trend-following
Lambda-HMM tracks Λ_F — the geometric instability of cross-asset rotation — and R_t — composite behavioral cascade conditions — in the same view. When both light up, the regime is critical.
Crashes require Q4 (high Λ_F and high R_t). Q3 alone is rotation; Q2 alone is crowded-but-stable.
Last refresh: 2026-05-12 22:01 UTC · Next refresh: 2026-05-13 22:00 UTC · Pipeline: ● OK
Trend-following regime. Λ_F is below the structural-instability threshold and R_t is benign. The default condition.
Two operating points on the precision/recall curve. The factor-model (top) is qsurf's canonical methodology — derived CMKT/CSMB/CMOM/CVOL covariance commutator; selective; correctly ignored the May 2021 mid-cycle dip; 4/4 on the IS-window 2015-2021 walk-forward backtest. The breadth signal (bottom; vonlambda's Method C) skips factor construction and uses standardized returns directly; more sensitive (catches more events, with more false positives) and used by vonlambda for cross-market generalization. The headline crypto regime call on this page comes from the factor-model, not the breadth signal — see SPEC v1.8 Path B.
Per-day Λ_F percentile, R_t composite, and quadrant call — the prospective walk-forward record as it accumulates.
Same breadth signal (vonlambda's Method C — standardized-returns covariance, expanding-window percentile) applied per market: commodities, equities, bonds, crypto, FX baskets. Aggregated into a 0–30 systemic score (0 = all stable, 30 = all critical). Use this to see whether qsurf's headline crypto call coheres with or diverges from the broader cross-asset regime. Note: this panel is a breadth-tuned diagnostic, designed for sensitivity across many asset classes; per-asset hit-rates vary. The headline crypto regime call on this page comes from qsurf's canonical factor-model (above), not from the breadth signal.
| Market | Λ_F % | Regime | As of | Universe |
|---|---|---|---|---|
| Commodities | P73.5 | NEUTRAL | 2026-05-11 | GLD, SLV, USO, UNG, CPER, DBA |
| Gold | P90.9 | CRITICAL | 2026-05-11 | GLD, SLV, GDX, GDXJ, UUP, TLT |
| Silver | P76.9 | ELEVATED | 2026-05-11 | SLV, GLD, SIL, DBB, UUP, TIP |
| Crypto (BTC) | P72.4 | NEUTRAL | 2026-05-11 | BTC-USD, ETH-USD, LTC-USD, XRP-USD |
| Ethereum | P68.4 | NEUTRAL | 2026-05-11 | ETH-USD, BTC-USD, LTC-USD, XRP-USD, BNB-USD, ADA-USD |
| US Equity (SPY) | P25.1 | STABLE | 2026-05-11 | XLF, XLK, XLE, XLV, XLI, XLP, XLY, XLB |
| UK Equity | P74.7 | NEUTRAL | 2026-05-11 | EWU, FXB, EWUS, EWL |
| Germany | P72.7 | NEUTRAL | 2026-05-11 | EWG, EWL, EWN, EWO, FXE |
| Bonds | P91.0 | CRITICAL | 2026-05-11 | SHY, IEF, TLT, LQD, HYG, TIP |
| Emerging Markets | P64.5 | NEUTRAL | 2026-05-11 | EEM, FXI, EWZ, EWW, EWY |
Same factor model applied to the SPEC §2 11-sector ETF universe. Independent signal — sector rotation regimes vs crypto regime hops.
Universe: XLF · XLK · XLE · XLV · XLI · XLY · XLP · XLU · XLRE · XLB · XLC
SPEC §5.4 R1 · every walk-forward refit canonicalizes HMM states via max-overlap; overlap-agreement audited per refit. Diagnostic ≥ 0.90; fallback floor ≥ 0.80.
Window: 2022-01-03 → 2026-04-10 (1071 trading days, T_refit=21) · Methodology version frozen at parameter-freeze date · Audit refreshes when calibration sub-sprints land new walk-forward runs.
