ProEA Lab · Topic hub
ICT concepts explained
ICT concepts are the trading vocabulary taught by ICT, the Inner Circle Trader: time (killzones on the New York clock), liquidity (the highs and lows price runs), displacement (a fast move that leaves fair value gaps) and PD arrays (gaps and blocks read in premium or discount). Most ICT models are one shape on different clocks: price runs one side of a range, closes back inside, then heads for the other side.
Our position: a precise way to name time and liquidity, not a proven edge. Each model below gets one written rule, a replay of it, and the tool that draws it, if one of ours does. What we measured is here too, failures included.
The learning path
First the clock. Then one shape on five clocks: price runs one side of a range, closes back inside, and heads for the other side. Last, a second chart and an entry. Every replay runs on simulated candles, read by the rule beside it.
00 · Before ICT
Smart money concepts first
ICT reads the same chart: swings and breaks of structure, sweeps, gaps and blocks, premium and discount. The pages below lean on these four: liquidity sweep · fair value gap · order block · premium and discount.
Start with smart money concepts
01 · Time
ICT killzones
Fixed windows on the New York clock: Asia 20:00–00:00, London 02:00–05:00, New York AM 08:30–11:00 and New York PM 13:30–16:00. They move with New York's daylight saving time, so in UTC they shift an hour for part of the year.
Learn ICT killzones
ICT killzones: four windows on the New York clock
One simulated day: the Asia, London, New York AM and New York PM windows draw their boxes on the New York clock, with the same moments in UTC and on your own clock below.
02 · One day
AMD: the power of 3
A range near the 00:00 open, a fake move out of one side that closes back inside, then the real move out of the other. On the daily candle, the fake move is the wick and the real one is the body.
Learn AMD (power of 3)
AMD: the power of 3 candle and its intraday path
One simulated day ranges near the 00:00 open, fakes out of one side and runs out of the other; beside it, the daily candle it makes: the fake move is the wick, the run is the body.
03 · The night
Judas swing
Between 00:00 and 05:00 New York, price runs one side of the Asian range (20:00–00:00), closes back inside within the hour, then closes through the 00:00 open the other way. No close back inside: a breakout.
Learn Judas swing
ICT Judas swing: the fake move at the open
One simulated night on the New York clock: the Asian range from 20:00 to 00:00, a run beyond one side that closes back inside within the hour, then a close back through the 00:00 open, with the same moments in UTC and on your own clock below.
04 · One hour
ICT Silver Bullet
One hour on the New York clock, 03:00, 10:00 or 14:00. The window runs one side; the first fair value gap toward the other side is the setup, entered when price comes back into it.
Learn ICT Silver Bullet
ICT Silver Bullet: three one-hour windows
One simulated day: the 03:00, 10:00 and 14:00 New York windows shaded on the New York clock, each inside a killzone, with the same hours in UTC and on your own clock below.
05 · Three candles
Candle range theory (CRT)
C1's high and low are the range. C2 trades beyond one side and closes back inside. C3 is expected to run to the other side. Until C2 closes, it could still be a breakout.
Candle range theory: three 4-hour candles
One simulated CRT on 15-minute candles with its 4-hour candles beside it: C1's high and low are the range, C2 trades beyond one side and closes back inside, C3 runs to C1's other side.
06 · 20 days
Turtle soup
Connors and Raschke's fade, which ICT borrowed: a new 20-day low breaks a low at least four sessions old, and a buy stop just above that old low, good for the day, buys it back. A new 20-day high mirrors it.
Learn turtle soup
Turtle soup: the 20-day rule, start to finish
One simulated turtle soup on daily candles, its day played 15 minutes at a time: a new 20-day low breaks an old low made at least 4 sessions earlier, a buy stop 5 ticks back above it fills before the close, the stop goes under the day's low, and the close names the day.
07 · Two markets
SMT divergence
Two markets that move together disagree at the same swing: one makes a lower low and the other doesn't (bullish), or the mirror with highs. The two swings sit within 2 candles of each other.
