Market regimes explained in plain English
A market regime is the general character of a market over a period of time. It isn't about one day's price. It describes how prices have been behaving across weeks or months.
Most people group regimes into three types.
| Regime | What prices do | Approaches usually suited to it |
|---|---|---|
| Trending | Keep moving one way, up or down, with fairly shallow pullbacks | Following the trend, often called momentum |
| Ranging | Bounce between a rough floor and ceiling without breaking out | Buying near the floor and selling near the ceiling, called mean reversion |
| Volatile | Swing hard in both directions, often around news or stress | Smaller positions and tighter risk limits |
The types can overlap. A market can trend upwards and be highly volatile at the same time.
Why systematic models try to identify regimes
A systematic trading model is a set of rules applied consistently, trade after trade. The trouble is that no single rule suits every market.
Rules that buy strength do well while a trend continues. In a choppy range they get whipsawed, buying near the top and selling near the bottom. Rules that buy weakness do the opposite: fine in a range, painful in a strong trend.
Volatility matters for a different reason. The same position carries more risk in a wild market than in a calm one. A model that knows volatility has risen can cut the size of its positions to keep risk roughly steady.
So identifying the regime first lets a model decide which rules to trust and how much to risk.
How models measure the market regime
There is no single test for a regime. Models usually combine several measures:
- Moving averages and their slope, to show the direction of travel
- Trend-strength indicators, such as the Average Directional Index, known as ADX
- Volatility measures, such as the standard deviation of returns or the average true range
- Statistical and machine-learning methods, such as hidden Markov models or clustering, which group past periods with similar behaviour and estimate which group today looks most like
The catch is that regimes are clearest in hindsight. A model only sees a change after it has started, so it is always estimating, never certain. Good systems are built to expect that and to change their minds when the evidence does.
There is a second trap. A model with too many moving parts can fit past data almost perfectly and then fail on new data, a problem known as overfitting. Simpler measures are often easier to test and to trust.
How iQuant Fund approached it
I was one of the co-founders of iQuant Fund, set up in February 2021 alongside Jason Noorman and other partners, and I served as its Chairman. It started with traders who wanted their own strategies to run automatically.
Its models used two tools together. One was machine learning. The other was Modern Portfolio Theory, which weighs risk and return across a whole portfolio rather than one holding at a time. The outputs included momentum signals, a reading of which regime a market was in, and diagnostics on whole portfolios. Coverage ran to more than 20 exchanges and roughly 5,000 listed shares. The early focus was the ASX, before coverage grew to include NASDAQ, the NYSE and Hong Kong.
As a member of both AWS Activate and Nvidia Inception, it had access to cloud computing and GPU capacity for training its models. Kanebridge ran a feature on the business, which is listed on my press coverage page.
As with my other companies, I don't write the code. Jason Noorman leads the technology. My role was building the business around it.
iQuant's clients were institutions only: family offices, funds and professional trading teams. It did not serve retail investors, and I make no claims here about its performance. The iQuant Fund page has the company's background, and the business is at iquant.fund.
Where my interest in market models started
Before iQuant, I co-founded Alan Masters with partners in Hong Kong. It sold subscriptions to technical analysis of crypto trading pairs, and I ran it from 2016 to 2017.
That work got me interested in a simple question about trading decisions: how much of the judgement could software handle consistently, without the fatigue and second-guessing people bring to it. Market regimes are a good example. A person can sense that a market feels different. A model can measure it the same way every time.
I see the same idea at work in property. Kinnara, the marketplace I founded, uses software to do work people would otherwise do slowly by hand, which I describe in one search across 76 portals.
This note is general information only. It is not financial advice. iQuant Fund served institutional clients only.