ICT Change in State of Delivery (CISD) Explained

A market can begin reversing a short sequence of price delivery before it breaks the swing high or low you are watching. If you treat every confirmation as a market structure shift, that earlier change is easy to miss. If you react to every opposite-colour candle, however, you can mistake ordinary fluctuation for a meaningful reversal.

ICT Change in State of Delivery, or CISD, provides a more specific reference: the opening price associated with the preceding directional candle sequence. The question is whether price has moved back through that reference in the opposite direction. It is a different question from whether an established swing extreme has been broken.

This guide explains how to select the reference consistently, distinguish a wick crossing from a completed-close confirmation, and connect the observation to an executable trade plan. The worked EUR/USD examples include a successful bullish retest and a failed bearish retest. They are constructed teaching examples, not historical results, live signals, or evidence of a guaranteed edge.

1. What is ICT Change in State of Delivery?

In ICT’s 2022 Mentorship Episode 3, the discussion around 24 to 25 minutes connects order blocks with changes in delivery. He identifies the opening price that begins a consecutive series of up-close candles, then discusses price moving below it. The bullish illustration reverses the idea, using the opening associated with the preceding down-close delivery.

The useful chart reference is therefore an opening price, not automatically the high or low of a candle. The original explanation discusses trading through that opening. The completed-close requirement used throughout this guide is an explicit confirmation filter for the study model. Stating that choice matters because an intrabar crossing and a candle that finishes beyond the level can produce different signal times.

What does “delivery” mean in practical chart work?

Within ICT’s framework, delivery describes the direction in which price is being offered and the price references it may seek next. A trader may interpret a shift away from a swept low as the beginning of a move toward higher liquidity. The chart itself shows the prices and their sequence; it does not reveal every institution’s positions or instructions.

You can use that framework without making claims the data cannot support. “Price closed above the selected down-sequence open” is observable. “A particular bank has finished accumulating” is not established by the same candle. Keeping the distinction clear makes your annotations easier to test and your trade review more useful.

A local change is not a promise of a daily reversal

A bullish CISD on a five-minute chart can represent a small recovery inside a bearish hourly move. Its significance depends on the surrounding range, session, and objective. Do not automatically promote a local signal into a new weekly bias. Define the expected destination before choosing how far to hold a trade.

2. How to select the reference opening price

Start with the directional sequence that leads into the extreme you are studying. A down-close candle closes below its own open; an up-close candle closes above its own open. These definitions concern each candle’s body direction. They do not mean merely that one closing price is lower or higher than the preceding candle’s close.

For a clean bullish example, identify the consecutive down-close candles that deliver price into the selected low. Mark the opening of the first candle in that uninterrupted sequence. For a clean bearish example, do the same with the consecutive up-close candles into the selected high. Preserve both the opening reference and the sequence’s extreme, because they serve different purposes.

ICT’s 2025 lecture on the price delivery continuum, around 32 minutes, also discusses the lowest opening price of consecutive up-close candles in a bearish order block. In a clean, continuous upward series, that is normally the opening that starts the run. The article’s examples use such clean sequences so the starting-open and extreme-open descriptions coincide.

Declare how you handle interruptions

For this guide’s reproducible model, an opposite-colour candle or a doji ends the consecutive run. A doji means an equal open and close at the feed’s quoted precision. We do not join separated runs because one tiny candle looks unimportant after the reversal succeeds. This is a study convention, not a claim that every discretionary ICT analysis must group candles that way.

If a price gap or unusual feed discontinuity makes the first opening different from the relevant extreme opening, exclude the sequence from this simple implementation. A more advanced version can define how to handle that case before testing it. Do not alternate between the earliest open, highest open, lowest open, and final candle’s open to obtain whichever signal looks best.

One qualifying candle can supply an opening reference; three candles are not a universal requirement. Longer runs simply make selection discipline more visible. Once the chosen extreme has formed and the opposite response begins, freeze the reference for that candidate. A later new extreme creates a revised candidate under a new timestamp, not a silent edit to the original record.

3. CISD versus MSS, order blocks, and displacement

A market structure shift examines a relevant swing high or low in its context. The CISD implementation here examines the opening of a selected candle sequence. Those levels can be close together, far apart, or crossed by the same candle. Their relative position determines which observation appears first.

Consider a down-close sequence beginning at 1.08318 while an earlier internal swing high stands at 1.08361. A candle that closes at 1.08329 has reclaimed the opening reference but has not closed above that swing high. If your rules require a swing-based MSS, the CISD observation alone does not satisfy them.

The relationship to an order block is closer. The original delivery discussion helps explain why a selected candle opening can remain relevant after price moves away. The CISD marks an event at that reference, while an order-block annotation describes the associated candle or grouped range. Do not count them as completely independent confirmations when they arise from the same price sequence.