Learn SMT divergence
SMT divergence: one lower low, one higher low
Two simulated markets that move together, on two stacked charts: both print the same swing low, then one makes a lower low and the other a higher low (bullish), or the same with highs (bearish).
08 · Entry
Optimal trade entry (OTE)
The 62% to 79% retracement of the leg that last broke structure, with 70.5% inside: past the 50% line, deep in discount for a long.
ICT optimal trade entry: Fibonacci on the leg
A simulated leg breaks structure; once its extreme is confirmed, a Fibonacci retracement draws on it and ICT's OTE band, 0.62 to 0.79 with 0.705 inside, lights up before the pullback taps it.
What we measured across ICT
None of the sweep-and-turn rules we tested has a validated edge after costs. The killzone hours sorted nothing; the force of the move did, on one market.
Two ICT ingredients, scored on zones
Our toolkit's 0–100 zone score, on 1,729 order blocks and fair value gaps from 50 sessions of gold futures (5-minute candles). Points: how much more often the top third of zones was respected than the bottom third. p: how often shuffled scores did as well. Thinned: neighbouring zones from one impulse removed.
| What we tested · result | Verdict |
|---|---|
| Born inside a killzone (the toolkit's two windows, read on the New York clock)+1.5 pts, p = 0.30 | Nothing found |
| Displacement: how hard the move that made the zone ran+18.0 pts thinned, p below 0.0005 | Passed, one market |
Source: zone score study
Sweep-and-turn rules, traded with costs
The closest tests we have to AMD, the Judas swing, CRT and turtle soup: a raid of a level, a close back inside, a move the other way. Each is a written rule run with commissions and slippage; none is one of those models exactly.
| What we tested · result | Verdict |
|---|---|
| Gold Sweep Desk: a raid of the Asia range, the London range or yesterday's high or low that closes back inside, then a close through the nearest opposing gapChosen on January to April: 40 trades, profit factor 1.13. Then May to June: 15 trades, −11.63R, profit factor 0.14. | No validated edge |
| Fractal Model: a 4-hour or daily candle sweeps the last one and closes back inside it, then a lower-timeframe close confirms (CRT's skeleton plus one condition)108 versions, 7 markets, 2018 to 2026: no validated edge after costs. Before costs, a coin flip. | No validated edge |
| Sweep Desk v1.0: sweeps allowed only inside the killzonesSeven trades, profit factor 0.93: too few to read. | Nothing found |
Sources: Gold Sweep Desk lab notes, Fractal Model study, Sweep Desk lab notes
Random prices, read by each page's rule
Our simulated markets have no orders, no stops and nobody behind them. Every count depends on how we built the simulation.
| What we tested · result | Verdict |
|---|---|
| AMD30 of 363 simulated days printed AMD, a count that says nothing about real markets. | Random prices print it |
| Judas swing74 of 291 simulated nights printed one, and on 56 of those price took the swing's extreme again before 08:00: counts that say nothing about real markets. | Random prices print it |
| CRT293 of 1,964 simulated 4-hour candles printed a CRT, and C3 reached the other side in 87 of them: counts that say nothing about real markets. | Random prices print it |
| Turtle soup214 turtle soups in ten simulated years, 84 of them stopped out the same day and 115 never filled: counts that say nothing about real markets. | Random prices print it |
| SMT divergence372 divergences in 3,247 swing comparisons over 364 simulated days, a count that says nothing about real markets. | Random prices print it |
| OTEThe pullback reached 0.62 in 503 of 1,812 simulated legs and went on past 0.79 in 354: counts that say nothing about real markets. | Random prices print it |
Source: each model's own page, linked in the path above.
Each test is one market or one window, and none is an ICT model exactly as its page writes it. We haven't tested killzones, AMD, the Judas swing, the Silver Bullet, CRT, turtle soup, SMT divergence or OTE as standalone rules.
