Market Regime State Detector

Last reviewed: May 17, 2026 by Options Analysis Suite Research.

Automated market regime classification system that reads the options market's vital signs daily. The core idea: options pricing models encode different assumptions about how markets behave. Diffusion models assume smooth volatility, jump models assume sudden crashes, surface models capture term structure. By calibrating all of them simultaneously and comparing how well each fits the observed market data, the system detects when market microstructure is shifting before it shows up in price action alone. Available with a Professional subscription.

Unlike the Options Market Scanner which takes a cross-sectional snapshot ("which tickers look interesting today?"), the regime detector takes a longitudinal view, tracking each symbol's own calibration history over time to detect state transitions. It operates at two levels: a systemic MARKET aggregate (SPY 50%, QQQ 30%, IWM 15%, DIA 5%) for overall market health, and 124 individual symbols across 12 scopes (sector, industry, bellwether, fixed income, macro, commodities, international, crypto, metals, factor, and thematic) for granular regime detection at the symbol level.

Pipeline

Every trading day, the pipeline calibrates 8 options pricing models across 124 symbols spanning 12 scopes (market, sector, industry, bellwether, fixed income, macro, commodities, international, crypto, metals, factor, and thematic) using end-of-day professional-grade options data. Intraday scans run at 5 intervals throughout the trading day (open, morning, midday, afternoon, pre-close) using live options feeds. For each symbol, it fetches the full options chain, filters by delta range (0.05 to 0.95) and bid-ask spread quality, groups into tenor buckets (7, 30, 90 DTE), and calibrates each model to minimize IV RMSE against observed implied volatilities. Seven models use target tenor buckets; eSSVI uses full expiry surface data with structure guards. The calibrated parameters and fit errors are stored daily, building a rolling history that enables z-score normalization of the extracted features.

8 Calibrated Models

The pricing frameworks are grouped into two families for regime detection. The smooth volatility family captures the volatility smile, skew, and term structure through continuous dynamics and surface parameterizations, without jumps. The jump/tail family adds discontinuous moves (crashes, gaps) on top of diffusion dynamics. When jump/tail models suddenly fit the observed market much better than smooth volatility models alone, it signals the market is pricing in discontinuous risk, a key regime shift indicator.

Baseline Reference

• Black-Scholes: Geometric Brownian motion with constant volatility. Serves as the baseline reference, excluded from both families and from model disagreement calculations. Provides the anchor for measuring how much additional complexity the market requires.

Smooth Volatility Family (Heston, SABR, eSSVI)

• Heston: Stochastic volatility with mean-reverting variance process. Captures the volatility smile through correlation between spot and variance (rho) and vol-of-vol (sigma_v). Day-over-day changes in rho and sigma_v are tracked as turbulence indicators.
• SABR: Stochastic Alpha Beta Rho model for volatility smile dynamics. Captures skew through the beta parameter and smile curvature through nu (vol-of-vol). Day-over-day nu changes contribute to the turbulence signal.
• eSSVI: Extended Surface SVI, which fits a full (strike x expiry) volatility surface simultaneously rather than per-expiry slices. Requires at least 30 IV points across at least 3 tenors. Reports IV RMSE only (no price RMSE) since it parameterizes the surface directly.

Jump/Tail Family (VG, Merton, Kou, Bates)

• Variance Gamma: Pure-jump Levy process with no diffusion component. Three parameters (sigma, theta, nu) capture volatility, skew, and kurtosis independently. Fits well when the market is pricing heavy tails and asymmetric returns.
• Merton Jump-Diffusion: Black-Scholes diffusion plus log-normally distributed jumps at Poisson arrival rate lambda. When lambda calibrates high, the market is pricing in crash risk. Day-over-day lambda changes are a key turbulence input.
• Kou Double-Exponential: Diffusion plus asymmetric exponential jump sizes, with separate parameters for upward and downward jumps. Captures put-call asymmetry in jump expectations. Seeded from Merton calibration for faster convergence.
• Bates: Heston stochastic volatility plus Merton-style jumps. Captures both smooth vol dynamics and discontinuous moves simultaneously. When Bates significantly outperforms Heston alone, it confirms jump risk is being priced.

8-Feature Regime Vector

From the calibration results and raw IV data, the system extracts 8 features that capture different dimensions of market stress. Each feature is z-scored against its own 60-day rolling history using robust statistics (median + scaled MAD) to be resistant to outliers. The composite stress score is a weighted sum of these z-scores.

