ICT Killzones in Forex: Trading London & New York

What Is an ICT Killzone (and Why Does Session Timing Matter)?

A killzone is a fixed window on the intraday clock in which your model is allowed to trade — and outside of which it is not. It is a filter, not a setup. Nothing about the window itself tells you to buy or sell; it only tells you when your setup is worth taking seriously.

The reason a window can do that work at all is that the currency market is open around the clock but is not staffed around the clock. Liquidity is a function of who is at their desk. When the London banks come online, real two-way flow arrives with them, and price starts moving with intent instead of drifting. When those desks go quiet, the same chart produces small, aimless ranges that look like setups on a 5-minute chart and behave like nothing at all.

ICT frames this through two ideas worth naming even though this page is not the place to teach them from scratch. IPDA — the interbank price delivery algorithm — is the assumption that price is being delivered toward pools of resting orders on a repeating schedule rather than wandering randomly. Power of three is the shape that delivery tends to take around a session open: a period of accumulation, a manipulative push that runs stops in the wrong direction, then the real distribution leg. If price is going to accumulate, manipulate and distribute around session opens, then the hours immediately after those opens are where the sequence is observable. That is the whole argument for the killzone. The hours are not arbitrary; they are the hours in which the pattern the model expects actually has a reason to form.

Two consequences follow, and both matter more than the hours themselves:

• A killzone is an exclusion rule first. Its main job is to delete the twenty-odd hours where your setup is statistically indistinguishable from noise. The "kill" is what happens to the rest of the day.
• It does not manufacture setups. A window with nothing in it is a window you sit out. Most traders break the framework here, not in the arithmetic of the hours.

One boundary before the hours: none of this transfers cleanly to a market that never closes. Running ICT across crypto and forex on a 24/7 playbook has to replace the session open with something else entirely, because there is no London desk arriving and no daily close to anchor to. That is its own subject; everything below is forex.

London Killzone: Exact Hours (ET, UTC, and the DST Fix)

02:00 to 05:00 New York time. That is the London killzone, and the "New York time" part is not a formatting choice — ICT quotes every window in New York local time, so New York's clock is the reference the whole framework hangs on. Convert from that, never to it.

Read the middle two columns again, because that is the part most sources skip. The window is fixed in New York time, which means its UTC equivalent shifts by one hour twice a year. A killzone alert you hard-coded to 07:00 UTC in January fires at 03:00 New York time in July — an hour after the window opened, with the sweep already behind you.

The daylight-saving rule, stated once

Three facts are all you need:

1. New York runs on EDT (UTC-4) from the second Sunday in March to the first Sunday in November, and on EST (UTC-5) the rest of the year.
2. London runs on BST (UTC+1) from the last Sunday in March to the last Sunday in October, and on GMT (UTC+0) the rest of the year.
3. The two do not switch on the same day. For roughly three weeks in March and one week at the end of October, New York and London are four hours apart instead of five.

That third fact is the one that quietly breaks people. In those desync weeks the London killzone opens at 06:00 London time, not 07:00 — because New York has already sprung forward while London has not. Every other week of the year the window sits at 07:00–10:00 London local, which is why so many UK-based traders never notice the problem until the one week it costs them the open.

The safe habit is to stop converting in your head. Set your charting platform's timezone to New York (America/New_York) and read the window natively, or work from UTC and re-derive it at each changeover. Never trust your broker's server clock: most MT4 and MT5 servers run on GMT+2 or GMT+3 and follow European daylight saving, which puts them out of step with the New York window for those same desync weeks — which is why the MT5 broker you run an ICT EA on matters as much as the hours themselves.

If you want the broader session map alongside these two windows, our forex market hours tool lays out the full trading day, and the killzones sit inside it.

New York Killzone: Exact Hours and the London Overlap

07:00 to 10:00 New York time — for forex. The qualifier matters, and it is the single biggest reason you have seen three different answers to this question.

Here is why the sources disagree, so you can stop collecting contradictory screenshots:

• 07:00–10:00 ET is the forex window. It starts when the New York FX desks pick up the book from London and runs into the middle of the overlap.
• 07:00–09:00 ET is the same window trimmed to its core. Some teachers quote only the first two hours because that is where displacement usually happens. Treating 09:00–10:00 as extra time rather than prime time is a reasonable simplification.
• 08:30–11:00 ET is a different instrument's window. It is built around US index futures: the 08:30 data releases and the 09:30 equity cash open. If you are trading EUR/USD, it is the wrong window — you would arrive after the FX move has already run.

