Scientific Methodology

Trend Up describes a market regime characterized by persistent positive directional movement with shallow drawdowns over extended windows.

• Total return: typically +8% to +25% over 4–6 months (annualized roughly +15% to +50%)
• Maximum drawdown: shallow, typically less than 10%, almost never exceeding 15%
• Realized annualized volatility: 8–25% for equities and commodities; crypto runs significantly higher (60–100%+)
• Persistence: at least 3 trading months of dominant directional movement, no decisive mid-case reversal

• SPY Bull Run 2017 H1 — +7.5%, max DD −3%
• QQQ Tech Rally 2017 — +13.9%, max DD −5%
• Gold Post-GFC Inflation Hedge 2010 — +7.9%, max DD −7.8%
• WTI Pre-GFC Oil Boom — +23.3%, max DD −10.6%
• BTC 2017 Parabolic ATH — +216%, max DD −34.9% (crypto outlier, broadens upper bound)

• Mean-reversion entries fading against persistent direction
• Counter-trend stops triggered repeatedly
• Cash-allocation systems lagging benchmark performance
• Volatility-targeting underweighting during low-vol bull runs
• Short-bias structures bleeding continuously

• vs. V-Recovery: Trend Up starts from neutral ground; V-Recovery starts after a drawdown of 15% or more.
• vs. Vol Compression: Vol Compression is about low realized volatility regardless of direction; Trend Up is about direction regardless of volatility.

Trend Down describes a market regime characterized by persistent negative directional movement that sustains or deepens across the case window without full recovery.

• Total return: typically −20% to −60% over 4–6 months (annualized roughly −40% to −95%)
• Maximum drawdown: −25% to −70%, with median around −50%
• Realized annualized volatility: elevated, typically 25–90%
• Persistence: dominant downward trajectory across the full window — drawdown deepens or sustains, no full V-recovery

• Gold Taper Tantrum 2013 — −19.4%, max DD −25%
• SPY Lehman GFC 2008 — −36.4%, max DD −42%
• WTI Saudi Oil War 2014 — −48%, max DD −53%
• BTC Crypto Winter 2018 — −52.5%, max DD −66%
• QQQ Dotcom Crash 2000 — −58.5%, max DD −67%

• Buy-the-dip systems re-entering into deepening drawdowns
• Long-only momentum systems failing to exit fast enough
• Dollar-cost-averaging extending drawdown duration without recovery
• Stop-loss systems confirming exits below average entry price
• Long-equity allocation systems persistently underperforming cash

• vs. Slow Bear: Trend Down may include sharp acceleration phases (e.g., late stages of Dotcom); Slow Bear specifically excludes them — it is the grinding-down version.
• vs. Sharp Crash: Trend Down persists across months; Sharp Crash is concentrated in days to weeks.

Sideways describes a market regime characterized by extended range-bound oscillation without decisive directional resolution.

• Total return: roughly −10% to +25% (around zero with mild dispersion)
• Maximum drawdown: typically −10% to −20% — pullbacks happen within the range
• Realized annualized volatility: moderate, 13–30%
• Persistence: at least 4 months without decisive directional resolution; price oscillates within bounded range
• Direction-to-vol ratio: net direction must be small relative to volatility — direction dominated by noise

• SPY China Sideways 2015 H2 — −7.7%, max DD −12%
• BTC Sideways 2023 — −1.6%, max DD −20%
• Gold Sideways 2014 — +8.2%, max DD −10%
• WTI Sideways 2018 — +23.9%, max DD −11% (upper-bound case, mild trend mixed in)

• Trend-following systems generating frequent false breakouts
• Wide stops repeatedly triggered around range edges
• Time-decay eroding options-based directional structures
• Trailing-stop momentum systems exiting at range floors
• Breakout-confirmation systems whipsawed in both directions

• vs. Whipsaw: Sideways captures the regime shape (range-bound); Whipsaw emphasizes the signal-noise problem for trend-followers (frequent reversals through any signal threshold).
• vs. Vol Compression: Vol Compression requires visibly suppressed realized volatility; Sideways allows moderate vol within the range.

Vol Expansion describes a market regime characterized by sustained elevation of realized volatility relative to the asset's recent baseline, decoupled from directional bias.