• Volatility Level (12%): Average ATM implied volatility across tenors. Higher ATM IV equals higher uncertainty equals more stress.
• Skew Pressure (13%): 25-delta put IV minus ATM IV at the ~30-day tenor, amplified by SABR model stress. SABR specifically models skew through its spot-vol correlation parameter (rho). When SABR cannot fit the observed skew (RMSE exceeds 0.10), the raw skew signal is amplified proportionally, confirming the skew pressure is genuinely extreme, not just noisy. No amplification when SABR fits well.
• Curvature (10%): 10-delta put IV minus 25-delta put IV at ~30 days, amplified by Heston model stress. Heston models smile curvature through its vol-of-vol parameter (sigma_v). When Heston cannot explain the observed curvature (RMSE exceeds 0.10), the raw curvature signal is amplified, confirming the tails being priced in are genuinely extreme. No amplification when Heston fits well.
• Term Structure (10%): (Short ATM IV minus Long ATM IV) / Long ATM IV. Positive values indicate an inverted term structure (backwardation) where near-term fear exceeds long-term expectations, a classic stress signal.
• Turbulence (13%): Day-over-day changes in key calibrated parameters (Heston sigma_v, Heston rho, SABR nu, Merton lambda), individually z-scored and averaged. High turbulence means the models' view of market dynamics is changing rapidly.
• Tail Dominance (12%): min(Heston, SABR, eSSVI IV RMSE) minus min(VG, Merton, Kou, Bates IV RMSE). Positive values mean jump/tail models fit the market better than smooth volatility models alone, indicating the market is pricing in discontinuous risk. Black-Scholes is excluded from both sides.
• Model Disagreement (20%): Vega-weighted cross-model IV standard deviation, normalized by ATM IV. When calibrated models disagree significantly on fair implied volatility for the same strikes, it signals structural uncertainty: the market doesn't fit cleanly into any single model's framework. Highest weight because it captures regime ambiguity directly.
• Surface Complexity (10%): eSSVI IV RMSE, which measures how far the market's implied volatility surface departs from a structured parametric surface. Unlike BS RMSE (which measures departure from flat vol), eSSVI captures residual complexity after accounting for skew, smile, and term structure. High eSSVI RMSE (including values above the fallback threshold) is itself informative, signaling the surface is too complex for even a structured surface model. Falls back to scaled BS RMSE only when eSSVI calibration is skipped due to insufficient data.

Classification and Hysteresis

The weighted composite score is classified into 5 states using z-score thresholds. To prevent noisy regime flipping during transitional periods, the system uses asymmetric enter/exit thresholds: quick to escalate, slow to de-escalate.

• CRISIS (enter at or above 2.5 sigma, exit below 2.0 sigma): Extreme stress, models breaking down.
• STRESS (enter at or above 1.5 sigma, exit below 1.0 sigma): Significant stress, jump models dominating.
• ELEVATED (enter at or above 0.5 sigma, exit below 0.0 sigma): Above-normal stress, hedging demand rising.
• NORMAL (enter at or above -0.5 sigma, exit below -1.0 sigma): Typical market conditions.
• CALM (below NORMAL): Below-average stress, low volatility environment.

Symbol Universe

124 symbols across 12 scopes provide broad cross-asset market coverage. The MARKET aggregate is a weighted composite (SPY 50%, QQQ 30%, IWM 15%, DIA 5%) that represents the overall market state.

• Market ETFs (4): SPY, QQQ, IWM, DIA.
• Sector ETFs (11): XLF, XLK, XLE, XLI, XLY, XLP, XLV, XLU, XLB, XLC, IYR.
• Industry ETFs (17): SMH, KRE, XHB, XBI, XRT, IGV, OIH, XME, GDX, ITB, URA, SOXX, IBB, KBE, XOP, JETS, VNQ.
• Bellwether Stocks (62): AAPL, MSFT, NVDA, AMZN, GOOG, META, JPM, GS, V, BAC, MA, XOM, CVX, UNH, LLY, ABBV, TSLA, HD, WMT, COST, CAT, BA, UBER, LMT, RTX, CRM, ORCL, NOW, ADBE, NFLX, DIS, AMD, AVGO, MU, COIN, PLTR, CRWD, PANW, ABNB, SHOP, BKNG, NEE, FCX, PLD, PG, PYPL, KO, MRK, INTC, T, VZ, PFE, SNOW, AMAT, QCOM, BX, XYZ, DDOG, HOOD, SMCI, APP, DASH.
• Fixed Income (4): TLT, IEF, HYG, LQD.
• Macro / Volatility (2): VXX, VIXY.
• Commodities (5): GLD, SLV, USO, UNG, GDXJ.
• International (8): EEM, EFA, FXI, EWZ, EWJ, EWG, EWY, MCHI.
• Crypto (1): IBIT.
• Metals and Mining (1): SIL.
• Factor ETFs (7): RSP, MTUM, VUG, MDY, IWF, SCHD, IJR.
• Thematic ETFs (2): TAN, KWEB.

Dashboard

• Stress Score History: Area chart of the composite stress score over time with color-coded regime bands, threshold reference lines, and a tooltip showing the hysteresis-aware regime label at each date.
• Feature Vector Display: Horizontal bar chart breaking down the current regime into its 8 constituent signals, each shown as a z-score with color coding by severity.
• Symbol Breakdown: Expandable tables for each scope showing per-symbol regime labels, stress scores, confidence levels, and top drivers. Click any symbol to view its model calibration details.
• Model Fit Quality: Per-symbol calibration results for all 8 models showing IV RMSE, price RMSE, convergence status, runtime, and fallback indicators.
• Landing Page Card: Compact regime summary card on the main landing page showing current state, stress score, top driver, and a 90-day interactive sparkline for at-a-glance market health monitoring.

This page is part of the Options Analysis Suite features overview. Browse the full documentation.