Gold Killzones: The London and NY Windows That Matter

The idea underneath is sound: Gold at 03:00 Paris and Gold at 15:00 Paris are different instruments. What is usually missing from killzone content is measurement. Which window actually carries the volume? How often does a London-window breakout hold? Does a sweep inside a killzone reverse more often than one outside it? We have the data to answer those questions for XAU/USD — several years of 5-minute spot data, a sample size next to every number — so this article maps the four classic killzones onto Gold’s real clock and says what each one has historically done. Where the folklore holds, we say so. Where it doesn’t, we say that too.

The four killzones, on the clock

Killzone times are conventionally quoted in New York time; our studies use Paris time, so here are both. The Paris column assumes the usual 6-hour offset — for a few weeks a year, when the US and Europe change their clocks on different dates, the gap is 5 hours and every box shifts by an hour on your chart. That is the most common reason a killzone “stops working” in March and late October.

Where the volume actually is

Across 653 normal trading days we measured the share of daily tick volume traded in each Paris-clock hour — the full shape is in our volume-by-hour study. Laid over the killzones:

The three busiest hours of the Gold day are 15:00 (8.4%), 16:00 (8.5%) and 14:00 (6.5%) — the New York open block. The New York AM killzone (13:00–16:00 Paris) catches the first two, the London Close killzone (16:00–18:00) catches the third, and together those three hours carry roughly 23% of the whole day’s volume.

The London open hour, 09:00 Paris, carries 4.5% — busy, but about half of a single New York-open hour. The folklore that London is “the big session” for Gold is not what the volume says: the London killzone is where the first meaningful participation arrives, not where most of the business gets done.

The Asian killzone sits in the quietest part of the curve: the thinnest hours of the day are 23:00 (1.0%), 00:00 (1.4%) and 22:00 (1.5%), and the 02:00–06:00 window that follows is not much busier. A level printed there was set by very little volume — which is exactly why it gets tested when the volume arrives.

On US data days (n=119), the 14:00 hour jumps to 10.7% and 15:00 to 10.5% — the 14:30 Paris (08:30 ET) release concentrates the day into the first ninety minutes of the New York killzone. A killzone box drawn without the economic calendar next to it is half a tool.

The London killzone: an expansion window that mostly pokes

The textbook London play is the Asian-range break: price takes out the overnight high or low during the London window and either reverses (the sweep) or runs (the breakout). We tested it in our Asia range breakout paper: 2,303 Asia-range breakouts on XAU/USD, Jan 2023 → Aug 2026, 923 valid days, with placebo levels as controls.

The result that matters here: when the Asia range (01:00–08:55 Paris) is broken by a wick during the morning window (09:00–14:25 Paris), price closes back inside within 30 minutes 75.7% of the time [72.2; 78.9], on 630 normal-day events. Three morning pokes in four fail inside half an hour. The “trap” everyone describes is real.

Two things complicate it. First, the failure rate depends on the horizon. Count only breaks where an M5 bar actually closes beyond the level and the 30-minute re-close rate drops to 62.9%. Ask where the day finishes and the wick-break “failure” is 50.2% — a coin flip; for close breaks, the day ends back on the original side 48.2% of the time [44.1; 52.3]. London generates many short-lived pokes, not a reliable direction for the day.

Second — the part the killzone framing hides — the same behaviour appears at any level, at any hour. Afternoon breaks of the same range re-close within 30 minutes 69.0% of the time [63.5; 74.1]. Breaks of phantom levels placed beyond the real Asia extremes, levels nobody is watching, re-close 76.0% [70.0; 81.1]. Breaks of the London first-hour range, tested between 10:00 and 14:25, re-close 73.9% [71.0; 76.5]. London is not producing a special trap. It is where you first see the M5 poke-and-fade behaviour Gold shows around every recent level, all day long.

The pre-release version is sharper. The fake-break rate of the 09:00–14:25 window runs 73.1% on normal days and 85.7% on mornings before NFP (n=28 breaks). A London-killzone breakout on an NFP morning has historically been a trap more than five times in six — not because London is weak, but because the market is waiting for 14:30.

The New York killzone: where the day gets decided

If London is where Gold pokes, New York is where it moves. The same study measured how far price travels after a break: after a morning close break, half of all moves never extend beyond 0.075 ATR before the first close back inside; after an afternoon close break the median is 0.135 ATR. Two hours after the break, price has closed at least 0.25 ATR beyond the level 10.5% of the time for morning breaks versus 21.7% for afternoon ones. The afternoon moves more, in both directions.

The reversal you were taught to expect in London actually arrives in New York. Of 571 morning close breaks on normal days, 23.8% saw the opposite edge of the Asia range broken in the afternoon — the US session reversed the morning move through the entire overnight range, and 95.6% of those days closed back on the original side. But 67.1% of morning breaks never saw the other edge broken; for those, the day closed back inside the range only 29.2% of the time [24.9; 34.0] and continued at least 0.25 ATR beyond the level 39.9% of the time [35.2; 44.9]. One morning break in four gets reversed by New York. Two in three don’t. You cannot tell which at 10:00 Paris.