Displacement describes the character of a move, such as an expansive directional candle with limited overlap. A fair value gap has its own three-candle geometry. Neither is automatically created every time price crosses an opening reference. If your trading model requires displacement and an FVG as additional filters, define and check them separately.

Use the site’s MSS guide for the swing-based framework and the order block guide for the related price array. Here the priority is to keep the opening reference, signal time, and entry time distinct.

4. Why context matters more than the number of CISD labels

In a congested range, price may cross several candle openings in both directions. A chart can accumulate many technically valid observations while offering little useful room between entry and objective. More labels do not necessarily mean more opportunities. Context is the process of deciding which candidate deserves evaluation.

For the examples in this guide, a relevant earlier high or low is marked before the setup. Price moves beyond that level and returns, creating a possible liquidity-sweep context. We then study the delivery sequence into the new extreme. A sweep is an eligibility condition for these examples, not part of the arithmetic that defines an opening price.

Higher timeframe areas can strengthen the reason for studying a particular location, but they need to be specified. Record the actual order block, FVG, or range boundary in your preparation notes. A vague statement that the setup agrees with the higher timeframe becomes difficult to audit after a loss.

Choose a destination that still offers room

A bullish opening reclaim directly beneath an already reached liquidity objective may offer little remaining reward. A bearish reclaim with a nearby higher timeframe bullish reference below it may face an obstacle before the proposed target. Assess the path and distance before sizing the position, rather than placing a target simply to manufacture an attractive ratio.

The institutional order flow interpretation is a hypothesis about that path. The BIS discussion of FX execution algorithms describes fragmentation and limited visibility across execution venues. One practical implication is that candles on a single retail feed cannot establish a complete map of institutional orders. Treat liquidity locations as inferred references and confirm actual order behaviour on your own account.

Context also defines scope. A five-minute bullish CISD may justify studying a move to an internal high while leaving the broader bearish bias intact. You do not need to claim a complete market reversal to test a local opportunity. You do need a clear endpoint and a reason to stop holding if the response fails.

5. How to identify a candidate without hindsight

1. Mark the location first. Record the earlier high or low and the higher timeframe reference that make the area relevant.
2. Identify the directional run into the extreme. Use the same body-direction and interruption rules on every candidate.
3. Save the opening reference and extreme. Include their prices and candle timestamps. Keep the internal swing reference separate.
4. Observe the first crossing. Record a wick through the opening if one occurs, but do not promote it to a completed-close signal.
5. Wait for the confirmation close. In this model, the candle must finish strictly beyond the opening, not exactly on it.
6. Evaluate the trade independently. Calculate the proposed entry, stop, target, costs, and expiry before submitting an order.

A common error is drawing the reference only after the reversal candle appears, then choosing the shortest run it has already crossed. Freeze the source candles when the candidate becomes identifiable. If you cannot explain why that run was selected before the outcome, the annotation is a retrospective description rather than a reproducible setup.

Another error is using a higher timeframe close before it exists. At 8:12 AM, a five-minute candle ending at 8:15 AM is still forming. Its temporary price above a reference can disappear by the close. Store the platform’s candle-label convention, because some charts label a bar by its opening time while your notes may refer to its ending time.

Keep the timeframe fixed for the first study. A one-minute chart can divide a five-minute body into several alternating candles and produce a different reference. That is a different observation, not proof that one chart is wrong. If a multi-timeframe rule is introduced later, state which timeframe supplies context and which supplies the confirmation.

6. A complete trading plan after confirmation

The worked examples use five-minute candles and require a completed close beyond the frozen opening reference after a marked liquidity sweep. They then place a limit order at the reference opening for a later retest. No entry is credited to a wick that occurred before the confirmation candle closed.

The protective stop is 1.3 pips beyond the selected sweep extreme. The target sits slightly before a premarked liquidity objective. The minimum planned reward-to-risk is 2R after the stated cost allowance. These are explicit educational settings, not claims that one buffer or ratio is optimal for all pairs and conditions.

An unfilled order expires after two additional five-minute candles or at the session cutoff, whichever comes first. Cancel it if the target is reached before entry, if a new adverse sweep extreme occurs, or if the execution filters fail. Allow one attempt per selected candidate. A losing attempt does not automatically authorise another order at the same level.

Keep management consistent with the chosen model

Once filled, use one full-position target and one protective stop. If neither is reached by the session cutoff, exit at the available executable price and record the actual result. A temporary close back across the reference is recorded but is not an additional exit rule in these examples. Adding that exit would create a different model and require a separate comparison.

For this study, the monitored spread must be at or below 0.8 pip at placement and while the order is pending. Exclude new orders during the ten minutes before and fifteen minutes after a scheduled high-impact announcement affecting either currency, and cancel pending orders as that exclusion period begins. Those filters reduce a defined category of exposure; they do not guarantee normal execution at other times.

