ICT Trading Explained: The Concepts, the Vocabulary and an Honest Assessment

ICT is a large, coherent vocabulary for talking about liquidity, timing and structure. Some of it is genuinely useful discipline; much of it is older ideas relabelled; and the story it tells about institutional order flow is a model, not a demonstrated fact.

ICT Trading at a glance

What it is and why it works

ICT stands for Inner Circle Trader, the online name of Michael Huddleston, who published an enormous body of free material through the 2010s and 2020s. It is not a single strategy but a framework: a way of describing markets, plus a set of models built from that description. Its central claim is that price does not wander randomly between technical levels but is delivered between pools of resting orders, and that this delivery happens within predictable windows of the trading day.

Here are the core terms in plain English, because most material assumes them. A liquidity pool is a cluster of stop orders, sitting above obvious highs or below obvious lows. A dealing range is the span between a significant high and low; the top half is premium and the bottom half is discount, and the rule of thumb is that you want to buy in discount and sell in premium rather than the reverse. An order block is the last opposing candle before a strong move, the last down candle before a rally, taken to be where a large participant positioned. A fair value gap is a three-candle window that price moved through too quickly to trade properly. Optimal trade entry is a retracement into roughly the 62% to 79% Fibonacci zone of an impulse. A killzone is a defined time window, the framework uses London around 02:00 to 05:00 New York time and New York around 07:00 to 10:00, where the framework expects the meaningful moves. Power of three describes a session as accumulation, then manipulation, then distribution, with the manipulation phase being a false move that traps traders before the real one; the Judas swing is that false move. SMT divergence is when two correlated instruments disagree, one making a new high while the other fails to.

Assembled, a typical ICT model runs: establish a daily bias, identify which liquidity pool price is likely heading for, wait for the killzone, watch for the Judas swing to sweep the opposite pool, wait for a change of character confirming the reversal, then enter on a retracement into a fair value gap or order block in discount, with the stop beyond the sweep and the target at the opposing pool.

Now the honest assessment, which this page exists to provide. The mechanics underneath are real. Stops genuinely do cluster at obvious levels; a triggered stop genuinely is a market order; large participants genuinely do transact where liquidity is; sessions genuinely do have different volatility characteristics. Those are structural facts, and a framework built on them is not built on nothing. What is not established is the specific narrative layered on top, that a single algorithm delivers price to engineered levels according to a knowable schedule. That is a model. It has never been demonstrated against institutional order flow data, and it does not need to be true for the underlying observations to be useful.

It is also worth being clear about originality. Order blocks are supply and demand zones, which are support and resistance with a story attached. Fair value gaps are imbalances, which Market Profile has called single prints for decades. Liquidity sweeps are Wyckoff’s springs and upthrusts, and Linda Raschke published a systematic version as Turtle Soup in the 1990s. Premium and discount is a 50% retracement rule. Optimal trade entry is a Fibonacci zone. None of this makes the framework worthless, a good vocabulary that organises real ideas has genuine value, but it should be received as a repackaging of established technique rather than as hidden institutional knowledge. And the practical hazard is specific: with this many concepts available, almost any chart can be explained after the fact, and a framework that can explain everything constrains nothing.

