// RESEARCH

Methodology, in the open.

Regime studies, backtests, and single-asset spotlights from the engine that scores the market nightly. Everything here is an observation of data and historical analogues — not a rating, target, or buy/sell recommendation. The decisions stay yours.

Backtest research · Flagship
Backtesting a regime system across 1,225 assets for 41 months
The current flagship study: probabilistic HMM regime detection tested out-of-sample from January 2023 through May 2026 across 1,225 assets in 17 markets, with Monte Carlo, Deflated Sharpe and cost-sensitivity checks alongside the headline figures.
23 May 2026 · on MediumRead →
Spotlight · Mega-cap tech
GOOGL at $3.83T: +41% in three months, and the OBV-vs-price divergence into I/O
A data observation on Alphabet: an OBV-vs-price divergence into Google I/O 2026, the HMM regime read, key levels, and three scenarios between now and the Q2 print.
20 May 2026Read →
Backtest research
Testing Hidden Markov Models across 530 assets
The origin-story study: a hand-tuned HMM regime classifier backtested across 530 global assets with zero look-ahead bias. What win rate does and does not mean — and the larger 1,225-asset study that now supersedes it.
Originally on MediumRead →
Strategy insight
Trading smarter: half the drawdown, double the returns
Regime-based exposure versus buy-and-hold across multiple cycles. In the tested window, switching on the HMM read cut maximum drawdown by roughly half — the more robust takeaway than any headline return multiple.
Originally on MediumRead →
Methodology deep-dive
A Hidden Markov Model analysis for gold
What “hidden states” really are, why you can never observe them directly, and how the mathematical structure of candlestick data surfaces market regimes — demonstrated on gold's price history.
Originally on MediumRead →

// Every study here uses the same regime engine that scores 1,230+ assets every night across nine dimensions — HMM regime, Technical, Trend, Support/Resistance, Fundamental, Sector Edge, Sentiment, Momentum, and Pattern. Backtests are out-of-sample and documented; the full trade list — today's entries and exits plus the year-to-date table — is published in the app on a free account.