Four acronyms, one family
All four describe the same thing from different angles: a place or a moment where a large execution is assumed to have left a trace. That is not a criticism, it is the structure of the corpus.
What really separates them is not their definition, it is when you can know whether you are wrong. Some can be checked the moment they appear, others only afterwards. That is the only sorting criterion that matters, and it is rarely stated.
Keep it in mind through the four sections that follow. For each acronym, ask the same question: at what moment can I say I was wrong?
The verifiable base is always the same, the one from the lesson on large orders: a big execution needs counterparty and leaves the trace of it. Everything else is interpretation laid on top.
⚠️ An abundant vocabulary gives the impression of a deep corpus. Above all it makes two methods hard to compare, because each can claim to be about something else. That is one more reason to bring everything back to the same test.
BPR: two scars on top of each other
BPR, balanced price range, names the overlap of two opposing imbalances on the same price area. Price crossed it once one way without negotiating, then once the other way.
Trading it rests on the idea that a double scar holds better than a single one, because two executions left a hole there instead of one. That is a reasonable assumption and it is not a demonstration.
On the chart a BPR is unambiguous: two opposing fair value gaps that overlap, and the zone is their intersection. Two people find the same zone, which is already better than half of this vocabulary.
Its weakness is arithmetic: BPRs are rare. Across three months of hourly history you will count a handful, not dozens. A pattern that does not show up often enough cannot be measured, and a pattern that cannot be measured cannot be improved.
That is the paradox of rare patterns: the more selective they are, the more reliable they look, and the fewer cases you have to check whether they are.
CRT: one candle as a whole range
CRT, candle range theory, applies the sweep idea to a single higher-timeframe candle: its high and its low become the bounds of a range, and everything that happens inside is read between those two levels.
Trading it means watching for one bound to be swept and then a return towards the other. It is a restatement of the liquidity sweep, brought down to the scale of one candle.
It has a merit the others lack: the bounds are known in advance. The high and low of the previous candle are two fixed prices, written before the session starts, and nothing will blur them after the fact.
That makes it testable without any special effort. You can count, across a hundred candles, how often a bound was exceeded and then reclaimed within the following candle. The result is a percentage, and it is the same for everyone.
⚠️ Testable does not mean profitable. A pattern that completes six times out of ten is still a loser if the reaction does not cover the stop it forces on you. Count both, not just the frequency.
SMT divergence: when two correlated markets disagree
An SMT divergence is a disagreement between two correlated instruments at the same swing point: one takes out its previous high or low, the other fails to. SMT is short for smart money technique, the name ICT vocabulary gives to that comparison. Traders often pair it with an inverse fair value gap in what is known as the IFVG and SMT model: the divergence says the move lacks support, the inverted gap says where price has already been rejected.
It comes in two forms. The bearish one is drawn below: instrument A makes a higher high while B stays under its own. The bullish one is its mirror at the lows: A trades through its previous low, B holds above. The reasoning is the same both ways: if two correlated markets stop answering each other, one of them is wrong, and it is usually the one that went further.
It is the only one of the four that measures without ambiguity, and that is what sets it apart. The divergence is a fact observable at the moment it appears: two charts, two highs, one comparison. No interpretation is needed to establish that it exists.
The choice of the two instruments decides everything, and three pairings come up in almost every course. On index futures, ES against NQ, the S&P 500 and Nasdaq-100 contracts. On currencies, EUR/USD against GBP/USD, two pairs quoted against the same dollar. And either pair against the dollar index, DXY, which moves the opposite way, so that a high on one chart is compared with a low on the other. The rule behind the list is general: two pairs sharing a currency, two indices from the same country, two expiries of the same contract. The stronger and steadier the correlation, the more the divergence means something.
Check that correlation before using it, and over the recent period rather than a long average: two markets correlated three years ago may not be any more, and the divergence then means nothing at all. The DXY pairing deserves a second look for another reason. The euro makes up 57.6% of the index, according to the weights published by ICE, which runs it, so EUR/USD and an inverted DXY are largely the same series, and a divergence between them mostly reports what the yen, the pound and the three other currencies of the basket did that hour.