For everything in this article, New York killzone means 07:00–10:00 New York time.

Why the London/New York Overlap Carries Extra Weight

Take the two sessions as whole trading days rather than as killzones. London's book runs roughly 07:00–16:00 UTC; New York's runs roughly 12:00–21:00 UTC. They are both open from 12:00 to 16:00 UTC — 07:00 to 11:00 New York time.

Now compare that to the New York killzone. The killzone is 12:00–15:00 UTC. It sits entirely inside the overlap. That is not a coincidence and it is the cleanest justification for the window: for those three hours the two deepest pools of currency liquidity in the world are staffed simultaneously, which is when spreads are tightest, size can actually be absorbed, and a displacement leg has enough fuel to reach the next pool instead of stalling halfway.

It cuts the other way too. Depth also means the stop runs are cleaner and faster. A sweep during the overlap can take out the level and reverse inside a single 5-minute candle, which is exactly why the sequence below insists on a body-close confirmation rather than an entry on the wick.

How to Trade Inside a Killzone: The Entry Sequence

The window is a filter. This is the thing the filter lets through. Run it in this order, every time — the sequence is what separates trading a killzone from being awake during one.

Mark Liquidity Before the Window Opens

This step happens before the clock, and doing it inside the window is the most common way traders end up drawing levels to justify a trade they have already decided to take.

What to mark, in priority order:

• The prior session's high and low. For the London killzone that is the Asian range. For the New York killzone it is the London killzone's own high and low.
• The prior day's high and low. These are the heaviest pools on the intraday chart and they routinely act as the target rather than the sweep.
• Equal highs or equal lows. Two or more touches at the same price leave an obvious shelf of stops. These are the levels most likely to be run first.

Three to five levels is plenty. If your chart has fifteen lines on it, every move will look like a sweep of something, which is the same as having no levels at all.

Wait for the Sweep and the Structure Shift

The window rarely opens by going where it is eventually going. The typical shape is a push that takes one of your marked levels, triggers the stops resting behind it, and then reverses hard. In ICT vocabulary that opening fake-out is the Judas swing — the setup that forms inside the killzone window and traps the traders who entered on the first move.

Two conditions, and both have to hold:

1. The sweep must actually take the level. Price needs to trade through your marked high or low and reject it — a wick through, not a body close beyond it. A body close beyond the level is not a sweep; it is a break, and you have the direction wrong.
2. The shift must be a close, not a touch. After the sweep, price has to break the most recent opposing swing point on your execution timeframe with a body close and obvious displacement — a fast, wide-range leg, not three overlapping candles limping across the level. A confirmation candle that closes back inside the range is not a shift.

If the sweep happens and no shift follows, there is no trade. That is a complete and correct outcome for a killzone, and it will be the outcome more often than not.

Enter at the Order Block or OTE Retracement

Displacement leaves an inefficiency behind it, and price usually comes back for some of it. You are entering on that return, not on the leg itself.

Two acceptable references, and you should pick one and stay with it rather than switching to whichever is closer to price:

• The order block — the last opposing candle before the displacement leg. Rest your limit at its opening price or its midpoint.
• The OTE zone — the 0.62 to 0.79 retracement of the leg measured from the swept extreme to the shift point. When the order block and the OTE zone overlap, that overlap is the highest-quality version of this entry.

Rest the order and let it fill. If price never returns to the zone, you did not miss a trade — you avoided a chase into the back half of the window, which is where the risk-to-reward on this model collapses.

Set the Stop and the Target

The stop-loss location is not a preference. It goes a few pips beyond the swept extreme — the wick that took your level — because that price is the invalidation point of the entire idea. If price trades back through it, the sweep was not a sweep. Putting the stop just beyond the order block instead is tempting, because it is tighter, and it is the single most reliable way to be stopped out of a trade that then works.

The take-profit side runs the same logic in reverse: you target the next pool of liquidity in the direction of the shift. In practice that means the opposite side of the range you marked before the window, then the prior day's high or low beyond it. A useful second exit is simply the window's close — if the move has not reached the pool by then, the conditions that produced it are over.

One Window, Start to Finish: An Illustrative EUR/USD Example

The numbers below are illustrative — a constructed example to show the sequence end to end, not a recorded trade or a backtest result.