The setup does not require a later MSS before entry. That is a deliberate distinction. A trader who wants both a CISD and a completed swing break should wait for both and recompute the entry opportunity afterward. A swing break that occurs after an order fills cannot be counted as evidence that authorised the earlier entry.

7. Complete bullish EUR/USD example: CISD before the swing break

This hypothetical winter London setup uses Ghana GMT. All times in the example identify completed five-minute candle closes. Preparation has identified a bullish higher timeframe area, an earlier low at 1.08264, an internal swing high at 1.08361, and a possible buy-side liquidity objective at 1.08584. These references exist before the entry sequence develops.

Freeze the source sequence

The candle closing at 8:00 AM opens at 1.08318 and closes down at 1.08296. The 8:05 AM candle opens at 1.08296 and closes at 1.08276. The 8:10 AM candle opens at 1.08276, trades below the earlier low to 1.08247, and closes at 1.08259. All three are down-close candles.

The reference is the first candle’s open, 1.08318. The sweep extreme is 1.08247. They are not interchangeable. The first is the delivery reference to be reclaimed; the second supplies the stop reference for this particular trade model.

At 8:15 AM, an upward response reaches 1.08326 but closes at 1.08307. Its wick crossed the opening reference, while its close finished 1.1 pips below it. Under the completed-close filter, no order is authorised. A trader testing an intrabar-crossing model would record an earlier event, but that is not the method used here.

Confirm, then wait for a later retest

The 8:20 AM candle closes at 1.08329, strictly above 1.08318. This confirms the bullish CISD under the article’s rule. Its high is 1.08341, still below the earlier internal swing high of 1.08361. The selected swing has therefore not yet been broken by this candle.

After that close, place a hypothetical buy limit at 1.08318. The stop is 1.08247 − 0.00013 = 1.08234, and the target is 1.08570, 1.4 pips before the premarked liquidity objective. Assume the session and execution filters pass and the target has not already been reached.

The candle closing at 8:25 AM trades down to 1.08309, allowing the later retest entry in the simplified reference-price sequence. The 8:30 AM candle then closes at 1.08372, above the old internal high. The swing-based confirmation is later than the entry in this path. It is additional information available during the trade, not a condition that existed at order placement.

The constructed continuation reaches 1.08577 during the candle closing at 8:50 AM, passing the target at 1.08570. The example closes the entire position at its planned target in the simplified calculation. It does not hold for the more distant liquidity reference just because the next few prices might look favourable.

Translate price risk into account risk

On a hypothetical US$5,000 account, a selected risk budget of 0.3% equals US$15. With a standard EUR/USD contract of 100,000 euros, pip value in a US dollar account is US$10 per standard lot. Using the 9.3-pip cost-adjusted loss estimate, position size is US$15 ÷ (9.3 × US$10) = 0.16129 standard lot.

If the broker permits 0.01-lot increments, rounding down to 0.16 lot gives an estimated loss of 9.3 × US$1.60 = US$14.88 and target profit of 24.3 × US$1.60 = US$38.88. The cost allowance combines spread, commissions expressed in pips, and assumed execution friction. Slippage beyond that estimate can increase the actual loss.

The 1.08318 reference does not become invalid merely because the entry candle briefly trades below it. The predetermined stop is at 1.08234. If your plan instead exits on a close back below the opening, state that before entry and calculate its outcomes separately. Mixing the two exit methods after seeing the path would make the example impossible to evaluate fairly.

8. Bearish worked example: a confirmed CISD can fail

On a separate hypothetical winter New York session, an up-close run develops near a bearish higher timeframe reference. Its first candle, closing at 12:15 PM Ghana GMT, opens at 1.09146 and closes at 1.09171. The next two candles close at 1.09203 and 1.09216. The final candle’s high reaches 1.09238, beyond an earlier high of 1.09224.

The clean sequence begins at 1.09146, which becomes the bearish delivery reference. A previously marked internal swing low is 1.09119. The intended lower liquidity objective is 1.08817. As in the bullish example, the opening reference, risk extreme, structural swing, and destination serve different roles.

The 12:30 PM candle trades down to 1.09137 but closes at 1.09158, so its wick crossing does not pass the close filter. At 12:35 PM, the next candle closes at 1.09131. It has finished below the opening reference, confirming the bearish CISD for this model, while its low of 1.09123 remains above the selected swing low.

After confirmation and the execution checks, place the hypothetical sell limit at 1.09146. The stop is 1.09238 + 0.00013 = 1.09251. The target is 1.08831, 1.4 pips above the premarked lower objective. Price risk is (1.09251 − 1.09146) ÷ 0.00010 = 10.5 pips; price reward is (1.09146 − 1.08831) ÷ 0.00010 = 31.5 pips.