A sweep that only happened on one side
An SMT divergence is a liquidity sweep that only happened on one of the two charts. The first instrument takes out its previous high, the second stops short of its own. Same clock, same reference level, two opposite outcomes.
Three conditions have to be written down before you look, otherwise you will find a divergence every time you go looking for one. The high you compare has to be the same event on both sides, identified by its timestamp and not by its shape. Both charts have to show the same timeframe and the same session hours. The side you keep is the one that failed to take its level, never the one that went past it.
The fourth condition is almost always missing, and it is invalidation. The reading dies the second the laggard takes its own level in turn. Write that price down before you act: it is what will tell you that you were wrong, not your discomfort.
The costliest mistake is comparing two charts whose session does not start at the same hour. A futures contract trades close to twenty-three hours out of twenty-four, a cash index only during the opening hours of its exchange. The high on one of them may have formed while the other was closed, and the divergence you logged then exists only in your software.
What that costs you is not the trade, it is the record. Cases that never happened enter your count, and the rate you draw from it no longer measures anything. A clean divergence still tells you neither when to enter nor where: it says a move lacks support, the rest comes from elsewhere.
An SMT divergence example on ES and NQ, and its limits
Take the two contracts on a 5-minute chart with the same session template. Overnight, ES left a low at 5,480.00 and NQ a low at 19,640.00. The numbers are there to show the arithmetic, not to be looked up on a chart.
At 9:35 New York time ES trades down to 5,476.50: its overnight low is taken by 3.5 points, fourteen ticks. In the same five minutes NQ bottoms at 19,652.25 and never reaches 19,640.00. That is a bullish SMT divergence: same timestamp, same reference, one low swept and the other not.
The instrument kept is NQ, the one that held, and the price that cancels the reading is known before anything else: 19,640.00, the low NQ refused to take. Buying NQ at 19,660.00 puts that level 20 points away, so 80 ticks at 5 dollars, so 400 dollars a contract before any allowance for noise.
The first limit is frequency. The lower the timeframe, the more swing points there are, and each is a chance for two contracts to disagree by a few ticks. We know of no published study that measures how often such a divergence is followed by a reversal rather than by the laggard catching up, so that rate is yours to count.
The second is that the two baskets are not the same. The Nasdaq-100 is weighted towards large technology companies, the S&P 500 spreads across every sector. On a day when technology moves on news of its own, NQ and ES part ways for a reason that has nothing to do with stops, and they can stay apart all session. The pound does the same on a Bank of England day: EUR/USD and GBP/USD diverge because one of the two currencies has news.
OTE: where to enter, never whether
OTE in trading is short for optimal trade entry, a retracement zone between two Fibonacci levels, usually 0.62 and 0.79 of the previous move.
The Fibonacci retracement is the tool that produces those levels. You measure a move from its origin to its extreme, and the grid derives the 0.38, 0.50, 0.62 and 0.79 steps of the ground covered. It is one of the most widely used indicators in technical analysis, and the only one ICT methods genuinely keep.
What the ICT approach does with it is specific, and that is the point worth remembering: it does not use the whole grid. The 0.38 and 0.50 Fibonacci levels are ignored, only the 0.62-0.79 band is kept, because that is where price has given back enough ground for an entry to leave a short stop. Here the Fibonacci retracement works as a placement rule, not as a buy or sell signal.
The 0.50 level still has a role, but it is not an entry level. In ICT vocabulary the middle of the range is called equilibrium: it cuts the move into two halves. Above it sits the premium zone, where price is expensive; below it, the discount zone, where price is cheap. The range in question is the dealing range, the move measured between a meaningful swing low and swing high.
The direct consequence: the OTE of a long, between 0.62 and 0.79 of retracement, always falls below equilibrium, so in the discount zone. You buy in the lower half of the move, never in the upper half. For a short it is the mirror image: the move is measured from the high to the low, and the OTE falls in the premium zone. It is the mechanical version of an old principle, buy low and sell high, and it is enough to rule out an entry taken in the wrong part of the range.
One caveat matters: these levels are not magic, and no property of the golden ratio explains a market. What gives them any reach is that many participants watch the same steps at the same time, and place their orders there. The grid describes a shared convention, it does not describe a law.
