Every hour tested separately, everything else held constant. Only 11 of the 24 hours finished in front — and the gap between the best and the worst is far larger than the gap between most strategies.

By session

Session	Hours (UTC)	Trades	Win rate	Net pips
Asia	00:00–06:59	4,846	37.4%	-2,041
London morning	07:00–11:59	3,137	35.6%	-4,339
London / New York overlap	12:00–16:59	3,188	38.7%	+890
New York afternoon	17:00–20:59	2,315	39.1%	+1,152
Late / rollover	21:00–23:59	2,211	40.3%	+2,397

This is the useful cut. A single hour is a thin sample and will happily tell you that 03:00 is magic; a session pools enough trades to be worth acting on.

Every hour

Hour (UTC)	Trades	Win rate	Profit factor	Net pips
00:00	860	36.5%	0.95	-466
01:00	686	37.3%	0.95	-427
02:00	573	39.3%	1.01	+88
03:00	605	37.4%	0.97	-222
04:00	595	38.3%	0.99	-80
05:00	723	37.1%	0.95	-418
06:00	804	36.9%	0.95	-517
07:00	703	37.0%	0.96	-330
08:00	649	34.1%	0.83	-1,396
09:00	530	37.9%	0.89	-670
10:00	596	33.2%	0.8	-1,524
11:00	659	35.8%	0.95	-419
12:00	832	38.0%	1.01	+102
13:00	644	39.9%	1.08	+539
14:00	633	38.5%	1.01	+77
15:00	581	39.6%	1.09	+592
16:00	498	37.6%	0.93	-420
17:00	565	40.7%	1.1	+624
18:00	572	40.6%	1.14	+894
19:00	606	36.8%	0.92	-602
20:00	572	38.6%	1.03	+236
21:00	691	42.0%	1.13	+986
22:00	801	40.3%	1.11	+948
23:00	719	38.7%	1.06	+462

What stands out

Best hours by net pips: 21:00 (+986), 22:00 (+948), 18:00 (+894). Worst: 10:00 (-1,524), 08:00 (-1,396), 09:00 (-670).

The pooled total across all hours is -1,941 pips, which is the honest headline: these are textbook strategies and they lose money overall. The point of the table is not that some hour is profitable in isolation — it is how much of the damage is concentrated, and how little of the day is carrying the result.

This contradicts the usual advice, and one reason is our own fault

Every trading course tells you to trade the London open and avoid the quiet hours. This table says the opposite: the London morning block is the worst of the five, and the late session is the best.

Before anyone rearranges their day around that, here is the most likely explanation — and it is a flaw in the test, not a discovery.

Our spread is a constant. The backtest charges each pair a fixed typical spread at every hour. Real spreads do not work that way: they widen sharply around the daily rollover, which is exactly the block that comes out best here. The session our test likes most is the session where our cost model is most optimistic. Some of that +2,397 pips is money a real account would have paid to the spread and never seen.

The strategies also have a bias. Two of the three tested are mean-reverting. Quiet hours suit mean reversion and punish trend following; the volatile London open does the reverse. A table built from trend-following strategies alone could easily invert this result, which means what you are looking at is partly a fact about these strategies rather than about the clock.

The other caveats

An hour filter is a parameter like any other. Pick the best three hours from this table, backtest only those, and you will get a beautiful curve that means nothing — you chose them because they were best in this sample. That is curve fitting with extra steps, and it is why our tester reports the last 30% of every window separately.

Six months is one seasonal cycle. Session behaviour shifts with volatility regimes, central bank calendars and daylight saving. A table built on a different six months would not look identical.

What probably does generalise is narrower than the table: your strategy has a volatility regime it prefers, the hour of day is a rough proxy for that regime, and the pairing between the two is worth knowing for your strategy. Not "trade at 21:00".

Test your own hours

The strategy tester has an hour filter — select the hours you actually trade, on the pair you actually trade, and see what your own window looks like rather than the average of everyone’s.

Which hours of the day actually make money