twenty-nine basis points a day less than the days both were invested, against every rule tested. On gold the difference was between five and seven basis points. On gasoline it was smaller — between two and ten basis points against three of the rules, and reversed by two basis points against the realised-volatility filter. Run the other way, with the days on which QEIv18 was invested split by the exposure-matched volatility rule's state, the conventional rule added no discrimination: on all four instruments the days the volatility rule had stood aside returned more on average than the days it held. Averages are frictionless close-to-close returns on the following session, in basis points per day, over the whole record; the number of days in each group is in the study record.

05 — Turnover and timing

It changes state less often than a volatility rule, and on different days.

At the same fraction of days hedged, the reading changed state about nine times a year on each energy instrument — 9.0 on crude, 8.7 on heating oil, 8.8 on gasoline — and 6.2 times on gold, while a realised-volatility rule calibrated to the same exposure changed state between fourteen and seventeen times a year. The reading does not move before the volatility rule: the median lead between corresponding changes of state was zero days. The two differ in which days they stand aside, not in how early.

≈ 9 / yr

QEIv18 changes of state, energy instruments

14–17 / yr

realised-volatility rule at the same exposure

0 days

median lead between corresponding changes of state

06 — Registered before the result

The test was registered before it was run.

The four rules, the correlation measure and the three descriptive bands were set out in a registration document on 24 September 2026 and fixed by SHA-256 hash before any figure was computed. The confirmatory follow-up — the same rules calibrated to matched exposure, a null of random state sequences at the same exposure and spell length, turnover, timing and the conditional test run in reverse — was registered as an addendum, again before its figures existed. Both hashes are reproduced below; the registration documents and the study code are available to qualified institutions on request.

Registration · 24 Sep 2026sha256 81cbc6cc7604a1db94dd093241f71cc3d3cf8db88ce595911293a4f5cbae3fff

Addendum · confirmatory follow-upsha256 9e123902fa6cf6f20d3729e0a6180276027762c5e57bbcd36cae9d39f24f5bd9

Why it matters. A threshold chosen after the figures are known can be moved to suit them; one hashed before they exist cannot be. The vocabulary on this page — uncorrelated, partially overlapping, substantially the same — was decided when the outcome was still unknown.

The boundary of the claim

What this page supports — and what it does not.

Stated plainly, because the discipline is the product. The page establishes independence from conventional rules by measurement; what a desk does with the reading remains the desk's own decision.

It supports: that the QEIv18 reading is uncorrelated with the four conventional overlay rules on every instrument it covers, under a measure and thresholds fixed in advance; that on the days a conventional rule holds the position, the reading carries information about the following day's return which the rule does not; and that it does so at roughly half the turnover of a volatility rule at the same exposure.

It does not claim: a guaranteed or forward return; a price forecast; that the reading is a fund, a strategy or a recommendation; or that any past result guarantees a future one. Risk-state information, not investment advice.

Uncorrelated with the rules a desk already runs.