How to trade it, step by step

1. Write your model down, reduce it to one setup, and journal every occurrence. The framework contains dozens of concepts and you cannot trade all of them. Choose one specific model (for example, a London killzone sweep of the Asian range low followed by a change of character and an entry in discount) and write out the exact conditions on paper. Everything not in that document is out of scope, and a chart that does not produce your setup is a day off. Then record every occurrence of that written model, whether you took it or not, and what happened. This is the only defence against the framework’s central hazard, which is that with enough concepts available every past chart can be explained and no honest record of your actual decisions ever accumulates.
2. Establish the higher timeframe bias first. On the daily and 4-hour, mark the significant high and low that define your dealing range, mark the 50% level, and identify where the obvious liquidity sits: equal highs, prior session extremes, clean swing points. Your bias is a statement about which pool price is likely to seek next, and it is made before the session opens, not during it.
3. Mark premium and discount and use it as a hard constraint. The midpoint of the dealing range divides premium above from discount below. The discipline is to take longs only in discount and shorts only in premium. This one rule does more practical work than most of the framework, because it prevents the standard retail behaviour of buying after a rally has already happened.
4. Trade only inside your chosen time window. Convert the killzone times into your broker’s server time and into your own local time, and write them down: the framework quotes them in New York time, and the offsets shift with daylight saving. Outside that window you do not trade, regardless of how good the chart looks. This constraint is where a substantial part of the framework’s practical benefit comes from, and it is the first thing traders abandon.
5. Wait for the sweep, not the setup you want. The model expects a move that takes out an obvious pool first (the Asian range low, the previous day’s high, a pair of equal lows) and then fails. Until that has happened there is nothing to trade. Positioning before the sweep is the most common way traders using this framework get caught on the wrong side of the day’s opening move.
6. Require a confirming change of character. After the sweep, drop to a lower timeframe and wait for price to break the most recent counter-swing in your intended direction. That break is what turns a sweep into a signal. Skipping it means you are catching a falling knife with a story attached, and it is the difference between the model and a hunch.
7. Enter on the retracement into a defined zone. Once the change of character has occurred, wait for price to pull back into the fair value gap the impulse left, or into the order block it came from, or into the 62–79% retracement of the leg. Any of these is defensible; using all three simultaneously and taking whichever fills first is not a model, it is three models with the losses hidden.
8. Place the stop beyond the sweep extreme and size from it. The invalidation point is the wick of the liquidity sweep, because price returning past it means the premise failed. Put the stop there with a margin, then derive the position size with the position size calculator. Never fit the stop to a lot size you had already decided on.
9. Target the opposing liquidity pool and check the arithmetic. If you are long after a sweep of the lows, the objective is the clustered orders above the nearest obvious highs. Measure that distance against your stop with the risk-reward calculator before entering; a technically perfect setup that pays less than it risks is still a trade to decline.

Size every one of those entries with the position size calculator and check the trade is worth taking with the risk/reward calculator before you commit.

The conditions it needs

A single written model, applied narrowly

The framework’s breadth is its main practical danger, and the fix is self-imposed narrowness. One setup, one instrument, one time window, written down in advance and unchanged for a meaningful sample of trades. Traders who use it this way report the same benefit people get from any well-specified plan: fewer trades, clearer feedback, and a record that means something.

Time-of-day discipline

The killzone concept has a genuine basis, even if the specific windows are conventions rather than discoveries. Volatility, spread and participation really do vary enormously by session, and restricting trading to the hours when the instrument you trade is actually liquid is sound regardless of framework. See the trading sessions guide for the underlying reality behind the labels.

Liquid instruments with clear session structure

The whole model assumes obvious levels with dense order clusters and a market large enough that reaching them is worth someone’s while. FX majors, index futures and CFDs, and gold satisfy that. In thin instruments the sweeps are often just poor liquidity and a wide spread, and the framework’s vocabulary will describe them convincingly while nothing meaningful is happening.

Premium and discount applied strictly

Of all the framework’s components, the location constraint may be the most quietly valuable. Refusing to buy in the upper half of a range and to sell in the lower half prevents the single most common retail error, entering after the move, and it does so with a rule that requires no interpretation.

A trader who keeps an honest record

Because the framework can explain any chart retrospectively, the only way to know whether your version of it works is a journal of the setups you defined in advance, including the ones you passed on. Without that, the concept count guarantees that your memory of the method will be far kinder than its actual results.