• Gating (required): Median realized volatility over the case window ≥ 1.5× the asset's normal-window vol
• Gating (required): Persistence — at least 2 distinct (non-overlapping) 30-day windows with vol ≥ 1.5× baseline; differentiates VOL_EXPANSION from a single concentrated SHARP_CRASH window
• Descriptive (not gating): peak rolling-30-day vol ≥ 3× baseline count; vol-of-vol level; max-DD-window characteristics — reported alongside, used for sub-classification ("sustained-elevated" vs "sustained-with-spikes") but not for FM-tag determination
• Direction: decoupled from regime — return can be sharply negative, sharply positive, or near zero
• VIX-specific: observable as VIX index level above its 90th percentile (separate criterion from realized-vol-of-VIX-returns)

Definition history: an earlier version required 2–5× baseline as a 6-month median plus ≥3 distinct windows with peaks ≥3×. Systematic anchor validation (2026-05) revealed that conjunction effectively required Lehman-class duration AND COVID-class peaks combined — rare even among real episodes. The current version separates the persistent-state core (gating) from spike behaviour (descriptive). Spike-character regimes are conceptually a separate axis (VOL_INSTABILITY, planned future failure-mode) rather than a sub-criterion of VOL_EXPANSION. Anchor re-tagging (Volmageddon → SHARP_CRASH; BTC ATH 2017 / DeFi Summer Bull → TREND_UP only) corrected mis-classifications where event-character did not match case-window measurement.

• SPY Vol Shock Feb 2018 — realized vol 17%
• QQQ Vol Shock Feb 2018 — realized vol 21%
• QQQ Tech Bear 2022 — realized vol 34%
• ETH Vol Shock 2018 — realized vol 105%
• BTC 2017 Parabolic ATH — realized vol 92%
• VIX Volmageddon 2018 — realized vol 289%
• VIX COVID 2020 — realized vol 187%

• Fixed-stop systems gapped through unfavorable prints
• Position-sizing models underestimating realized risk
• Mean-reversion entries fading against expanding moves
• Short-volatility carry structures facing uncapped losses
• Risk-parity allocations rebalancing into concentrated positions

• vs. Sharp Crash: Sharp Crash is direction (down) plus speed; Vol Expansion is direction-agnostic.
• vs. Whipsaw: Whipsaw emphasizes oscillation around levels; Vol Expansion emphasizes magnitude regardless of pattern.

Vol Compression describes a market regime characterized by sustained suppression of realized volatility well below the asset's recent baseline.

• Realized annualized volatility: 0.4–0.7× the asset's baseline (equities 7–12% vs. baseline 16%; gold 8–15%; crypto 30–50% vs. baseline 80%+)
• Persistence: typically >2 months of sustained low realized vol; brief calm windows do not qualify
• Direction: can be slightly positive, neutral, or VIX-specific (low VIX is the regime)
• Drawdowns: shallow, typically <8% for non-VIX assets

• SPY Bull Run 2017 H1 — realized vol 7.1%, max DD −3%
• QQQ Tech Rally 2017 — realized vol 10.1%, max DD −5%
• VIX Low Vol 2017 — VIX index level persistently below 12 (regime defined by the index level itself, not by the realized volatility of VIX returns — which remains structurally high)

Only two robust empirical anchors for non-VIX assets (SPY and QQQ 2017); the regime is otherwise inferred from synthetic profiles. Bullish low-vol periods cluster naturally — this category overlaps with Trend Up by construction.

• Breakout strategies generating few signals, mostly false
• Time-stop systems exiting before signal materializes
• Short-volatility systems accumulating without warning of regime change
• Momentum signals weak relative to noise floor
• Trend-followers underperforming passive allocation

• vs. Sideways: Sideways allows moderate volatility within the range; Vol Compression requires visibly suppressed vol.
• vs. Trend Up: Trend Up is about return direction; many Trend Up regimes are also Vol Compression (low-vol bull) but not all.

Sharp Crash describes a market regime characterized by accelerated downside dislocation over compressed trading windows.

