ICT Concepts

ICT — short for "Inner Circle Trader," the alias of Michael J. Huddleston — is less a single strategy than a shared vocabulary for reading price action. Its concepts describe where resting orders accumulate, how fast moves leave unfilled areas behind, and when during the day those areas tend to be revisited. Traders assemble the pieces into concrete setups, which is why two strategies both labelled "ICT" can look almost nothing alike.

The premise underneath all of it is that price travels between pools of liquidity. Stops rest above swing highs (buy-side liquidity) and below swing lows (sell-side liquidity), clustering at obvious places: equal highs and lows, the previous day's or week's extremes, session highs and lows. An ICT model frames a directional move as price reaching for one of those pools, then repricing back toward the inefficiency it left behind on the way.

## How the model works

Most ICT setups share the same four-step skeleton, whatever name the pattern carries.

**Bias.** A higher timeframe read establishes the intended direction: which side of liquidity is still untouched, where price sits inside the current dealing range (above its 50% equilibrium is "premium," below is "discount"), and how today relates to the prior day's range.

**Liquidity event.** Price sweeps an obvious high or low, running the stops resting there — often early in a session. When that sweep is rejected rather than continued, it is read as a false move rather than a breakout.

**Displacement and structure shift.** A fast, one-sided move in the opposite direction breaks the most recent minor swing point (a market structure shift, or break of structure). Because the move is fast, it usually leaves a fair value gap: a three-candle imbalance where the first and third candle ranges do not overlap.

**Entry into an array.** Instead of chasing, the trader waits for a retracement into a reference zone — the fair value gap, an order block (the last opposing candle before displacement), a breaker, or a retracement band such as the 62–79% "optimal trade entry" area. The stop usually sits beyond the swept extreme; the target is the liquidity pool on the other side or the next unfilled imbalance.

Time is treated as a variable in its own right. Session windows ("killzones" around the London and New York opens), the accumulation–manipulation–distribution sequence known as the power of three, and daily-bias routines all restrict when a setup is allowed to trigger.

## Main variants

Four families cover most published versions. *Sweep-reversal models* trade the failure of a liquidity raid. *Continuation models* use fair value gaps, order blocks and breakers as pullback entries inside an established trend; the "unicorn" variant requires a breaker and a gap to overlap. *Bias and framework models* are not entry systems at all — they produce a daily or session direction that some other trigger then executes. *Correlation models* use SMT divergence: two related instruments are compared, and a high or low made by one but not the other is read as a failed sweep. Cutting across all four, indicator-assisted versions auto-detect gaps, blocks and structure, which is usually what makes a version mechanical enough to test.

## What separates one implementation from another

The concepts are shared; the definitions are not. Versions differ in the timeframe pair (a 4-hour bias with 5-minute entries behaves nothing like 15-minute to 1-minute), in what qualifies as a valid gap or order block (minimum size, whether a wick or a close counts as mitigation), in whether a structure shift is required after the sweep or the array alone suffices, in which sessions are permitted, and in how targets are set — fixed multiples of risk versus the opposite liquidity pool. Instrument choice matters too: session logic built for index futures does not transfer cleanly to a market that trades around the clock.

## Common mistakes

The most frequent is hindsight selection: on a finished chart the "correct" gap is obvious, but a rule set has to say in advance which of several candidates counts. Stacking confluences is the second — each added filter cuts the sample until there is nothing left to evaluate. Others include mishandling time zones and daylight saving so session windows drift, treating every sweep as a reversal instead of requiring the rejection, leaving invalidation undefined, and relying on structure tools that redraw after the fact.

## Evaluating and backtesting a version

Start by rewriting the setup as rules a machine could follow; if the gap, sweep and structure definitions cannot be stated unambiguously, the results will not be reproducible. Check any indicator for repainting — compare where signals sit on history against where they appeared live, and confirm on bar close. Expect small samples: a model restricted to one hour of one session produces few trades a year, so state how many the conclusion rests on. Run the same rules outside their intended session and instrument to see whether the timing logic is doing real work. Include spread, commission and realistic slippage, since tight stops paired with distant targets are unusually cost-sensitive. Returns from liquidity-targeting models tend to be skewed, so read the equity curve with that in mind and compare it against a stripped-down baseline to see what the extra machinery adds.

Strategies in this concept (78)

Frequently asked questions

Is ICT one strategy or a set of concepts?

A set. The name covers a vocabulary — liquidity, imbalance, market structure and session timing — that traders combine in different ways. Two setups can both be called "ICT" and share nothing beyond terminology, so the specific entry, stop and target rules matter far more than the label.

What is a fair value gap, and why do ICT setups rely on it?

It is a three-candle imbalance in which the first and third candle ranges do not overlap, leaving a price band that was traversed quickly in one direction. ICT setups use it as a reference area for pullback entries, on the reasoning that price often revisits it. The definition details — wick or body, minimum size, partial or full fill — vary between implementations and change results materially.

How do ICT concepts relate to Smart Money Concepts (SMC)?

The vocabularies overlap heavily: order blocks, breaks of structure and liquidity sweeps appear in both. SMC is the broader, largely derivative label, while ICT versions usually add explicit session timing and named models such as the power of three or SMT divergence. In practice the two labels are often used interchangeably, including by the videos catalogued here.

Do ICT concepts work on any market and timeframe?

The structural ideas are timeframe-agnostic, but the session-based parts assume a market with a defined open and a daily rhythm — index futures and FX in particular. On instruments trading around the clock, killzone and daily-bias logic has to be re-anchored or dropped. Retest on the instrument you intend to trade rather than assuming the behaviour transfers.

Can ICT setups be automated or backtested?

Partially. Sweeps, displacement, structure shifts and gaps can be coded; the discretionary layer — the narrative, and choosing which of several arrays to use — resists it. Anything you cannot state unambiguously cannot be tested, and a mechanical version is usually a simplification of what a discretionary trader is actually doing, so test the version you would trade, not the idea in general.

How many concepts should a single setup combine?

Fewer than most published versions use. Every added confluence reduces trade frequency, and a session-restricted model with four filters may produce only a handful of trades a year — too few to draw conclusions from. Start with the smallest rule set that expresses the idea and add filters only when testing shows they earn their place.

All strategy concepts · Explore all strategies