When it fails

• It can explain everything, so it constrains nothing. With order blocks, breaker blocks, mitigation blocks, fair value gaps, inversion gaps, three kinds of liquidity and several structural readings all available, any past chart can be narrated convincingly. A framework that is never wrong in hindsight gives you no information in advance, and this is the single largest problem with how ICT is typically used.
• The institutional narrative is unproven. The claim that a single algorithm delivers price to engineered levels on a schedule has never been demonstrated against actual institutional order flow data, and it is not required for the useful parts to work. Treat it as a model that organises real observations, not as a description of what is happening inside banks. Anyone presenting it as established fact is overstating what is known.
• Rules are applied retrospectively. The gaps that held are obvious afterwards; the ones that failed disappear from view. Educational examples are almost always drawn on completed charts, which makes the models look far cleaner than they are live. Marking your levels before the session and recording your decisions in real time is the only way to find out what you can actually execute.
• Concept-hopping after losses. When a setup fails, the framework offers an immediate explanation; it was the wrong order block, the higher timeframe had not reached premium, the real target was the external range liquidity. That flexibility is comforting and corrosive, because it prevents the loss from ever being informative. If every losing trade has a reason that was invisible beforehand, the model is not being tested.
• Trading outside the time window. The killzone discipline is one of the framework’s more defensible features and it is the first thing abandoned. Setups appear all day; the ones outside the liquid hours arrive with worse spreads, thinner participation and less follow-through, and they are exactly the trades a written plan exists to prevent.
• Paying a great deal for repackaged material. The original body of ICT teaching is free and enormous. A substantial resale industry has grown around it, offering the same concepts at high prices, sometimes alongside performance claims that cannot be verified. Before paying anyone, check whether what you are buying exists free at source, and treat any specific return or win-rate figure as unsubstantiated unless independently audited.

Which markets this works best on

• EUR/USD: The deepest FX market, with the tightest spreads and the clearest session liquidity structure.
• GBP/USD: Wide enough ranges and a heavily watched London open, which suits the killzone framing.
• NAS100 (Nasdaq 100): The New York open regularly sweeps overnight extremes before setting the day’s direction.
• Gold (XAU/USD): Round numbers concentrate orders heavily and the reactions to US data are unusually clean.
• US30 (Dow Jones): Simpler swing structure than the broader indices, which makes the models easier to read live.

For different levels of experience

If you are brand new

If you have arrived here from social media, you have probably seen ICT presented as the hidden truth about how markets work. Here is a fairer description. ICT is a large set of names for things (order blocks, fair value gaps, killzones, liquidity pools) that mostly describe ideas which already existed under other names.

The genuinely useful core is simple and worth learning. Stop-loss orders pile up in obvious places, above recent highs and below recent lows. Because a triggered stop turns into a real order, those piles attract price. Markets move most during certain hours. And it is better to buy in the lower half of a range than the upper half. All four of those are true and all four will improve your trading.

What you should be careful about is the story wrapped around them: that an algorithm run by institutions is deliberately delivering price to specific levels on a schedule. Nobody has ever shown that to be true, and you do not need it to be true for the useful parts to work.

The practical risk for a beginner is drowning. There are dozens of concepts, and with that many available you can explain any chart after the event, which feels like understanding and is not. If you want to start, pick one idea (buy only in the lower half of the range, or trade only during the London hours) and use it properly for a month. Learn risk management and market structure first. Those work everywhere and nobody will sell them to you at a premium.

If your results are inconsistent

The intermediate trader’s problem with ICT is almost never that they know too little. It is that they know too much and have never narrowed it down. When you have order blocks, breaker blocks, mitigation blocks, fair value gaps, inversion gaps, internal and external liquidity and several structural readings all live at once, you will find a justification for any trade you feel like taking, and you will do it sincerely.

The fix is aggressive reduction. Choose one model, write it out as a checklist with specific conditions, and trade only that for a defined sample: fifty occurrences, say. Everything outside the checklist is not a trade. This feels like giving up most of the framework, and it is, and that is the point: you cannot evaluate a method that changes shape after every loss.

Be especially watchful for retrospective reasoning. When a setup fails, the framework will always supply an explanation that was not visible beforehand. Notice yourself doing it. A useful habit is to write your reasoning down before entering and then refuse to add to it afterwards; the gap between the two is where the honest information about your trading lives.

On execution, the two components with the clearest practical value are the time filter and the premium-discount constraint. Restricting yourself to liquid hours and refusing to buy in the top half of the range will change your results more than any refinement to which candle counts as the order block. And keep the target logic, aiming at the opposite pool of liquidity rather than at an arbitrary multiple, because that is one of the framework’s genuinely good ideas.

