RegimeEdge Validation

How the engine performed
when it mattered most

A plain-English breakdown of RegimeEdge across five major market crises. No jargon. Just the structure, the signal, and the honest result.

16 years of data 4,073 trading days 5 crisis periods audited SPY · QQQ · GLD · TLT · IWM
−0.90% avg price move in the 20 days
before a flip to caution
175 total regime flips studied
across 16 years of SPY
20 days average early-warning window
before escalation flip

On average, price has moved only −0.90% when the engine first flags caution. The engine detects structural deterioration before it becomes obvious in price.

Crisis-by-crisis breakdown
Five major events. Real signal history. No smoothing or hindsight adjustments.
2020 COVID Crash
Jan 2020 — Jun 2020  ·  SPY
−33.7% drawdown 16 days early

As markets were still near all-time highs in late January and early February, the engine began detecting a structural shift in the underlying market environment — rising stress indicators, falling signal coherence, and an early forecast volatility pickup. By the time price started falling meaningfully in late February, the engine had already moved to a cautious regime.

The signal was in reduce_risk five trading days before the bottom on March 23, and the full recovery signal followed as the market structure rebuilt over April and May — before price had fully confirmed the reversal. This is the clearest example of the engine reading structure, not reacting to price.

−33.7% SPY max drawdown
16 days lead time before trough
reduce_risk signal 5 days before bottom
The engine also caught the recovery early — the signal began moving toward calm before price had fully turned, giving time to re-engage exposure methodically rather than chasing.
2022 Bear Market
Dec 2021 — Oct 2022  ·  SPY
−24.5% drawdown 213 days early

This was the engine's most impressive lead time. Over seven months before the October 2022 trough, the engine detected the structural deterioration that would eventually become one of the worst bear markets in a decade. The signal moved to cautious in late 2021 when price was still near highs — well before the consensus had turned bearish.

Throughout 2022, the engine remained in an elevated or severe risk regime with high confidence — not flip-flopping on every rally attempt. This stability is what allowed the signal to be actionable rather than noisy. The October bottom was correctly flagged with a reduce_risk signal, and the de-escalation sequence began in Q4 as conditions rebuilt.

−24.5% SPY max drawdown
213 days lead time before trough
reduce_risk signal 5 days before bottom
The 2022 bear is the clearest demonstration that the engine does not need a crash to respond — it reads the slow deterioration in market structure months before drawdown accumulates.
2018 Q4 Selloff
Oct 2018 — Jan 2019  ·  SPY
−19.4% drawdown 45 days early

The engine was in a calm regime through mid-2018, then began detecting structural deterioration in mid-October — 45 days before the December 24 trough. As the selloff accelerated into the worst December since the Great Depression, the signal escalated from elevated to severe risk, the most defensive posture available.

The recovery was tracked with equal precision: the engine flagged the bottoming process near the December low and began signaling a structural improvement before price had fully turned. By late January the regime had de-escalated as market conditions rebuilt.

−19.4% SPY max drawdown
45 days lead time before trough
reduce_exposure signal at trough
2011 Debt Ceiling Crisis
Jun 2011 — Oct 2011  ·  SPY
−18.4% drawdown 39 days early

The U.S. debt ceiling standoff and subsequent S&P downgrade of U.S. debt caused a sharp summer correction. The engine moved to a defensive posture 39 days ahead of the August trough, as the structural signals showed increasing stress well before the political situation became front-page news.

At the deepest point of the drawdown, the signal was in reduce_exposure — the most defensive posture — correctly reflecting the severity of the environment. The recovery signal began in early September as conditions stabilised.

−18.4% SPY max drawdown
39 days lead time before trough
reduce_exposure signal at trough
2015 Flash Crash
Aug 2015  ·  SPY
−11.9% drawdown Missed — speed crash

The August 2015 flash crash was a single-session event driven by overnight selling in Chinese markets, triggering a cascade at the U.S. open. The drop was so fast and so concentrated that no daily regime signal — ours or any other — could realistically detect it in advance.

The engine was in a hold posture going into the crash and moved cautious after the initial drop. This is correct behavior for a regime-detection system: it reads underlying market structure over days and weeks, not intraday shocks. We publish this miss because honest validation requires it.

−11.9% SPY drawdown
0 days lead time (missed)
hold signal before crash
Why we include this: A single-day flash crash driven by overnight overseas selling is outside the detection window of any daily regime system. The engine is designed to protect against sustained structural deterioration — not to predict one-off shocks. Including the miss is part of how we build trust.
How the engine detects regime shifts
Plain-English explanation of what the engine is reading — without the internal signal names.

Structural stress builds before price falls

In the 20 trading days before the engine flips to a cautious regime, market structure is already deteriorating — forecast volatility is rising, realised volatility is picking up, and internal signal coherence is falling. On average, price has moved only −0.90% at that point. The engine is reading the structure, not reacting to the price move.

Recovery confirmation is deliberate

De-escalation from stress back to calm happens more slowly than the initial escalation. The engine waits for multiple structural conditions to confirm before calling "all clear." On average, price is already up +1.60% before a de-escalation flip — the engine prioritises not calling recoveries too early over being first.

Stability over sensitivity

Across 16 years and 175 regime flips, the average regime lasts 2–5 weeks. The engine does not flip-flop on day-to-day volatility — it waits for sustained structural change. This makes the signal actionable for real trading decisions rather than producing noise.

Not a price prediction

RegimeEdge does not predict whether the market will go up or down. It identifies the current risk environment and flags when conditions suggest increasing or decreasing your exposure. The edge is in risk reduction during stressed regimes, not direction forecasting.

Where the engine excels — and where it struggles
Tested across 12 assets over 16 years. The pattern is consistent.

Liquid, structurally-driven assets

The engine performs best on SPY, financials (XLF), and oil (USO) — assets that respond to macro regime shifts. Vol edge is strong (+50–70% forecast vol on cautious days vs calm), and drawdown protection is meaningful. These are the cleanest signals in the system.

Highest conviction when assets align

When SPY, GLD, TLT, and QQQ simultaneously signal caution, the negative-return probability nearly doubles — from ~23% to 41%. Cross-asset agreement is the strongest state the engine can reach. It has occurred on 190 days across 16 years — rare, but decisive when it fires.

A practical time horizon per regime

Reduce-risk is a short-term warning — median 7 days, typical range 3–13. Reduce-exposure commits longer — median 30 days, up to 87. When the engine escalates that far, history says the stressed environment persists for weeks. One is a flag, the other is a sustained bear signal.

Gold, silver, and emerging markets: supporting role

These assets follow their own safe-haven or supply-shock dynamics that don't align cleanly with the structural equity model. They work best as confirming indicators alongside equity signals — when GLD is stable while SPY is cautious, that's additional conviction, not a standalone directional call.

The honest limits of any regime signal

No daily signal can catch single-session flash crashes driven by overnight external shocks (as the 2015 event shows). The engine is designed to protect against sustained structural deterioration — the kind that builds over days and weeks before becoming obvious in price.

Fast, exogenous shocks remain a genuine risk for any daily regime system. We believe being explicit about this builds more trust than pretending it doesn't exist. All validation numbers on this page are from live historical signal data — no backcurve fitting or hindsight adjustment.

Last validated April 2026 · SPY · 4,073 trading days · hello@regimeedge.com for methodology questions

Simple output. Serious edge.

One daily regime signal per ticker. Before the open. Early access from $79/mo, grandfathered forever.