SignalTrace

Live Regime-Model Evidence Gap

John Montgomery5 min read

Live Regime-Model Evidence Gap

The Live Regime-Model Evidence Gap is this vault’s name for a structural asymmetry in the public record. It separates two claims that are often conflated: (a) that real money managers use regime classification as a risk-management and asset-allocation input — which is real and openly disclosed by named firms; and (b) that a Markov chain, hidden Markov model or regime-switching model functions as a standalone, profitable live trading system — for which there is essentially no credible public evidence, despite the claim being common in marketing and trading-blog material. Resolving whether the standalone-profit claim has live substantiation is the single most important open question in the vault, and the honest answer from public disclosure is no, not substantiated — while acknowledging that fund secrecy means this is an absence-of-evidence finding, not proof of absence.

The evidence for claim (a) is strong and named. Bridgewater Associates’ official firm history The All Weather Story discloses an economic-environment “four boxes” framework (growth × inflation, each rising/falling) as the foundation of its flagship strategy — but explicitly as a passive, non-predictive portfolio: “this was the best portfolio Ray and his close associates could build without any requirement to predict future conditions.” State Street’s Investment Solutions Group has published its Market Regime Indicator (MRI) quarterly since at least 2012: a 0–100% risk-aversion classifier built from implied equity/currency volatility and fixed-income spreads, sorting markets into five regimes (Crisis, High Risk Aversion, Normal, Low Risk Aversion, Euphoria), used as “one of its inputs into ISG’s global tactical asset allocation decision-making process.” BlackRock’s Investment Institute frames a “new regime” requiring more dynamic allocation, and BlackRock Systematic discloses macro machine-learning models designed to “adapt better to regime shifts.” Index and data vendors offer comparable tools (MSCI and Bloomberg regime/scenario analytics; the Financial Turbulence Index). The pattern is unambiguous: regime classification is a confirmed, openly described component of live institutional processes.

The evidence for claim (b) is the gap. Across these disclosures, regime models appear as one input among many to portfolio construction or risk control — never as a documented standalone strategy with an audited, out-of-sample, post-cost live track record. Bridgewater’s regime product is deliberately passive; State Street’s MRI is an allocation input whose validation language (“over twelve months of rigorous testing … tracked historical market stress events”) is a back-test, not a live performance audit; BlackRock’s most detailed regime-model disclosure attaches the standard hindsight back-test disclaimer (“Unlike actual performance returns, they do not reflect actual trading, liquidity constraints, fees and other costs … It is not likely that similar results could be achieved in the future”). Crucially, AQR Capital Management — a major quant manager — publishes the opposite of a timing endorsement: Asness, Ilmanen & Maloney’s “Market Timing: Sin a Little” and Asness’s “The Siren Song of Factor Timing” argue that timing the market or factors is “very difficult to do well,” that the accurate answer to “can you time these?” is “mostly no,” and that one should at most “sin a little.” A leading systematic firm thus disclosing skepticism toward regime-style timing is itself evidence against the standalone-profit claim.

This gap is the live-trading counterpart of the vault’s academic findings. The peer-reviewed and working-paper literature (Bulla et al. 2010, Shu Yu and Mulvey 2024, Zakamulin 2016, Dacco and Satchell 1999) consistently shows that regime models deliver risk reduction — lower volatility and shallower drawdowns — far more reliably than after-cost directional alpha, and that apparent alpha often collapses once look-ahead bias is removed and costs applied. The public conduct of named live managers matches that academic verdict exactly: they deploy regime classification for de-risking and allocation, not as a published alpha engine. Two cautions keep the verdict honest. First, secrecy cuts both ways: profitable proprietary signals are precisely what a hedge fund would never disclose, so Two Sigma, Man Group or others could run internal Markov/HMM components — the gap is an absence of public evidence. Second, and decisively, the burden of proof sits with the claim: no credible public live track record substantiates a standalone profitable Markov/HMM trading system, so the marketing-grade version of that claim should be treated as unproven until such evidence appears.

Live Regime-Model Evidence Gap [relates] Regime Classification Bridgewater Associates [supports] Live Regime-Model Evidence Gap State Street Associates [supports] Live Regime-Model Evidence Gap BlackRock [supports] Live Regime-Model Evidence Gap AQR Capital Management [supports] Live Regime-Model Evidence Gap Zakamulin 2016 [supports] Live Regime-Model Evidence Gap Live Regime-Model Evidence Gap [contradicts] Markov Chain Trading Model

Connections

Sources

Graph View

Backlinks