Tools that draw these
Only what we checked in each tool’s Pine source. Where nothing we make draws a concept, the row says so.
| Concept | What draws it |
|---|---|
| Killzones |
|
| AMD | Nothing we make draws it whole. These draw its parts:
|
| Judas swing | Nothing we make draws it whole. These draw its parts:
|
| Silver Bullet |
|
| CRT | Nothing we make draws it whole. The closest:
|
| Turtle soup | Nothing we make draws it whole. These draw the levels it raids:
|
| SMT divergence | Nothing we make draws it whole. These can require an SMT check against a second market, off by default:
|
| OTE | Nothing we make draws it whole. None draws Fibonacci levels. These draw the leg:
|
Questions traders ask
What are ICT concepts?
The trading vocabulary taught by ICT, the Inner Circle Trader: time (killzones and one-hour windows on the New York clock), liquidity (the highs and lows price runs before it turns), displacement (a fast move of big candles that leaves fair value gaps) and PD arrays (gaps and blocks read in the premium or discount half of a range).
Is ICT the same as smart money concepts?
Mostly the same vocabulary: liquidity, order blocks, fair value gaps, premium and discount. ICT adds the clock: killzones, the Silver Bullet hours, the midnight open. Learn smart money concepts first; every model here is built from them.
Does ICT trading work?
We haven't found a tested yes. In our gold study, zones born inside a killzone did no better than shuffled scores (p = 0.30). Our sweep-and-turn rules, the closest tests we have to AMD, the Judas swing and CRT, found no validated edge after costs. Use the definitions to read a chart, and test any rule on your own market.
What should I learn first in ICT?
The clock: ICT killzones, on New York time. Then the same sweep on five clocks: one day (AMD), the night (Judas swing), one hour (Silver Bullet), three candles (CRT) and 20 days (turtle soup). SMT divergence and OTE come last: a check on a second chart, and an entry.
What time zone does ICT use?
New York time, America/New_York, daylight saving time included. On UTC and most other clocks every window moves an hour for part of the year: the London killzone, 02:00–05:00 in New York, is 07:00–10:00 UTC from November to March and 06:00–09:00 UTC from March to November.
What win rate does an ICT strategy have?
We can't tell you, and we won't quote one. The win rates in the guides we read, 55–65% for the Silver Bullet and 60–70% for turtle soup, show no trades behind them. We haven't tested any ICT model as a standalone rule.
The terms in our definition, as innercircletrader.net defines them, checked 2026-10-04: ICT concepts and displacement (updated 9 May 2026); PD array (“Premium and Discount Arrangement”, updated 7 May 2026).
The studies behind the numbers
Do Order Blocks Work? We Backtested Our Own SMC Zone Score on 1,729 Zones
Our SMC toolkit stamps every order block and FVG with a 0–100 score. We ported its exact rules to a test rig, ran 1,729 zones from 50 sessions of real gold-futures data, and shuffled the scores 2,000 times to see if the grade beats chance. The composite doesn't — and two of its five ingredients absolutely do. Full design, every number reproducible from committed code, and what we're changing because of it.
Gold Sweep Desk: What We Measured Before Release.
Two pre-registered studies of a gold liquidity-sweep rule on 2026 data: every candidate, every period we ran, and why no setting became a tested default.
We turned the TTrades-style Fractal Model into rules and tested 108 versions on 7 markets. Before costs, it was a coin flip.
C2 sweeps C1, the lower timeframe confirms with a CISD, the candle closes back: we coded the public model as one mechanical rule, wrote the test down before running it, and ran it on 8¾ years of gold, FX, indices and Bitcoin. Every number, the losing ones first.
We Built a Liquidity Sweep Engine. Then We Measured Both Confirmations.
A thread said finding the liquidity is easy and the entries and exits are where the gold is. So we built the liquidity sweep sequence with a real stop and both targets, ran it with costs on gold and bitcoin, and published every number, including the ones that lose.