"The market's in a different regime now" is one of those phrases that sounds precise and usually isn't. Here's what it means when it's used carefully, and how a regime actually gets classified rather than just asserted.
The plain-English definition
A market regime is a persistent set of conditions — direction, volatility, and how long those have held — that make one style of trading work and another style fail. It's not a mood or a vibe. It's a classification with concrete inputs, the same way "winter" is a classification of weather conditions, not just a feeling that it's cold.
Three regimes come up constantly:
- Trending — price is moving persistently in one direction (up or down) with follow-through, not just noise.
- Ranging — price oscillates between a rough ceiling and floor without a sustained directional move.
- High-volatility — the size of price swings has expanded, regardless of direction, often around news or macro events.
A given market can be trending and high-volatility at once (a violent breakout), or ranging and low-volatility (a quiet summer chop) — regime isn't a single axis, it's usually described along at least two: direction and volatility.
What actually goes into a classification
A regime label isn't a guess — a systematic classifier looks at things like:
- Position relative to a moving average (e.g. is price above or below its 200-day average) — a first-pass proxy for direction.
- The moving average's own slope over a recent window — trending up, trending down, or flat.
- A volatility measure — Bollinger Band width, ATR (average true range), or realized volatility over a trailing window — to separate "quiet" from "turbulent."
- Persistence — how long the current condition has held, since a one-day spike isn't the same as a three-week shift.
Put together, a simple deterministic rule might say: price above a rising 200-day average with band width below a threshold = TRENDING_BULL / low-vol. Price oscillating around a flat average = RANGING. Band width above a threshold regardless of direction = HIGH_VOLATILITY. More sophisticated classifiers add live context — news flow, cross-asset signals, sentiment — on top of the pure price-action read.
Why the label matters more than most indicators
Most trading indicators (RSI, MACD, moving average crossovers) answer "what should I do right now." Regime classification answers a prior, more important question: "which of my strategies is even allowed to have an opinion right now." A dip-buying strategy validated in ranging markets has no business running in a confirmed downtrend just because its RSI trigger fired — the precondition for its edge isn't present.
The honest caveat
No regime classifier is ground truth. A deterministic, price-only classifier is a proxy — useful for consistent backtesting, but it will occasionally disagree with a classifier that also weighs live news and cross-market context, and both will occasionally be "wrong" in hindsight. The useful move isn't finding the perfect classifier; it's being explicit that regime tagging is a probabilistic read, tracking how it performs, and gating strategy eligibility on it rather than ignoring it.
That's the model Atradian runs on: every backtest is broken down by classified regime rather than blended into one number, and the live bot only allows a strategy to trade in the regimes it was actually validated against — visible in the product, not just claimed in a blog post.