Volatility Clustering: Does a Wide Day Predict the Next One?
Volatility is not scattered randomly across days: quiet stretches run in series and so do violent ones. We tested this on data — split days into quartiles by range and looked at what the following day turned out to be.
Volatility clustering: how the test was set up
For each pair we took daily ranges, marked the top and bottom quartile by size, then computed the median range of the day immediately following a wide one and immediately following a narrow one. No indicators: only the actual ranges of adjacent days are compared.
The measurement
- Ratio of the range after a wide day to the range after a narrow one: median 1.38
- The effect held on 12 pairs out of 12
- Weakest: GBPUSD — 1.14 (79 pips after a wide day against 69)
- Strongest: EURCHF — 1.87 (50 pips after a wide day against 27)
Measured across 518 daily bars for each of 12 currency pairs. The sample is public, the methodology is in how we run backtests, and the calculation is reproduced by a script in the repository.
The result holds on every pair
A ratio above one came out on all twelve pairs without exception, with a median of 1.38. The effect is weakest on GBPUSD (1.14) and strongest on EURCHF (1.87). That consistency sets this measurement apart from most checks in our section: this is not a faint regularity at the edge of noise but an effect visible in every instrument.
What it means in practice
Volatility is predictable and direction is not — and those are not the same thing. Knowing that tomorrow's range is likely to be wide says nothing about where price will go. It can be used in two places: stop distance and position size, but not in choosing direction.
How to apply it
Tie the stop to current volatility rather than to a fixed number of pips, and recompute size after the stop changes. This property is the whole basis of average-true-range stops: more on the ATR stop.
What the measurement does not show
It does not say how long a cluster lasts: two days, a week or a month. It shows no direction and gives no trading signal. And it says nothing about what happens at a regime change — which is precisely where clusters break.
This material is for educational purposes and is not individual investment advice.
Frequently asked questions
What is volatility clustering?
A property of markets in which quiet and violent periods run in series rather than being randomly mixed. In our measurement the median range after a wide day was 1.38 times that after a narrow one.
Can you build a trading signal on it?
No. Clustering concerns range, not direction. It is useful for stop distance and position size but does not answer where price will go.
How robust is the effect?
It held on all 12 pairs in the sample, which distinguishes it from most regularities we have tested. But the measurement does not show how long a cluster lasts or what happens at a regime change.
New backtest studies — to your inbox
When a new reproducible study comes out or the data is updated, we send a short email. No spam, no "signals", no selling your address.
By leaving your email you agree to receive occasional emails from the portal. Unsubscribe in one click from any of them.