If you are experienced

The defensible content reduces to three things, all of which predate the branding. First, stop-cascade dynamics: clustered resting stops create locally convex marketable flow, execution against them is efficient for size, and the resulting impact is temporary rather than informative. This is documented microstructure and it is the strongest leg the framework stands on. Second, intraday seasonality: session-conditional volatility and liquidity patterns are well established, and the killzone windows are a coarse but broadly correct mapping onto them. Third, the discipline effect of a rule set that specifies location, time and invalidation in advance, which improves outcomes independently of whether the model’s theory is correct.

The problematic part is epistemic rather than mechanical. The framework has a very large rule surface, no published out-of-sample validation, and a strong tendency toward unfalsifiable post-hoc attribution; a failed setup is reclassified rather than counted. That combination makes it structurally difficult to distinguish edge from narrative fit, and it is why the concept-count itself should be treated as a risk factor. If you intend to test any of it, the components must be mechanised: a fixed order-block definition, a fixed gap definition, a fixed swing rule, a fixed time window, evaluated on out-of-sample data with realistic spread and slippage, with every occurrence counted rather than the ones you noticed.

On instrument choice, note that several concepts implicitly assume observable participation. Where you have genuine exchange volume (futures, index products) a volume profile measures the same thinness the fair value gap infers from three candles, and does so directly. In spot FX, MT5 gives tick volume rather than transacted size, so any claim about where institutions were filled is inference on inference. The framework is usable there; the confidence attached to it should be lower.

Risk management for this strategy

The risk profile of ICT trading is unusual because the largest danger is not a market outcome but a reasoning failure. A framework with dozens of interlocking concepts can rationalise any position, including the one you should be exiting. The practical defence is procedural: write the setup and the invalidation before entering, and refuse to revise the reasoning while the trade is live. If you find yourself reclassifying which order block mattered after price has moved against you, that is not analysis, and the position should be closed.

On stop placement, the framework is at its best. The invalidation point is well defined (beyond the extreme of the liquidity sweep, because price returning past it means the premise has failed) and having a genuine structural invalidation is a real advantage over methods that use arbitrary distances. Respect it, put the stop beyond the wick with a margin for the spread, and derive the size from that distance rather than the reverse.

Two exposure notes. First, killzone trading concentrates activity into short windows around session opens, which are also the moments when spreads widen and fills degrade, so assume slippage rather than treating it as unusual. Second, the models produce correlated positions easily; a sweep on EUR/USD, GBP/USD and the DAX at the same London open is one idea in three instruments, and total exposure should be managed as such.

Finally, a risk that is not about the market at all. There is a significant resale industry attached to this framework, and some of it carries performance claims that cannot be verified. The original teaching material is free. Treat any specific return figure as unsubstantiated unless it has been independently audited, and be aware that funded-account programmes marketed alongside these strategies have their own rules and costs: see our prop firm comparison before committing.

Where Market Structure Pro fits

The central practical problem with ICT is not that the concepts are wrong. It is that there are too many of them, and a framework that can account for any chart after the fact gives you nothing to lean on before it. The moment that actually decides the trade (is this sweep resolving into a move, or is this market simply chopping around a level?) is the one the framework answers with narrative rather than with a measurement.

Market Structure Pro answers it with a single verdict. It fuses twenty-seven tools into one output (TRADE, TRANSITION or NO TRADE) with a confidence percentage, an A/B/C grade and a plain-English explanation of what is supporting or limiting the read. It gives you one thing to disagree with, which is precisely what a framework with dozens of simultaneous concepts fails to provide. The dedicated ranging filter exists to return NO TRADE in dead or choppy conditions, which is where the models generate their highest signal count and their lowest follow-through.

It is session-aware, which maps directly onto the killzone discipline: a setup appearing outside the liquid hours is graded for the conditions it is genuinely in rather than treated the same as one at the London open. It is spread-aware, which matters because the killzones coincide with the moments spreads widen most, and a tight sweep entry can be invalidated by execution cost alone.

