At a glance
The short version
Power of 3 says every daily candle is built in three moves: accumulation near the open, a manipulation move the wrong way (the Judas swing), then distribution in the real direction. On a finished chart it looks true almost every day. We measured it on five years of our own data to find out which part is real and which part is just what a daily candle looks like.
Power of 3, stated so a replay can score it
Take one day. The open is the reference. On a bullish day, price first trades below the open (the Judas swing takes the sell stops under it), then turns and spends the rest of the day building the candle's body upward, closing near the high. A bearish day is the mirror: a run above the open first, then the real move down. Accumulation is the quiet stretch around the open where neither side has committed.
ICT traders anchor this on one of two opens: the midnight New York open or the 18:00 New York open that starts the futures and forex day. Pick one and keep it. The statistics below use the 18:00 open, because that is where the daily candle actually begins on the charts most people trade.
For a backtest, the claim reduces to three things you can write down before the day happens: which open you use, how far below (or above) it a move has to go to count as manipulation, and what has to happen for the manipulation to be "done". Everything else is hindsight.
What five years of data says
We took every New York trading day from 20 September 2021 to 18 September 2026 on NAS100, gold and EURUSD (about 1,290 days each, hourly bars) and asked when each day's high and low formed.
| Measured | NAS100 | Gold | EURUSD |
|---|---|---|---|
| Up days where the low came before the high | 88.9% | 89.6% | 88.2% |
| Down days where the high came before the low | 89.5% | 87.9% | 89.6% |
| Up days whose low was in before 10:00 New York | 79.6% | 93.0% | 91.1% |
| Down days whose high was in before 10:00 New York | 80.0% | 90.2% | 91.1% |
| Typical dip below the open on an up day (share of the day's range) | 24% | 21% | 21% |
| Up days whose low printed in the first hour (no Judas swing at all) | 21.7% | 22.1% | 17.3% |
New York trading day 18:00–17:00, open = 18:00. "Typical" is the median. Measured on CRTLAB's own hourly price history.
Why that table flatters Power of 3
Nine up days in ten made their low first. That sounds like proof, and it is mostly arithmetic. A day that closes near its high has to have made its low earlier. It could hardly be otherwise. So "the low came first on bullish days" is true of almost any definition of a bullish day, with or without ICT.
The more useful rows are the other ones. On NAS100, four up days in five had their low in before 10:00 New York, and on gold and EURUSD it was more than nine in ten. The move that sets the day's extreme usually happens early. That is the part of Power of 3 worth building a rule on: if you are going to catch the Judas swing, it is going to be in London or the first half hour of New York, not after lunch.
And about one up day in five never had a Judas swing to speak of: the low was in inside the very first hour and price never went back. A rule that waits for manipulation below the open misses those days entirely, which is fine as long as your backtest counts them as missed rather than forgetting them.
The tradeable version is rarer than the videos suggest
The table describes finished days. The trade has to be taken during the day, before you know which kind it is. So we tested the textbook Judas swing directly: before 10:00 New York, price drops at least 20% of the previous day's range below the open (without first running that far above it), and by 10:00 it is back above the open.
Over five years that exact sequence happened on 40 NAS100 days, 27 gold days and 27 EURUSD days, a few percent of all days. On those days NAS100 closed up 57.5% of the time against a 53.8% base rate for any day; gold 59.3% against 52.7%; EURUSD 51.9% against 48.8%. The mirror setup (a run above the open that is rejected by 10:00) closed down 53.1% of the time on NAS100, 62.5% on gold and 57.7% on EURUSD.
That is a small lift on a small sample. At 40 trades, a win rate can easily be fifteen points away from the truth, which the sample size calculator will show you. What was consistent: after a reclaimed Judas swing, NAS100 went on to take out the pre-10:00 high on 85% of those days. The day's direction is a coin with a slight bias; the next pool of liquidity is a much better bet. That is a target rule, and it is testable.
None of this says Power of 3 doesn't work. It says the version most people trade is defined too loosely to know. Your version, with your open, your threshold and your entry trigger, is the only one worth measuring.
How to backtest Power of 3 in CRTLAB
The button on this page opens NAS100 on a 5-minute chart with Key Levels and Liquidity Sweeps already on. Key Levels draws the midnight open by default; switch on the true day open (18:00 New York) in its settings if that is your anchor. Sweeps mark the moment price runs a prior high or low and comes back, which is the Judas swing, detected candle by candle without looking ahead.
The HTF candle beside the price scale is the other half. Set it to Daily in its settings and you watch the day's candle build as you replay: the wick forming under the open, then the body. It is the clearest way to see Power of 3 happen in real time instead of reading it off a finished chart.
Then log every day, not just the good ones: which open, how far the manipulation went, whether and when it was reclaimed, whether you entered, the result in R, and whether there was an 8:30 or 10:00 release. The empty days are the denominator. Without them you have a highlight reel.
Where it fits with the rest of ICT
Power of 3 is a frame for the day, not an entry. The entries people pair it with are the ones with their own pages here: the Silver Bullet inside the 10:00 hour, a fair value gap left by the move off the Judas low, or a market structure shift on the lower timeframe. The kill zones tell you when the manipulation usually happens, and the data above agrees with them: early.
If you trade CRT, you will recognise the shape. A higher-timeframe candle that sweeps the previous candle's range and closes back inside is Power of 3 on a single candle, seen from one timeframe up.
Frequently asked questions
What is Power of 3 in ICT?
The idea that each daily candle forms in three phases: accumulation around the open, manipulation (a false move against the day's direction, the Judas swing) and distribution (the real move, which builds the candle's body). On a bullish day that is open, low, high, close in that order; on a bearish day, open, high, low, close.
What is the ICT Judas swing?
The manipulation leg of Power of 3: an early move against the day's eventual direction that runs the stops beyond the open or a nearby high or low, then reverses. It usually happens in the London session or the first part of the New York session. On our data, the low of an up day was in before 10:00 New York on about 80% of NAS100 days and over 90% of gold and EURUSD days.
Does Power of 3 actually work?
The shape is real but partly automatic: any day that closes near its high made its low earlier. The tradeable version (a Judas swing below the open that is reclaimed before 10:00) is rare, about 27–40 days in five years on the markets we tested, and the day-direction edge it gave was small. Taking out the pre-10:00 high after a reclaim was far more consistent (85% on NAS100). Test your own definition before trading it.
Which open does Power of 3 use, midnight or 18:00?
Both are taught. The midnight New York open is the classic ICT reference; the 18:00 New York open is where the futures and forex day actually begins. The difference matters most on days with a big overnight move. Pick one, write it into your rules, and never switch mid-sample.
Can I backtest Power of 3 for free?
CRTLAB is free to try. You get NAS100, gold, EURUSD and every other market, with key levels, liquidity sweeps and a live daily HTF candle on the chart, with every tool the product has, for your first 10,000 candles of replay — roughly two hours of stepping through a chart, no card needed. After the sample, unlimited replay is $14.99 a month, cancel any time.
What timeframe is best for Power of 3?
The frame is the daily candle; the execution is usually on the 1 to 15-minute chart. Replay on the timeframe you would actually enter on, with the daily HTF candle beside price so you see which phase you are in.