What Is SMT Divergence? (ICT)
SMT divergence is the ICT name for two closely related markets disagreeing at a turning point: one takes out its prior high or low while the other fails to. Traders read the failure as a sign the move is weak, and the SPY and QQQ record since 1999 is the way to check that reading.
SMT divergence is two closely related markets disagreeing at a high or a low: one pushes past its previous extreme and the other does not. In ICT teaching the failure is read as weakness, so a new high in one index that the other refuses to match is treated as a warning that the high may not hold.
The letters are usually expanded as “smart money technique”. The idea underneath is older than the label. Dow theory asked the industrials and the transports to confirm each other, and this is the same test run on a shorter clock.
How it works
Pick two markets that normally move together. The usual pairs are the S&P 500 and Nasdaq 100 futures, two major currency pairs, or a currency pair against the dollar index read upside down. The pair has to be tightly linked, because the whole reading depends on the two agreeing most of the time.
Mark the same swing on both. Take the most recent swing high in each market, formed over the same stretch of time.
Watch what happens when price comes back. If both take out their prior high, the move is confirmed. If one takes out its high and the other stalls below its own, that disagreement is the divergence.
Bearish and bullish are mirror images. At highs, one market making a higher high while the other makes a lower high is bearish SMT. At lows, one making a lower low while the other holds a higher low is bullish SMT.
The market that made the new extreme is often described as having run the stops above that level, which is why the idea sits next to the liquidity sweep in ICT material.
Why correlation is the whole idea
Divergence only means something against a background of agreement. Two unrelated stocks disagree every day, and nobody calls that a signal.
SPY and QQQ are close to the ideal pair. Their daily returns carried a correlation of 0.85 from March 1999 to September 2026, and they closed in the same direction on 84.1% of days. By calendar year the figure ran from 0.79 in 2017 to 0.97 in 2025, and it stands at 0.92 so far in 2026.
The flip side is that they differ in what they hold. The Nasdaq 100 leans far harder into technology than the S&P 500, so a week of strong chip earnings or a sharp move in rates can pull the two apart for reasons unrelated to anyone’s stops. A divergence has to be read with that in mind.
Which market fails matters less than people think
It is tempting to treat one index as the leader whose failure matters more. The record does not give much to choose between them.
Of the 314 bearish events counted below, SPY made the new high alone 150 times and QQQ 164 times. On the bullish side the split was 93 and 84. Neither index is a reliable leader, which fits two funds that share most of their largest holdings.
A worked example
14 August 2026, SPY against QQQ, daily bars. The prior 20-session high for each fund was set the day before, 13 August.
- QQQ traded up to $734.39, which is $0.43 above its prior 20-session high of $733.96. It took its high.
- SPY traded up to $778.80, which is $0.57 short of its prior 20-session high of $779.37. It did not.
That is a bearish SMT divergence under the rule this page measures. SPY closed the day at $776.34.
What followed. Five sessions later, on 21 August, SPY closed at $765.72, down 1.37% ($765.72 / $776.34 - 1). Ten sessions later, on 28 August, it closed at $769.35, down 0.90%.
Read it as one case, not a proof. The divergence did precede a small decline. The next section shows how often that happened across every case since 1999, and it is the only way to know whether this one was typical.
The original data
The rule, written so it can be rerun. Daily bars for SPY and QQQ on the 6,930 sessions both traded, 10 March 1999 to 25 September 2026. A fund “takes its high” when that day’s high is strictly above the highest high of the 20 sessions before it; lows mirror that. A bearish SMT day is one where exactly one fund takes its high. A day counts as a new event only if no day of the same kind came in the 5 sessions before it, so one cluster counts once. Every event is listed in the SMT event file.
Divergence is ordinary. On 1,923 days at least one fund made a new 20-session high, and on 960 of them (49.9%) only one did. On the 848 days with a new 20-session low, 377 (44.5%) were one-sided. After removing clusters that leaves 314 bearish and 177 bullish events, about 17.8 a year.
Bearish events against any day. SPY was lower 5 sessions later after 141 of 313 events with a full window, 45.0%, against 42.6% of all days (2,941 of 6,905). Ten sessions later it was lower after 123 of 313, 39.3%, against 40.2% of all days (2,772 of 6,900). A two-sided binomial test puts both results well within chance (p = 0.39 and p = 0.77).
When both funds confirmed the high, 229 events, SPY was lower 5 sessions later 39.7% of the time and 10 sessions later 37.6%. That is slightly less often than after a divergence, and the difference is small.
Bullish events leaned the wrong way. SPY was higher 5 sessions after 88 of 177 bullish events, 49.7%, against 57.3% of all days (p = 0.048). Ten sessions later the share was 57.6% against 59.8% (p = 0.59). The 5-session result is the only one on this page under the usual 5% line, only just under it, and it points against the bullish reading rather than for it. Across the four divergence tests reported here, one result that close to the line can easily come from chance, so it is weak evidence either way.
Search demand is thin but real. In the 24,971-video study behind this site, 8 titles contain “SMT”, from 6 channels, at a median of about 59,700 views. Only one of the 8 carries a description, and it does not discuss what a failed divergence looks like.
When it fails
The weaker market catches up
3 August 2026 is the counter-example. SPY traded up to $758.58, $3.00 above its prior 20-session high of $755.58 set on 15 July. QQQ reached only $701.59, far below its own prior high of $726.39 from 10 July. By the rule that is a large bearish divergence. SPY closed at $757.67 that day, then at $771.33 the next day, up 1.80%, and ten sessions later it was at $772.67, up 1.98%.
The lagging fund did not signal weakness. It caught up, and the divergence closed from below.
A daily rule is not an intraday rule
SMT is usually taught on intraday charts of index futures. The test on this page uses daily ETF bars, so it answers a different question. It does show that the daily version adds nothing measurable, and an intraday claim needs its own test on intraday data before it is trusted.
The swing you pick decides the answer
Change the lookback from 20 sessions and a different set of days becomes “divergent”. That freedom is the biggest risk in the idea: after the move, it is always possible to find some swing on some pair that disagreed. Fix the rule before looking at the chart, as the data section does, or the divergence will be found in hindsight every time.
Correlation breaks when it matters
Pairs decouple during sector rotations and rate shocks, precisely when a trader most wants the signal. In a year when the two funds move together less, disagreement becomes normal rather than informative.
Related
ICT is the framework this label belongs to, and that page separates its genuinely new ideas from the renamed ones.
A liquidity sweep is what the market making the new high is usually said to be doing, and its close-back-inside rule is a useful second trigger to pair with a divergence.
Correlation explains the number that makes this idea work at all, and why it weakens under stress.
For the same disagreement read between price and an oscillator instead of between two markets, see RSI divergence and the step-by-step guide to spotting a divergence.
I treat a divergence between two indexes as a reason to watch a level more closely, never as a signal on its own. Write down which market made the new high and which one failed, then wait for price to confirm the failure before you act on it.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.