In practice the Fibonacci retracement tool is drawn straight onto the chart: you connect the start of the move to its extreme and the steps appear on their own. It is one of the drawing tools shipped with the Tradoshi chart, free plan included.
It is therefore an entry zone, not a signal: it tells you where to enter if the scenario holds, never whether it holds. The two get confused constantly, and that is precisely what makes this reading so easy to teach and so disappointing to apply.
What it genuinely gives you is a coherent stop placement. If you enter inside the 0.62-0.79 Fibonacci band, the level that invalidates your scenario sits just beyond, and your risk is bounded by geometry rather than by your comfort.
That is already a great deal, and it is not what you are sold. A well-defined entry zone improves your reward-to-risk at a constant hit rate; it does not improve the hit rate.
The scenario itself has to come from elsewhere: from readable market structure, an identified sweep, a session sequence. The zone says where, something else has to say whether.
The only one you can check on the spot
Go back to the question set at the start: at what moment can you know you were wrong? A BPR can be spotted on the spot but shows up too rarely to be measured. CRT can be spotted on the spot and measures easily.
OTE says nothing about the scenario, so it can be neither right nor wrong on its own: it inherits the validity of whatever came before it.
SMT is the only one whose signal can be established at the moment it appears, without waiting for what follows, and on a fact nobody can interpret differently. Two highs, one taken out, the other not.
That does not make it the most profitable, it makes it the most honest to test. A signal verifiable on the spot can be counted across a history without your judgment getting involved, and that is the condition for the count to mean anything.
⚠️ Knowing these words does not make you money, and that has to be said plainly. They describe traces after the fact, with a precision that creates the illusion of prediction. If you cannot say in advance what would make the reading wrong, you do not have a method, you have a story.
Practising: thirty cases of each
The protocol is the same for all four, and that is what makes them comparable. Pick one instrument and one timeframe, and do not change them again.
Collect thirty CRT cases and thirty SMT cases across a history, moving candle by candle. For each one, note what you would have done live, before looking at what followed.
Add BPR if you can find thirty of them, and use OTE only as an entry placement on the cases from the first two. That is its real use.
Then compare three numbers per acronym: how often it appears, how often it completes, and above all the ratio between the average win and the average loss. The third decides, the first two explain it.
Sort the whole thing by setup type. That is exactly what a trading journal does, and without that sorting the claim "I trade SMT" stays an intention rather than a measurement.
Frequently asked questions
What does SMT stand for in trading?
SMT stands for smart money technique. In ICT vocabulary it names one thing only, the divergence between two correlated instruments at the same swing point, which is why the acronym is almost always followed by the word divergence.
Is an SMT divergence the same as an RSI divergence?
No. An RSI divergence compares a price with an indicator computed from that same price, so both sides of the comparison come from one series. An SMT divergence compares two prices, two instruments quoted separately, with no indicator involved. The only thing they share is the word.
Which timeframe should you use for SMT divergence?
The comparison is valid on any timeframe, provided both charts use the same one and the same session hours. What changes is the reference: on an intraday chart it is usually the previous session's high or low, which both instruments have by construction. The lower you go, the more swing points there are, hence the more divergences, and the larger the sample you need before the count means anything.
Does SMT divergence work on crypto?
The natural pairing there is Bitcoin against Ether, the two largest cryptocurrencies, and the same condition applies: check the recent correlation before trusting the comparison. Crypto removes one trap, since both markets trade around the clock and no session opens later than the other. It adds another: the same coin prints slightly different highs and lows from one exchange to the next, so take both charts from the same exchange.
Key takeaways
- All four acronyms describe the same thing from different angles. What separates them is when you can know you were wrong.
- BPR is the overlap of two opposing imbalances. Unambiguous to spot, but too rare to be measured seriously.
- CRT takes the high and low of a higher-timeframe candle as bounds. Its levels are known in advance, so it tests without effort.
- SMT compares two correlated instruments. It is the only one whose signal is established on the spot, on a fact nobody can reinterpret.
- OTE says where to enter, never whether. It improves your reward-to-risk, not your hit rate.
- Check the correlation before using SMT, and over the recent period: two markets correlated three years ago may not be any more.
Going further
These blog articles dig into this lesson's ideas, one subject per article.