Overnight, EUR/USD builds an Asian range between 1.0838 and 1.0865. You mark both edges before 02:00 New York time and go and make coffee.

1. 02:15 NY. Price pushes down through 1.0838 and prints a low at 1.0831, then closes back above the level. The low is swept; the sell stops under the range are gone.
2. 02:45 NY. A wide-range 15-minute candle closes at 1.0854, above the last lower high at 1.0851. That is the structure shift — a close, with displacement, not a wick.
3. 03:15 NY. Price retraces into the order block at 1.0841–1.0845, which overlaps the 0.62–0.79 retracement of the leg. Your resting limit fills at 1.0843.
4. Stop at 1.0827, four pips beyond the 1.0831 sweep low. Risk: 16 pips.
5. Target at 1.0879, the prior day's high and the next pool above the Asian range. Reward: 36 pips, or 2.25 times the risk.
6. 04:30 NY. Price reaches 1.0881 and the target fills, inside the window.

Notice what the chart makes obvious that the numbered list does not: the stop is a long way below the entry relative to how tight the order block is. That distance is the price of the model. Shrink it and the sweep that sets up your trade is also the move that closes it.

Which Pairs Actually Move in Each Killzone?

Right window, wrong instrument is still a losing combination. A killzone works because specific desks are staffed, and those desks trade specific currencies. Point the framework at a pair whose natural liquidity providers are asleep and you get the ranges without the follow-through.

London killzone (02:00–05:00 NY). The European desks are the ones arriving, so European currencies are where the flow is:

• EUR/USD and GBP/USD — the deepest books in the window, and the cleanest sweeps of the Asian range.
• GBP/JPY and EUR/JPY — larger ranges, wider stops. Workable, but the stop distance forces a smaller position for the same risk.
• EUR/GBP and USD/CHF — narrower and slower, but they respect the Asian range as cleanly as anything else in the window.

New York killzone (07:00–10:00 NY). The dollar is the common factor, and the overlap means the European pairs are still liquid:

• EUR/USD and GBP/USD carry over from London, now with both books open.
• USD/JPY and USD/CAD come into their own here — USD/CAD in particular, because North American data and the oil complex both hit in this window.

What to leave alone. AUD, NZD and their crosses had their liquidity hours while you were asleep; by the London killzone the Pacific desks have gone home. Exotics behave even worse: the sweep is real but the spread eats the first third of the target. One pair per window is a defensible rule. Two is the practical maximum before you are watching charts instead of executing a plan.

Risk Rules Specific to Killzone Trading

A time filter is not risk management. In some ways it makes risk worse — you have compressed your trading into three hours, which concentrates decision-making into exactly the period when price is moving fastest.

Six rules that are specific to this model rather than generic advice:

1. The stop is placed by the sweep, not by your account size. If the correct stop is 30 pips away and 30 pips is more than you want to risk, the answer is a smaller position, never a closer stop.
2. Size from the stop distance. Fix your risk per trade as a percentage of the account, then let the stop distance decide the lot size. A position size calculator does this in one step, and it is the difference between a 16-pip stop and a 40-pip stop costing you the same money.
3. Set a reward-to-risk floor and refuse trades under it. Two-to-one is a sensible floor for this model, because the entry is a retracement and the target is a defined pool — if the nearest pool is only 1.2 times your risk away, the setup is valid and the trade is not.
4. Cap the window at one setup, two at most. The sequence produces one high-quality entry per window on a given pair. The second is usually the same idea re-entered at a worse price after it failed.
5. Flatten or stop managing at the window's close. A killzone trade that has not reached its target by 05:00 or 10:00 New York time has lost the conditions that created it. Take what is there, move to break-even, or close it — but stop treating it as a killzone trade.
6. Track outcomes in R-multiples, not pips. A 16-pip win and a 40-pip win are the same trade if both were 2R, and pips will lie to you about which pair is actually working.

None of this makes losing runs go away. A window filter changes when you trade, not whether a sequence of valid setups can fail one after another — read our risk warning before sizing anything on this model.

Killzone Mistakes That Quietly Wreck the Strategy

These are failure modes of the framework itself, not general trading errors — and each one produces a losing month while the trader believes they are following the rules.