Each time the diagnostic crosses into ELEVATED/CRITICAL territory (Λ_F percentile ≥75) an episode opens. After 90 days the episode resolves to one of TRUE_POSITIVE (a Q4 call followed by a ≥10% drawdown), PARTIAL (Q3 call + drawdown), FALSE_POSITIVE (no drawdown), or EXPIRED (insufficient forward data). Outcomes are written back here automatically — alerts are not unrecorded.
| Episode | Opened | Max regime | Max Λ_F % | Status | Outcome | Drawdown |
|---|
Showing 10 most-recent episodes (open + resolved). Trigger: Λ_F percentile ≥ 75. Forward window: 90 days. Correction threshold: ≥10% drawdown on the cross-asset average close from the open-date anchor.
Q4-CRITICAL calls vs realized 60-day forward drawdowns ≤ -10%. Floor for §10 G5: 50%.
Prospective record begins accumulating from launch.
Open formula, open spec, open audit log. The retrospective fit is a sanity check; the prospective record is the evidence stream.
The diagnostic detects when the geometric structure of the asset-rotation matrix becomes unstable — when the principal factors begin to rotate, fold, or collapse onto each other. Retrospective validation shows strong signal on institutional rotation events (Nov 2021 top, Oct 2025 top, Dec 2018 Fed Panic) and correctly silent on negative-control mid-cycle dips (May 2021, Jun 2022 Luna aftermath). The diagnostic is materially weaker on pure exogenous shocks without an institutional precursor — see the research / validation page for the full event ledger and documented limitations. Forward tracking is ongoing.
R_t is a behavioral-cascade composite of cross-asset fear, crypto reflexivity, macro stress, and risk-off proxies — VIX (0.40) + BTC volume |z| (0.30) + yield curve (0.20) + DXY momentum (0.10). Hysteresis bands (enter ≥65 / exit ≤55) and 5d EMA smoothing prevent whipsaw.
Subject of US provisional patent application 63/903,809 (R. J. Mathews, filed October 22, 2025). Full backtest specification and the prospective walk-forward record are published as a pre-registered research document on the launch date.
A · B ≠ B · A. Routine in matrix algebra; the degree of non-commutativity is what the diagnostic measures.[A, B]A·B − B·A. Zero iff the operations commute; non-zero entries quantify how much "order matters." Λ_F uses the commutator between the rolling factor covariance and its time derivative.‖·‖_FF_t — factor covariance matrixt.dF_t/dt — time derivativeF_t rotates rapidly (institutional rebalancing); near zero when the structure is stable.R_t — reflexivity compositeTwo diagnostics, one honest disclosure. The methodology has documented limits — they're as load-bearing as the strengths.
The covariance commutator [F_t, dF_t/dt] picks up when the principal factors of the asset-rotation matrix begin to turn over. Empirically this fires before institutional positioning changes show up in price — a structural-instability early warning, not a trend signal.
VIX percentile + crypto-volume |z| + yield-curve stress + DXY momentum form a composite reflexivity score. Q4 (the only "crash setup" call) requires BOTH Λ_F high AND R_t high — structure unstable and crowd vulnerable. Either alone is rotation or fragility, not cascade.
COVID-March 2020, Terra/Luna collapse, 3AC unwind, FTX implosion — all of these moved markets faster than the institutional-rotation cycle the diagnostic tracks. Documented as known limitations, not retroactively explained-away. The diagnostic is for positioning regimes, not news shocks. See the research page for the full event ledger including failures.
Daily regime calls run live (Cloud Scheduler at 22:00 UTC). The R_t reflexivity composite is v1.7 calibration; refinements ship in v1.8 alongside the public OOS report at the 2026-06-05 cutover.
Direct research questions: mail.rjmathews@gmail.com. No payment is collected pre-launch; waitlist signups are notified when each tier opens.