ICT and SMC concepts explained for algo testing
ICT and SMC concepts such as order blocks, liquidity sweeps, fair value gaps, structure breaks and SMT divergence are discretionary ideas with no single standard definition. To test one, replace every judgement with a number so a computer finds the same pattern a person would, then backtest it with costs. AKQX is not affiliated with any educator and claims no concept is profitable.
Not affiliated. ICT and SMC are discretionary trading concepts taught by many independent educators. AKQX is not affiliated with, endorsed by or connected to any of them, uses none of their trademarks, and makes no claim that any concept is profitable. This guide explains the vocabulary and how to test it. It does not teach anyone's method.
Key takeaways
- These concepts are discretionary and defined differently by different people, so they cannot be tested until you fix a definition.
- Testing means replacing every judgement with a number and applying it mechanically to a long history.
- Hindsight is the main enemy. Every past move has an obvious pattern in retrospect.
- With this many adjustable settings, overfitting is the main risk after hindsight.
- Whether any of it has an edge is an empirical question, and the answer may be no.
Why are these ideas so hard to test?
Most of the teaching around these concepts is deliberately discretionary. A sentence such as “wait for price to return to the order block, with displacement, in the right session, if structure agrees” is something a person can act on and a computer cannot, because each clause hides a judgement: which block, how much displacement, whose idea of structure.
That is also why a trader can spend years on them without learning whether their version works. With no rules there is nothing to measure, so every loss can be explained as a misread and every win as confirmation. Making the rules explicit is not about proving the concepts right. It is about letting the data answer back.
The vocabulary, and the choice each term hides
Each concept below has a short, commonly used meaning and a list of decisions you have to make before it can be tested. The full definitions are in the glossary.
| Concept | Commonly meant as | Decisions you must write down |
|---|---|---|
| Order block | The last opposing candle before a strong move, treated as a reaction zone | Body or full range, how strong the move away must be, what counts as a valid return |
| Liquidity sweep | A brief move through an obvious high or low, then a reversal back inside | Which levels count, how far beyond, how many candles to close back inside |
| Fair value gap | A three-candle pattern whose outer ranges do not overlap | Minimum gap size, whether a partial fill counts, how long it stays valid |
| Break of structure | A close beyond the last swing high or low, in the trend's direction | Swing length, wick or close, which timeframe |
| Market structure shift | A break of a swing level against the trend | The same swing choices, plus what confirms it |
| SMT divergence | Two correlated markets disagreeing at a swing point | Which pair, how a swing is defined, how long the correlation is checked |
| Kill zone | A time window, often a session open, where setups are preferred | Exact times, which timezone, daylight saving |
How to turn one into a rule
Take a single concept and walk it through four moves.
- Write the definition in numbers. Not “a strong move” but “a move of at least 1.5 times the average candle range within three candles”.
- Make it mechanical. Two people, or one person on two different days, applying the rule must mark exactly the same patterns. If they would not, it is still a judgement.
- Make it causal. The rule may only use information available at the moment of the decision. A pattern that needs the next candle to be recognised cannot be traded.
- Attach an entry, a stop, a target and a size. A pattern is not a strategy until it says what to do.
Then apply it to a long stretch of history and count honestly, including the cases that failed. The steps for that are in how to backtest a crypto strategy.
The two traps
Hindsight
Draw the zones on a chart after the move and every rally has an obvious order block, every reversal an obvious sweep. The pattern is only informative if a rule could have identified it before the outcome was known. This is why the causal check above matters more than any other.
Too many settings
Between swing length, move strength, session windows, retest tolerance and filters, a rule built from these concepts can have a dozen adjustable settings. Every one is another degree of freedom for history to be fitted by. Keep the rule small, nudge each setting to check the result degrades smoothly, and test on data the tuning never saw. The full method is in the overfitting guide.
How AKQX approaches it
AKQX provides these ideas as configurable blocks so the definition has to be written down: order blocks, liquidity levels and sweeps, fair value gaps, structure breaks and shifts, SMT divergence, the True Day Open filter, and multi-timeframe confirmation where a slower candle gates a faster entry. It also has two AKQX-specific terms, with their own working definitions: a PSP, a divergence between two correlated pairs whose candles close in opposite directions, and the True Day Open, the 00:00 UTC open used as a daily anchor. Both are described on the ICT and SMC algo builder page.
Results are scored out of 350 points and rejected automatically on negative expectancy, a profit factor below 1.0, a drawdown above 30 percent, or weak walk-forward, Monte Carlo or overfitting scores. A passing score is a description of a simulation. It is not a recommendation and it is not a forecast.
A worked example, with made-up numbers
Illustrative only. A trader defines a liquidity sweep as: price trades at least 0.2% beyond the highest high of the previous 20 candles, then closes back below that high within two candles.
Run across a long history on one timeframe, the rule finds 180 events. Of those, price falls by at least the size of the sweep within the next five candles in 96 cases and does not in 84. That is a 53% hit rate before any costs, on a definition chosen in advance. The honest reading is that 53% over 180 events is close enough to a coin flip that costs alone could erase it, so the next steps are the ones above: try neighbouring values of 0.2% and 20, check a period the rule was not chosen from, and subtract fees and slippage. Whether the concept has an edge in this form is still undecided, which is a perfectly good outcome for a test to produce.
A reminder about what this is. AKQX is software only. It is not a SEBI-registered Investment Adviser or Research Analyst and gives no investment advice. Backtest and paper-trading results are simulated and do not guarantee future results. Crypto products are unregulated and can be highly risky, and there may be no regulatory recourse for any loss.
Questions about ict and smc concepts for algo testing
Can ICT and SMC concepts be backtested?
The parts that can be written as explicit rules can. A concept becomes testable once every judgement in it is replaced by a number: which candle, how far price must move, what counts as a valid return. The parts that depend on a trader's eye cannot be tested, and a test cannot rescue them.
Do ICT and SMC concepts make money?
Nobody can honestly say that in general, and this guide does not. They are discretionary frameworks taught by many independent people, with different definitions. Whether your interpretation has any edge is an empirical question, and the point of testing is to find out rather than assume.
Is AKQX affiliated with any ICT or SMC educator?
No. AKQX is not affiliated with, endorsed by or connected to any ICT or SMC educator or course, uses no one's trademarks, and makes no claim that any concept is profitable. It provides a way to test your own interpretation.
Why do different sources define an order block differently?
Because it is a discretionary idea with no standard definition. Whether the zone is the candle's body or full range, how strong the move away must be, and what makes a return valid are all choices. That is exactly why they must be fixed in writing before testing.
What is the hardest part of testing these ideas?
Hindsight. Every past rally has an obvious order block and every failed breakout looks like a liquidity sweep once you know the outcome. A rule has to identify the pattern at the moment it forms, using only information available then.
Do I need to code to test them on AKQX?
No. The detectors are visual blocks you place and configure: order blocks, liquidity levels and sweeps, fair value gaps, structure breaks, SMT divergence, time filters and AKQX's own PSP and True Day Open filters. You arrange the logic and the platform backtests it.
Keep reading
Test your own interpretation, free
A free account includes 3 saved strategies and 10 backtests a month, with no card and no broker connection.