Research · a claim, tested

Should you only trade when the market is busy?

It is the most repeated filter in this business: wait for volatility, do not trade a dead market. We can test it directly, because we publish a set of windows we think are worth watching. Do they work better when the market arrives busy? No — they work worse, in 14 of 14 markets, and neither state covers its costs.

55.9%a flagged window is right, after a big previous hour

58.0%after a quiet one

-2.03points, the cost of waiting for the busy state

0.80× / 0.71×what each state is worth against one round trip

What “busy” has to mean

The whole test turns on this. If you split hours into busy and quiet by their raw size, you have not measured a regime — you have rediscovered the London session, because 13:00 is bigger than 03:00 in every market and every week. The split would just be a clock wearing a different name.

So every hour is divided by the median of its OWN weekday × hour bucket first. A 03:00 that ran twice what 03:00 usually runs is busy; a 13:00 that ran half what 13:00 usually runs is quiet. The question becomes whether an hour that was big for itself is followed by a window that behaves better — which is the question the advice is actually making a claim about.

It points the wrong way

Across 77,714 flagged-window hours that followed a big hour and 76,738 that followed a quiet one, the flag is right 55.9% of the time in the busy state against 58.0% in the quiet one. That is -2.03 points the wrong way, and 14 of 14 markets lean the same direction. The strongest is NZD/USD at -3.84.

⚠ The control is what stops this being an artefact. Run the identical split on the windows we do not flag — 1,367,570 hours of ordinary buckets — and the gap is -0.18 points. Essentially nothing. So this is not "every window is worse after a big hour"; it is specific to the windows we publish, which is the claim being tested.

Market	Flagged windows	After a big hour	After a quiet hour	Δ	p	Busy: vs round trip	Quiet: vs round trip
NZD/USD	13	55.7%	59.5%	-3.8	0	0.57×	0.54×
Silver (XAG/USD)	37	55.2%	58.3%	-3.0	0	0.40×	0.43×
USD/CHF	16	58.2%	61.2%	-3.0	0.007	0.94×	0.95×
GBP/JPY	8	56.0%	58.8%	-2.8	0.025	0.87×	1.06×
EUR/GBP	16	61.7%	63.8%	-2.2	0.092	1.00×	0.83×
EUR/USD	13	54.7%	56.6%	-1.9	0.028	0.89×	0.96×
AUD/USD	13	54.9%	56.8%	-1.9	0.022	0.93×	0.81×
USD/JPY	10	56.3%	58.0%	-1.7	0.15	1.57×	1.21×
Gold (XAU/USD)	19	55.4%	56.4%	-1.0	0.177	0.58×	0.42×
USD/CAD	16	56.1%	57.0%	-1.0	0.19	0.89×	0.76×
EUR/JPY	9	55.4%	56.1%	-0.8	0.522	1.25×	0.98×
BTC/USD	3	58.1%	58.8%	-0.7	0.835	1.60×	0.90×
GBP/USD	10	55.9%	56.3%	-0.4	0.677	1.01×	0.68×
ETH/USD	2	58.3%	58.3%	-0.1	1	1.37×	0.46×

The last two columns are the ones that end the discussion: a value under 1.00× means the average outcome does not cover a single round trip in that market.

Where the difference actually lives

Split the same test by the kind of hour and nearly all of it is in the thin band — the Sunday reopen and the hours from 20:00 UTC, where the spread is widest and this site already tells you to be careful. There the gap is -3.29 points. In ordinary hours it is -0.56, which is close to nothing.

That is worth saying plainly: the busy-market filter is not a filter, it is a restatement of the thin-band warning. What looks like a regime effect is mostly the cost of trading at the hours nobody is at their desk.

And now the part that makes it interesting

Point the same regime split at how far the market moves rather than which way it closes, and it reverses. After an hour that ran twice its own bucket median, the next hour runs 1.47× its usual size — and after a quiet hour, 0.82×. That effect is real in every market we hold and worth a round trip in 17 of them.

So the two halves of this product answer to opposite regimes. Direction is marginally better when the market is quiet. Size is decisively better when it is busy. Using one as a filter for the other is the mistake — and “wait for volatility” is exactly that mistake, borrowing a true fact about movement to justify a claim about direction.

⚠ And the clustering does not decay: it is 1.47× at one hour and still 1.35× at twenty-four. Something that does not fade is a state, not a reaction — so it describes the weather the market is in, not a signal to act on.

What we would actually say

Do not use volatility as a filter on a directional window. It costs 2.03 points on our own flagged set, and the 0.18-point control says that is about these windows rather than about all hours.
Neither state pays anyway. 0.80× and 0.71× of a round trip — this is a line about where not to bother, not a setting to tune.
Check whether a “regime” is just a clock. Any busy/quiet split done on raw size rediscovers the trading sessions. Normalise by the bucket's own median first, or you are testing what time it is.
The advice is not baseless — it is aimed at the wrong axis. Busy hours really are bigger hours, which matters for how far price travels and what a stop costs. It just says nothing useful about which way the hour closes.

Full method in section 16 of the evidence page, the clustering result in the size sections, and every claim we have tested on the reports index.

More on how we decide what counts

What spread kills a seasonal edge?The hour's MOVE beats a round trip in 471 of 471 liquid windows. The directional EDGE beats it in 4.Does a hot streak mean anything?A streak is as predictive at four weeks' lag as at one. Memory decays; drift does not — which is why there is no streak feature on this site.Do weekend gaps fill?"Gaps fill 90% of the time" is true and empty — a matched midweek move fills just as often. The indices do the opposite.Does yesterday's high or low matter?68,688 breaks across 19 markets. The level holds 50.4% of the time — slightly less than an arbitrary level the same distance away.

Every claim we have tested, including the ones that held, is on the reports index.

We publish the tests that failed as loudly as the ones that held. Every number on this site carries its sample size, and every claim carries what it was measured against.

Read the whole evidence page →