• Trading every candle inside the window. The window grants permission to look, not to act. Without a sweep and a shift, there is no setup, and three hours of screen time is not an obligation to use it.
• Reading the clock off your broker's server. MT4 and MT5 server time is usually GMT+2 or GMT+3 on European daylight-saving rules. For a few weeks each spring and autumn that puts your "02:00" an hour away from ICT's 02:00. Anchor to New York or to UTC, never to the platform default.
• Forcing an entry with no price-delivery reference present. No order block, no imbalance, no OTE overlap — just "it swept, so I bought". Those trades are the bulk of the false signals traders blame on the model.
• Treating the 08:30 New York data release as a killzone setup. It falls inside the New York window, it produces enormous displacement, and it is a different phenomenon with different risks — slippage and spread widening on release are execution problems, not structure. Either apply a news filter around scheduled releases or accept that you are trading news, not the model.
• Widening the stop to survive. The stop was placed at the invalidation point. Moving it means the trade is already wrong and you have decided not to pay for it yet.
• Holding past the close "to give it room". After 05:00 or 10:00 New York time, the desks that produced the move are handing off. Whatever happens next is a different session's business.

Checking a Live Signal Against the Killzone Window

Once the hours are fixed in your head, there is a cheap way to keep yourself honest without sitting at the chart for the entire window: read timestamps.

Every signal on our free forex signal feed carries a real-time timestamp. Take that timestamp, convert it to New York time using the same rule from the top of this page, and you know immediately whether the setup formed inside a killzone or in the dead hours. A EUR/USD signal stamped 07:42 UTC in January is 02:42 New York time — forty-two minutes into the London killzone, which is exactly where the sequence above expects the sweep to have happened. The same signal stamped 17:00 UTC is not a killzone event at all, whatever the chart looks like.

That is the honest scope of it: the feed timestamps signals in real time so you can check them against the window. It is a feed for manual review, not an automated killzone-entry system. If you want the sequence itself executed for you — the sweep detected, the shift confirmed, the limit rested at the OTE zone while you sleep — that is the job of an Expert Advisor with a proper trading session filter, and no signal feed replaces one.

The Takeaway: Trade the Window, Not the Whole Day

The setup that chopped you out at 11:40 was not a bad setup. It was a valid pattern in an invalid hour, and no amount of refinement to your entry criteria fixes a timing problem.

So fix the timing first. Set your platform to New York time. Write 02:00–05:00 and 07:00–10:00 somewhere you will see them. Mark your levels before the window opens, wait for the sweep and the close that confirms it, and let the window's end be the end of your trading day. Two windows a day is not a limitation of this model — it is the model.

FAQ

Do ICT killzone times change with daylight saving?

The windows themselves never move: they are always 02:00–05:00 and 07:00–10:00 in New York local time. What moves is everything derived from them. Because New York switches to EDT on the second Sunday in March and back on the first Sunday in November, the UTC equivalent of each window shifts by an hour twice a year, and any alert, indicator or script hard-coded in UTC will fire at the wrong time until you update it.

What if the London killzone lands in the middle of my night?

Trade the New York killzone instead, and treat London's high and low as levels rather than as a session you participate in. The sequence is identical; the only change is which prior session you mark. Traders in Asia-Pacific time zones usually find the reverse — London is the workable window and New York is the one that lands at 3 a.m.

Can a bot or EA trade killzones for me?

Yes, and the time filter is the easiest part to automate — a session filter that only allows entries between two clock times is a few lines of code. The hard parts are the parts a human does visually: defining what counts as a sweep, distinguishing a genuine displacement close from a slow drift through a level, and placing the retracement limit. Automate the clock first; that alone removes the out-of-window trades that cause most of the damage.

Do killzones work the same way on crypto?

No, and the reason is structural rather than a matter of tuning the hours. Crypto has no session open, no daily close and no desk handover, so the accumulation-manipulation-distribution shape that anchors a killzone has nothing to anchor to. Some traders map the forex windows onto crypto anyway because a large share of derivative volume still tracks traditional hours, but that is a different framework and a different argument.

How many setups should I expect in one killzone?

Frequently zero. A complete window with a sweep of a marked level, a displacement close through structure and a retracement that fills your limit is not a daily occurrence on a single pair. If you are finding two or three per window, you have almost certainly loosened the definition of one of the three conditions — most often the shift.

Sources & Further Reading

Want to go deeper? These independent, authoritative sources shaped this guide — each one is worth reading in full:

• NIST - Daylight Saving Time in the United States, with the exact start and end rules
• TradingView Pine Script documentation - defining and testing session time windows in code
• MQL5 Reference - TimeCurrent, TimeGMT and related date/time functions an EA can use to build a session filter