Every figure on the platform rests on a sixteen-year record and a set of tests designed to break a weak result. This is the full basis: the year-by-year record, the head-to-head against the methods a desk already uses, the behaviour through real crises, the falsification battery, and the sealed forward chain that proves the calls are made in advance.
Maximum drawdown and total return under buy-and-hold versus the QEIv18 structural overlay, 2010–2026. Drawdown is reduced on every instrument; energy also exceeds buy-and-hold on return. All figures are frictionless — each desk models execution cost against its own fills. Lower drawdown is better.
| Instrument | Contract | B&H return | QEIv18 return | B&H max DD | QEIv18 max DD |
|---|---|---|---|---|---|
| Energy complex | |||||
| Crude | NYMEX CL · 2nd-mo | +25% | +199% | 80% | 69% |
| Heating Oil | NYMEX HO | +82% | +1065% | 76% | 52% |
| Gasoline | NYMEX RB | +54% | +70% | 84% | 69% |
| Gold — macro risk reference | |||||
| Gold reference | COMEX GC | +244% | +239.9% | 45% | 25% |
A crash-cushion is judged on the depth of loss it prevents, not the number of winning years. Across the full 2010–2026 window, QEIv18’s own worst drawdown was lower than buy-and-hold on every instrument tested — by 11 to 24 percentage points.
| Instrument | Buy-and-hold max drawdown | QEIv18 max drawdown | Reduction |
|---|---|---|---|
| Heating Oil HO | 76% | 52% | −24 pp |
| Gold GC ref | 45% | 25% | −20 pp |
| Gasoline RB | 84% | 69% | −15 pp |
| Crude CL | 80% | 69% | −11 pp |
The honest test of a risk overlay is not whether it makes money every year — nothing does. It is what happens when the market falls. Across sixteen years, QEIv18 never had a down year in a rising energy market: every negative year was a year energy itself fell. Below, each instrument's overlay result is set against simple buy-and-hold, year by year — the down markets first, where the protection has to earn its place, then the rising markets. Down years and the few misses are shown plainly, in red.
In every one of these years, buy-and-hold was negative across the complex. The overlay reduced the loss in 14 of 18 instrument-years; the four it did not — 2014 gasoline, 2020 crude, 2023 gasoline, 2024 crude and heating oil — are shown in red, no exceptions made. (Each cell is green where the overlay lost less than holding.)
| Year | Crude B&H | Crude QEIv18 | Heating Oil B&H | Heating Oil QEIv18 | Gasoline B&H | Gasoline QEIv18 |
|---|---|---|---|---|---|---|
| 2014 | −45.2% | −35.7% | −39.2% | −13.6% | −47.8% | −54.3% |
| 2015 | −29.4% | −24.7% | −41.5% | +1.1% | −13.7% | −9.3% |
| 2018 | −23.2% | +2.6% | −18.5% | −0.8% | −26.3% | −25.1% |
| 2020 | −20.4% | −42.6% | −26.8% | −24.5% | −15.8% | +53.0% |
| 2023 | −11.2% | −2.6% | −24.6% | +2.7% | −15.8% | −30.2% |
| 2024 | −0.2% | −10.5% | −8.5% | −17.3% | −4.0% | −2.6% |
Green = the overlay lost less than holding (or turned positive); red = the overlay matched or trailed holding. In a down market, a smaller loss is the product working. The standout: heating oil 2015, where holding lost 41% and the overlay finished +1.1% — and gasoline 2020, holding −16% versus the overlay +53%.
In rising markets the overlay stays invested and participates; it gives back a little to the cost of staying alert, and on the strong structural years it has exceeded buy-and-hold outright.