And it is non-repainting: state locks on the closed bar. That is the property that matters most here, because the framework’s besetting weakness is retrospective reinterpretation. A verdict that was recorded at the time you acted and cannot be revised afterwards is a direct, mechanical check on the temptation to rewrite the reasoning after the outcome. Market Structure Pro is decision support: it places no trades, it is not a signal service, and it guarantees nothing.

Frequently asked questions

What is ICT trading?

ICT stands for Inner Circle Trader, the online name of Michael Huddleston, who published a large body of free trading material. It is a framework rather than a single strategy, built on the idea that price moves between pools of clustered stop orders within specific windows of the trading day. Its vocabulary includes order blocks, fair value gaps, liquidity sweeps, killzones and premium and discount pricing.

Is ICT trading legit?

The mechanics it rests on are real: stops do cluster at obvious levels, triggered stops do become market orders, and sessions do have different liquidity characteristics. What is not established is the specific narrative that a single institutional algorithm delivers price to engineered levels on a schedule, that has never been demonstrated against actual order flow data. Treat it as a useful organising framework rather than as proven institutional mechanics.

What are ICT killzones and what times are they?

Killzones are defined windows when the framework expects the significant moves. The commonly quoted ones are the London killzone at roughly 02:00 to 05:00 New York time and the New York killzone at roughly 07:00 to 10:00. The underlying idea is sound, liquidity and volatility genuinely do vary by session, though the specific boundaries are conventions rather than discoveries. Convert them carefully, since your broker’s server clock is usually neither London nor New York time.

What does premium and discount mean in ICT?

Take a significant high and low and find the midpoint. Above the midpoint is premium, below it is discount. The rule is to look for buys in discount and sells in premium, so you are entering at a relatively better price within the range. It is a 50% retracement concept with new labels, and it is arguably the most practically valuable single component of the framework because it prevents entering after a move has already happened.

Is ICT just a rebrand of older concepts?

Largely, yes, and this is worth saying plainly. Order blocks are supply and demand zones, which are support and resistance with a story attached. Fair value gaps are imbalances, called single prints in Market Profile for decades. Liquidity sweeps are Wyckoff’s springs and upthrusts, published systematically as Turtle Soup in the 1990s. That does not make the framework useless, organising real ideas into a coherent vocabulary has value, but it is a repackaging rather than hidden knowledge.

What is optimal trade entry?

Optimal trade entry, or OTE, is a retracement into roughly the 62% to 79% Fibonacci zone of an impulsive move, used as an entry area on the reasoning that it offers a favourable price with a defined invalidation beyond the origin of the move. It is a standard Fibonacci retracement concept with a specific band selected. Like all retracement entries, its weakness is that strong moves sometimes never retrace that far.

What is the power of three and the Judas swing?

Power of three describes a session in three phases: accumulation, where price ranges; manipulation, where it makes a false move that traps traders; and distribution, where the real move happens. The Judas swing is that false move. The observation behind it is genuine, sessions frequently do sweep an obvious level before reversing, but the three-phase structure is a description that fits many sessions after the fact and rather fewer in advance.

Do I need to pay for ICT training?

No. The original material is free and extensive, and much of the paid ecosystem resells the same concepts. Before buying anything, check whether the content exists at source without charge. Be particularly cautious about any specific win rate or return figure, since such claims are almost never independently audited and are not verifiable from a screenshot or a trading history export.

Why does ICT seem to work on past charts but not live?

Because the framework contains enough concepts that any completed chart can be explained convincingly, while in real time you face many candidate levels and no way to know which matters. Failed setups also vanish from view in hindsight, whereas successful ones are what get drawn in educational examples. The only remedy is marking levels before the session and journalling every occurrence of a written model, including the ones you decided to skip.

Related reading

• Smart Money Concepts: The broader family this framework belongs to, assessed on the same honest terms.
• Liquidity Sweeps: The single mechanism most of the framework is built on, explained without the narrative.
• Fair Value Gaps: The entry zone the models rely on most, and its real limitations.
• BOS and CHoCH: The confirmation step that turns a sweep into an actual entry.
• Risk Management: The part no framework can replace, and the one that decides whether you survive learning it.