The gross ratio is 31.5 ÷ 10.5 = 3R, or 1:3. With a 0.9-pip round-trip allowance, the estimated loss is 11.4 pips and estimated target profit is 30.6 pips. The adjusted ratio is 30.6 ÷ 11.4 = 2.68R, which passes the model’s minimum.

The candle closing at 12:40 PM reaches 1.09165, passing the sell limit in this simplified price sequence. Instead of continuing lower, the next candle rises to 1.09258, beyond the stop at 1.09251. The result is an estimated 11.4-pip loss including the allowance. No later move can turn that already stopped trade into a winner.

The setup’s failure does not mean the earlier close never happened. The CISD signal was present under the declared rule, and the trade lost. Keep both facts. The selected swing low was not broken before entry, so a separate model requiring an MSS would have made a different decision. That difference should be compared across a complete sample rather than used to relabel only this losing chart.

9. Ghana timing, chart feeds, and executable prices

Ghana uses GMT year-round. New York changes between UTC−5 during standard time and UTC−4 during daylight time. Session windows anchored to the New York clock therefore move by one hour on a Ghana clock. Keep the date and time convention in every journal entry.

The familiar 7 to 10 AM London window and noon to 3 PM New York window in Ghana are winter conversions of these New York anchors. Check the Ghana time zone reference and New York clock-change reference for the clock rules. A window anchored to London’s local business opening is a different convention, especially around weeks when the two cities change clocks on different dates.

Broker feeds and candle boundaries also matter. Two platforms can construct different one-minute sequences from different quotes or aggregation settings. Use one consistent feed to define the source run and the confirmation. Do not choose the reference on one feed and then use another feed solely because it avoids a stop or supplies a more favourable entry.

Record the quote side used for each calculation

A long opens at the ask and closes at the bid; a short opens at the bid and closes at the ask. A bid-chart wick touching a buy limit does not necessarily mean the ask reached it. Likewise, a short’s stop can execute at the ask even when the displayed bid high appears below the stop.

The worked examples use reference prices plus a separate cost allowance. If your test already uses executable bid-and-ask entry and exit prices, the spread is embedded in those prices. Add only costs that are not already included. The CFTC’s forex customer advisory discusses leverage and dealer-dependent trading conditions; its US regulatory framework does not establish a broker’s authorisation in Ghana. Verify the terms and execution behaviour of the actual account being studied.

10. How to test CISD without changing the rules after a loss

Build a record around the candidate, not just around completed trades. Store the source sequence, its first open, the sweep extreme, the earlier liquidity level, the separate swing reference, and the time at which each became known. Then record the first intrabar crossing, the first qualifying close, order placement, fill or expiry, and final outcome.

This allows you to answer two distinct questions: how often does a delivery reference get reclaimed, and how often does a particular entry plan produce an acceptable result? A signal can occur without a retest. A retest can fill without reaching the target. An order can expire before a later successful move. Those outcomes all belong in the sample.

Compare confirmation filters on the same candidates

If you want to compare an intrabar-crossing rule with a completed-close rule, begin with the same frozen source sequences. Record their signal times separately. Do the same for a version requiring a later MSS. Keep stop definitions, costs, holding periods, and risk budgets explicit so you know what changed between the models.

Waiting for more evidence may avoid some failed early entries, but it can also remove a favourable retest or leave less distance to the objective. Earlier does not mean better, and later does not mean safer in every path. Report fill rate and net outcome per eligible candidate as well as results per filled trade.

Do not infer the order of events from a single candle

If one bar contains the entry, stop, and target, its open, high, low, and close may not establish which exit happened first. Use more detailed timestamped data where available. Otherwise, flag the path as ambiguous or apply a conservative convention chosen before reviewing results. Do not award the target because the candle eventually closes in the expected direction.

The same principle applies to a new adverse extreme before a pending order fills. The cancellation rule matters only if you know when that extreme occurred relative to the fill. A candle that contains both prices requires more detail. A precise-looking entry line cannot make missing intrabar information appear.

Keep revisions visible

If you later decide to ignore small opposite-colour candles, require an FVG, or tighten the stop, date the rule change and retain the older results. Evaluate the revision on a later set of observations. Repeatedly adjusting the source grouping until every past reversal produces a clean signal is a way of fitting the description to the chart, not establishing a useful decision process.

Start with replay and demo records, then examine sensitivity to costs, spread variation, and missed fills. A pattern that appears useful only with ideal execution may not survive actual account conditions. The aim is a repeatable method that admits uncertainty, rather than a chart covered with labels that can explain either direction after it happens.

11. Frequently asked questions

Continue with the guides to fair value gaps, ICT order blocks, and market structure shifts. Use the London Kill Zone guide and Judas Swing explanation to organise session context. Then study CISD with a frozen opening reference, a stated confirmation rule, and a record that preserves both successful and failed attempts.