| Year | Crude B&H | Crude QEIv18 | Heating Oil B&H | Heating Oil QEIv18 | Gasoline B&H | Gasoline QEIv18 |
|---|---|---|---|---|---|---|
| 2010 | +27.6% | +14.1% | +29.4% | +21.2% | +23.1% | −0.7% |
| 2011 | +7.6% | +18.3% | +15.3% | +14.5% | +9.6% | +3.2% |
| 2012 | −7.1% | +16.0% | +4.3% | −3.2% | +5.4% | +8.3% |
| 2013 | +7.1% | +14.4% | +1.2% | −14.5% | −0.8% | +2.3% |
| 2016 | +43.4% | +85.8% | +56.5% | +64.7% | +32.9% | +29.9% |
| 2017 | +9.7% | −2.0% | +20.9% | +38.7% | +7.2% | +9.2% |
| 2019 | +32.0% | +22.3% | +20.2% | +11.7% | +27.9% | +35.1% |
| 2021 | +54.8% | +58.3% | +57.0% | +79.1% | +56.5% | +54.8% |
| 2022 | +7.2% | +17.6% | +44.3% | +129.6% | +11.0% | +1.3% |
| 2025 | −19.9% | −5.6% | −8.6% | +29.5% | −14.9% | −8.2% |
| 2026 | +58.2% | +71.2% | +68.4% | +34.1% | +78.8% | +67.9% |
The most common way a backtest deceives is hindsight — choosing a method’s settings with knowledge of what already happened. A rule that only worked because its parameters were picked in hindsight collapses once that knowledge is denied. QEIv18’s thresholds were fixed in advance and never fit to commodity data, so there is no hindsight to remove — the same untouched contract runs on every instrument, and it holds. That is the signature of a real effect rather than a fitted one, and it is what the SHA-256 provenance record makes independently checkable.
The full hindsight-removed comparison against conventional systematic overlays is included in the per-instrument analysis available on request.
A risk tool earns its place in the events that hurt. The per-instrument analysis documents QEIv18’s recorded daily states through the major dislocations of the window — the 2020 crash, the 2022 bear, and the energy shocks of 2025–2026 — with exact dates and instruments, each traceable to the sealed record. Past detection of a named event does not guarantee detection of a future one; the value is the method, applied identically every day, not any single call.
Full event-by-event daily states are included in the per-instrument PDFs available on request.
The honest truth, stated plainly: energy markets are volatile, and in choppy sideways markets the system produces false signals — it steps out and gives back small amounts. That is the known cost of staying alert. What sixteen years of unfitted testing shows is where it counts — when structure actually broke, the system was substantially out of the way. Every month below is grouped by what the market itself did.
| Market regime (what the market did) | Months | QEIv18 avg | Market avg | What it means |
|---|---|---|---|---|
| Energy book — crude · heating oil · gasoline | ||||
| Big down months (market ≤ −5%) | 49 | −5.1% | −10.6% | Cushioned — roughly half the loss |
| Choppy months (−5% to +5%) | 85 | −0.1% | +0.2% | Small give-back — the alertness premium |
| Big up months (market ≥ +5%) | 59 | +7.8% | +11.3% | Rides the rally, trails slightly |
And in the months that actually destroy capital — the worst the market delivered across sixteen years — the system cushioned the fall by measured amounts:
In March 2020, energy fell −46.7% outright; QEIv18 held the loss to −22.5%, saving 24 points in the single worst month on record. Across all twelve of the deepest market months, the system reduced every one.
On the same engine, gold's structural breaks have historically coincided with broader risk-off regimes — September 2011, the 2020 shock. Shown as macro context for the energy book, not a traded signal; you decide how to use it.
A genuine structural signal and a curve fitted to the past behave differently under stress. QEIv18 was put through the tests that separate them.
| Test | What it checks | Result |
|---|---|---|
| Out-of-sample half | Holds on data the method never saw | holds |
| Fixed thresholds | No parameters fit to commodity returns | untouched |
| Shuffled returns | Effect must vanish on randomised data | collapses, as required |
| Cost robustness | Energy edge persists when institutional execution cost is applied | persists |
| Threshold sensitivity | Result stable to ±20% on the levels | stable |
A backtest, however clean, is history. The forward record removes the last doubt: each day's states are committed to a tamper-evident chain and revealed only on a one-week delay — so the record cannot be edited after the market moves.
The day's structural states are produced by the frozen contract at close.
A SHA-256 digest of the sealed states is posted publicly the same night.
The states themselves stay private through the T+7 window — the trading value has passed, the evidential value has not.
The states are released; anyone can re-hash them and confirm the match to the digest posted in advance.
The complete computation behind these numbers — raw price data, the structural engine, the monthly record, and these documents — was audited line by line and cryptographically fingerprinted. The engine uses strictly causal calculations with thresholds fixed in advance, never fit to commodity data. No look-ahead, no per-instrument tuning, no hand-entered results. The entire chain reduces to a single hash you can re-verify against the published artifacts.
The certificate records the audit findings, the registered artifact hashes, and how to verify them. It attests to the integrity and reproducibility of the computation — it is not a statement about future performance.
Stated plainly, because the discipline is the product. The evidence demonstrates a structural risk-state signal with a reduced-drawdown record and a sealed forward chain. It is not a promise of future return.
The figures here are the public summary. Qualified institutions can review the complete per-instrument numbers, the comparison test, and the live sealed chain.