• Gating (required): Peak-to-trough drawdown ≥ 20% in any rolling 30-day window. DD computed as max over all overlapping 30-day windows. Multiple ≥20% windows within 60 days consolidate into a single crash event.
• Gating (required): Realized volatility during the crash window ≥ 1.5× the asset's normal-window vol (asset-agnostic).
• Sub-classification (descriptive, not gating): "tail-intensified" (excess kurtosis ≥ 1.5 OR daily-return 1st percentile ≤ −3%) versus "broad-distribution" (otherwise).
• Reported alongside: vol percentile against asset's own rolling-vol distribution; min-window DD (5–10 days) for flash-event detection.
• Near-crash band: 15–20% rolling-30d DD is reported as "near-crash dynamics observed" without the SHARP_CRASH tag.
• Recovery within the case window is permitted — SHARP_CRASH classifies the dislocation event itself, not the post-event trajectory.

The definition captures observable price dynamics. Underlying microstructure effects (liquidity, gap-structure, order-book stress) are not directly modeled here — see LIQUIDITY_STRESS for that dimension. Anchor validation (Lehman 2008 rolling-30d DD −32%, COVID 2020 −34%, Luna 2022 −40%, FTX 2022 −26%) confirmed all four listed historical SHARP_CRASH anchors satisfy the ≥20% DD-component threshold with margin. Note: while the DD component is met by all four, the full gating (DD ≥20% AND crash-window vol ≥1.5× baseline) is asset-asymmetric — for high-baseline assets like BTC the vol component does not always activate even in clear stress events (e.g. Luna 2022 crash-window vol 0.92×, FTX 2022 0.85×). See /verifiability for the empirical-identifiability table. Definition revised 2026-05.

• Lehman Brothers collapse — Sep–Oct 2008
• COVID-19 onset — Feb–Mar 2020
• Luna / UST collapse — May 2022
• FTX collapse — Nov 2022 (secondary, exchange-failure shock)

• Signal lag exceeding downside acceleration
• Oversold re-entry during unresolved downside continuation
• Liquidity-gap stop execution at unfavorable prints
• Fixed volatility thresholds destabilized by expanding realized volatility
• Position-sizing assumptions invalidated by realized risk

• vs. Slow Bear: Sharp Crash compresses the drawdown into days or weeks; Slow Bear extends comparable losses over months without the same acceleration.
• vs. Vol Expansion: Vol Expansion can be directionally neutral; Sharp Crash combines directional dislocation with vol expansion.
• vs. Liquidity Stress: Liquidity Stress can occur without large directional moves; Sharp Crash usually carries liquidity stress as a co-symptom.

Slow Bear describes a market regime characterized by gradually accumulating downside losses without a single concentrated crash event.

• Total return: typically −25% to −60% over ~6 months (annualized roughly −50% to −95%)
• Maximum drawdown: −25% to −70%, accumulating gradually rather than concentrated
• Realized annualized volatility: 20–90% — usually elevated but without sharp single-day shocks dominating
• Persistence: drawdown sustains or deepens over 3+ months without recovery
• Distinguishing trait: drawdown growth is monotonic-ish, not punctuated by a single crash event

• SPY Fed-Bear 2022 H2 — 0% return, max DD −17% (slow profile)
• QQQ Tech Bear 2022 — −29.8%, max DD −32%
• WTI Saudi Oil War 2014 — −48%, max DD −53%
• BTC Crypto Winter 2018 — −52.5%, max DD −66%
• QQQ Dotcom Crash 2000 — −58.5%, max DD −67% (also Trend Down + Sharp Crash due to severity)

• Long-equity allocation underperforming cash for extended periods
• Trend-followers cycling in/out with each minor rally
• Mean-reversion bottom-fishing entries repeatedly stopping out
• Drawdown psychology favoring capitulation near terminal lows
• Volatility-targeting under-allocating during deep drawdowns when vol stays moderate

• vs. Trend Down: Trend Down may include sharp acceleration phases (e.g., late stages of Dotcom); Slow Bear specifically excludes them.
• vs. Sharp Crash: Sharp Crash concentrates loss in days to weeks; Slow Bear distributes loss across 3+ months.

V-Recovery describes a path pattern characterized by a meaningful drawdown immediately followed by a rapid retracement to near or above pre-drawdown levels.

