Accumulation, manipulation, distribution โ usually shortened to AMD, and also taught as the "power of three" โ is a framework that describes a trading session or period in three phases: a quiet range where positions are built, a false move that pushes price out of that range and reverses, and a directional move that runs to the opposite side.
The framework comes from the Inner Circle Trader material popularised by Michael J. Huddleston, and draws on older Wyckoff ideas about accumulation and distribution by large operators. The vocabulary is newer than the concept.
The description is intuitive and matches many charts. It is also, as this page will get to, far easier to apply to a finished chart than to a live one โ which is the single most important thing to understand before trading it.
The three phases
Accumulation. A period of low range and low momentum. Price moves sideways, often overnight or in a quiet session, and a range establishes with a clear high and low. In the framework's reading, this is where larger participants build positions without moving price against themselves.
Manipulation. Price breaks out of the range โ and the break fails. The framework reads this as deliberate: the move takes out the stops resting beyond the range and fills the remaining orders needed, then reverses. This is also called the judas swing, and it is the phase the whole model turns on.
Distribution. Price moves decisively in the opposite direction to the failed break, running through the other side of the range and beyond. This is the move the framework says the first two phases were preparing.
The template is a false break followed by a real move the other way โ a structure that traders have described in many vocabularies over the years. What AMD adds is a specific narrative about why, and a session-based timing overlay.
The session overlay
The framework maps the phases onto trading hours, most often for index futures and major forex pairs:
| Phase | Typical window (New York time) | Description |
|---|---|---|
| Accumulation | Asian session into early London | Low range, level building |
| Manipulation | London open, or the New York open | The false break |
| Distribution | London and New York sessions | The directional move |
There is a real observation underneath this. Liquidity and participation genuinely do vary by session, ranges genuinely are narrower in Asian hours, and volatility genuinely does expand around the London and New York opens โ that part is measurable and not controversial.
What is an interpretation, not a measurement, is that the expansion is a deliberate manipulation designed to trigger stops before the real move. Volatility around a session open has simpler explanations: participants arriving, order flow concentrating, and economic releases clustering in those windows. The economic calendar explains a large share of the moves the model attributes to intent.
What is measurable and what is narrative
Being clear about this makes the framework more useful, not less.
| Claim | Status |
|---|---|
| Ranges are narrower in the Asian session | Measurable, generally true |
| Volatility expands around session opens | Measurable, generally true |
| Stops cluster beyond obvious range highs and lows | Reasonable inference from how traders place stops |
| Price often breaks a range and reverses | Observable, and happens constantly |
| A specific actor engineers this to fill orders | Narrative, not verifiable from a chart |
| The three phases occur in a reliable daily sequence | Not established; many sessions do not fit |
The trading logic survives without the narrative. "Ranges get broken and the break often fails, especially around session opens when volatility expands" is enough to build a rule on, and it does not require believing anything about intent.
Trading the structure
Mark the range. Define the accumulation window by clock time before the session, not after. The high and low of that window are the levels. Choosing the window afterwards is where hindsight enters.
Wait for the break and the failure. The setup is not the break. It is price trading beyond the range and then closing back inside โ a rejection. Entering on the break itself means taking every break, most of which continue.
Entry and stop. Entry commonly comes on the reclaim of the range boundary or on a lower-timeframe structure shift after the sweep. The stop belongs beyond the extreme of the false break, plus a buffer. That extreme is the invalidation: if price goes back through it, the break was not false.
Target. The opposite side of the range is the standard first objective, with extensions beyond it if the session trends.
The stop placement has a practical consequence worth stating. A manipulation leg can be large โ that is what makes it effective at taking stops โ so the distance from your entry to beyond its extreme is often much wider than a normal intraday stop. Sizing by habit rather than from that distance is how a correct read becomes an oversized loss. Fix the stop from the structure, then compute the size from your risk percentage using the lot size calculator; for index traders the NASDAQ 100 and S&P 500 pages use the correct point values.
The hindsight problem
This is the central weakness, and it applies to the whole framework rather than to any one setup.
On a completed chart the three phases are obvious. You can see which range was the accumulation, which push was the manipulation and which move was the distribution, because you can see how it ended. Live, at the moment of the break, you cannot tell a manipulation from a genuine breakout โ they are the same event until one of them reverses.
The framework's flexibility compounds this. If price breaks up and reverses down, that is manipulation. If price breaks up and keeps going, that can be relabelled as distribution with the accumulation placed earlier. A model that has an explanation for every outcome cannot be wrong, and a model that cannot be wrong cannot be tested.
The only way to get honest information out of it is to constrain it in advance: fix the accumulation window by the clock, fix what counts as a failed break, fix the entry trigger and the invalidation, write them down, and then measure. Under those constraints it becomes a testable rule about range breaks failing near session opens โ which is a real thing to study. Backtesting a specified rule tells you something; reviewing charts and labelling the phases afterwards tells you only that you can label charts.
How it fails
The break is real. Price leaves the range and trends. There was no manipulation phase; there was a breakout.
No range forms. The accumulation window is not quiet, and there is no clean high and low to work from. Forcing a range onto a directional session produces arbitrary levels.
Both sides get swept. Price takes out the high, reverses, takes out the low, reverses again. Traders positioned after the first sweep are stopped, and so are those positioned after the second.
The move comes from news. A release during the window produces the expansion, and the direction has nothing to do with the range structure. The economic calendar is the check on this before the session, not after.
Relabelling. The phases are reassigned after the fact so the framework still describes the day. This feels like understanding and produces no usable information.
Making it testable
Log each AMD trade with the accumulation window you used, the range in pips or points, which side was swept, the time of the sweep, the stop distance, and the result in R. Two questions matter more than the rest: whether your results differ between London-open and New York-open setups, and whether the days you skipped because no clean range formed would have been losers.
Fips's trading journal records trades in R and groups them by setup so those comparisons come out of a filter, and account analysis shows the distribution rather than the average โ with a framework this dependent on session context, the average conceals more than it reveals.
The concepts here connect directly to the wider ICT framework, and the manipulation leg is often described in terms of the fair value gap left behind by the reversal.
Frequently asked questions
What does AMD stand for in trading?
Accumulation, manipulation, distribution: a quiet range where positions are built, a false break that takes out stops beyond it, and a directional move to the other side. It is also taught as the "power of three".
Is AMD the same as Wyckoff?
It draws on Wyckoff's accumulation and distribution concepts but is not the same. Wyckoff described multi-week campaigns by large operators with a detailed phase structure; AMD compresses a similar idea into a single session and adds the manipulation leg as an explicit phase.
Can the phases be identified in real time?
The accumulation range can, because it is defined by a clock window. The manipulation leg cannot be identified until it has reversed โ before that it is indistinguishable from a breakout. This is the framework's main practical limitation.
Does AMD only apply to a daily session?
It is most often taught on a daily session cycle, but the same template โ range, failed break, move the other way โ is applied on weekly and intraday scales. The smaller the scale, the more occurrences and the more noise.
Is the manipulation phase really deliberate?
That is the framework's interpretation, and it is not something a chart can establish. The observable facts are that stops cluster beyond obvious levels and that volatility expands at session opens. Whether any participant acts with the intent the model describes is not verifiable from price data.