The New York window is also where scheduled volatility lives. On the 25 FOMC days we measured, roughly 79% of the day’s total range happened after the 20:00 Paris (2pm ET) announcement, against 20% on normal days. Across 181 US releases at 08:30 ET, NFP produced a median 5-minute range of $13.9 and continued in the direction of that first move for 15 minutes 93.5% of the time. On a data day the New York killzone is a release window with its own rules — see what actually happens on FOMC days.

Sweeps inside killzones: timing does not rescue the reversal story

The strongest killzone claim concerns liquidity sweeps: a sweep inside a killzone is supposed to be the high-probability reversal. We measured 2,434 first sweeps of eight level families across 807 days and asked how often price closed back on the original side of the level by end of day. The unconditional answer was about 50% — and the clock does not change it.

Sweeps during the London morning reversed by end of day 44% of the time [38–49], n=304. Sweeps during the New York morning reversed 49% [44–53], n=459. Neither window produces a reversal edge; the London morning, if anything, leans slightly toward continuation. London’s own levels, swept later in the day, did no better: London high 48% [42–55] (n=244), London low 47% [40–54] (n=200). Every interval either contains 50% or sits just below it.

The only fade-shaped edge in that study is conditional on regime, not on the window: the Asia low swept in a Transition regime reversed 65% of the time [53–76], n=63. That event often happens inside the London killzone, which is probably where the folklore comes from — but the edge belongs to the level and the regime, not to the box.

A killzone tells you when a sweep is likely — that is when participation arrives to run the overnight stops. It does not tell you the sweep will reverse. Only the first claim survives the data.

What the windows really change: the behaviour regime

Here is the finding that, in our view, justifies drawing killzones at all: the windows do not hand you direction, they change the kind of market you are in — and that decides which tools work.

The clearest evidence is mean reversion. Across 6,945 touches of the ±2σ daily VWAP band over 830 days, price got back to VWAP before the end of the day 95% of the time when the touch happened in Asia, 78% in London, and 30% in New York. On a fixed 60-minute horizon, which removes the “more day left” effect, the ranking holds: Asia 34%, London 14%, New York 12% (full breakdown here). An extension in the Asian killzone is, historically, something that comes back. The same extension in the New York killzone usually does not.

Put the pieces together and the boxes read like this: the Asian window builds thin-volume levels and snaps back when stretched; the London window tests those levels and fails most tests within 30 minutes, at the base rate of any level; the New York window carries the volume peak, the largest moves in either direction and most of the reversing, with mean reversion sharply weaker. That is what killzones are for: not a signal, a regime clock. Fade tools where fades historically worked, continuation tools where moves historically travelled, patience before releases.

Setting it up on TradingView

We publish a free script for the boxes themselves: Session Killzones ICT - Clean, on our TradingView profile. It draws the windows and nothing else — no signals, no arrows, no claims.

For the measurement layer, three of our indicators do what this article describes. Intraday Timing Stats shows, on the history you have loaded, in which hour the high and low of day have historically formed. RVOL Time-of-Day compares the current bar’s volume to the norm for this exact time slot rather than a flat average — the only way to tell whether a 09:00 Paris burst is heavy or just the usual London arrival. And the Session Liquidity Regime dashboard puts the current session, the active killzone, the session levels and the regime read (ATR, ADX, Chop) in one panel. The free SLR Dashboard Lite is the entry point to that dashboard; if you want session context on your chart before paying for anything, start there.

Everything in the suite was developed hand-in-hand with a full-time Gold day trader, and every historical frequency the indicators display comes with its sample size — 18 indicators today, more added every month.

Limits, then the bottom line

Every number above is a historical frequency on one instrument — XAU/USD spot, Dukascopy data, 2023 to mid-2026 — measured on bar closes with no simulated trades, stops or targets; a 75.7% re-close rate is not a 75.7% win rate for any trade plan. Killzone boundaries are conventions: our studies use Paris-time windows (Asia range 01:00–08:55, morning 09:00–14:25, afternoon from 14:30), which the ICT boxes overlap without matching minute for minute. Tick volume is a proxy on spot Gold. And 2023–2026 was a Gold bull market in which the morning-versus-afternoon differences were strongest in 2023 and close to zero since — re-measure before relying on anything.

Killzones are a clock, and the clock matters on Gold: the New York window carries the day’s volume and its largest moves, the London window produces mostly short-lived pokes at the overnight extremes, and the Asian window builds thin-volume levels that revert when stretched. What the clock does not do is tell you direction. London-window breakouts end the day back inside the range about half the time, sweeps in either window reverse about half the time, and the reversal that does happen comes from New York, one morning break in four. Draw the boxes — then put a sample size next to whatever you expect to happen inside them. For how the sessions themselves line up on Gold, start with Gold trading sessions explained.

Delta Terminal ResearchStudies run on 3.5 years of XAU/USD spot data, developed hand-in-hand with a full-time Gold day trader. We publish the negative results too.