• Maximum drawdown during the case: −15% to −35% (peak-to-trough)
• Recovery from low: at least 80% retracement of the drawdown by case-end (price returns to within ~5% of pre-crash peak)
• Time-from-low to substantial recovery: typically 2–4 months (sharp rebound, not slow)
• Final case-window return: typically flat to +25% (the V-shape lands somewhere near or above starting price)
• Realized annualized volatility: 20–50% during the V

• SPY COVID Crash 2020 — +0.7%, max DD −34%
• QQQ COVID + Tech V-Recovery — +19.5%, max DD −29%
• Gold COVID + ATH 2020 — +25%, max DD −12.5%

All three anchors are COVID 2020 across asset classes — this limits regime-diversity validation.

• Trend-following systems exiting near lows, missing the recovery rally
• Drawdown-based de-risking reducing exposure before snap-back
• Cash-on-the-sidelines systems waiting for confirmation that arrives late
• Risk-parity rebalancing dampening recovery participation
• Stop-loss execution at terminal lows preceding the rebound

• vs. Sharp Crash: Sharp Crash is the down-leg only; V-Recovery is the down-leg plus the rapid up-leg as a single regime.
• vs. Trend Up: V-Recovery starts after a meaningful drawdown; Trend Up starts from neutral ground.

Whipsaw describes a market regime characterized by repeated directional reversals at signal-relevant levels, generating false confirmation signals for trend-following systems.

• Multiple directional reversals: at least 3–5 sign-changes of >5% magnitude over 3–6 months
• Net case return: near zero, typically ±10% (no decisive direction)
• Maximum drawdown: typically −10% to −20% — pullbacks happen but recover
• Realized annualized volatility: moderate, 15–30%
• Distinguishing trait: false signals — strategies dependent on directional confirmation are stopped out repeatedly

• SPY China Sideways 2015 H2 — −7.7%, max DD −12%
• BTC Sideways 2023 — −1.6%, max DD −20%
• Gold Sideways 2014 — +8.2%, max DD −10%
• WTI Sideways 2018 — +23.9%, max DD −11%

Whipsaw shares its anchor set with Sideways by design — these modes are not differentiated by separate cases but by perspective: Sideways describes the regime shape, Whipsaw describes the failure of trend-following signals within that shape. Sign-change counts per case are not currently measured in our pipeline.

• Moving-average crossovers generating repeated false signals
• Breakout-confirmation systems stopped at range boundaries
• Trailing-stops triggered at noise levels
• Mean-reversion at extremes works, but with high signal cost
• Time-stop systems exiting just before signal materializes

• vs. Sideways: Sideways emphasizes the bounded range; Whipsaw emphasizes the failure of trend-following signals within that range.
• vs. Vol Expansion: Whipsaw has moderate volatility with reversals; Vol Expansion can be uni-directional with high volatility.

Liquidity Stress describes a market regime characterized by deteriorating execution conditions — widening spreads, gap-down opens, and reduced fill quality — typically co-occurring with significant directional drawdowns.

• Bid/ask spread widening: ≥3× the asset's baseline spread during stress windows
• Gap-down opens: 2 or more days with overnight gaps >3% within the case window
• Volatility: typically co-occurs with Vol Expansion (realized vol 50%+)
• Drawdown: typically deep in our anchor set — −25% to −80%
• Distinguishing trait: execution-slippage matters — strategies relying on fill-quality assumptions underperform their model

Spread and gap-open thresholds are indicative microstructure heuristics drawn from market-microstructure literature; they are not directly measured in our case pipeline. Liquidity-stress identification in our catalog relies on historical co-occurrence of these conditions during the named episodes.

• Gold Taper Tantrum 2013 — max DD −25%
• BTC FTX Collapse 2022 — max DD −26%
• SPY Lehman GFC 2008 — max DD −42%
• BTC Luna Collapse 2022 — max DD −55%
• QQQ Dotcom Crash 2000 — max DD −67%
• WTI Negative Oil 2020 — max DD −78.5% (canonical extreme: futures settled negative)

• Limit-order entries missing execution at expected prices
• Stop-loss orders executing at prints far from trigger level
• Scaling-in plans facing partial fills with worse cost basis
• Daily-rebalanced systems facing widening transaction costs
• Multi-leg structures facing execution slippage that breaks the model

• vs. Sharp Crash: Sharp Crash is about price magnitude; Liquidity Stress is about execution conditions — they often co-occur but are diagnostically different.
• vs. Vol Expansion: Vol Expansion is realized-volatility focused; Liquidity Stress focuses on the microstructure breakdown that accompanies it.