SMT divergence: how to read it between NQ and ES
SMT divergence appears when two correlated assets stop confirming each other at an extreme. Here is how to read it on futures and use it as confirmation, not as a signal.
SMT divergence is a non-confirmation between two correlated instruments. NQ prints a new high and ES does not. Or gold sweeps its previous low and silver does not. One takes out its extreme and the other stays behind. It is often described as a "crack in correlation".
SMT stands for "smart money technique", a term from the ICT vocabulary. For those who use it, that crack suggests the new extreme may be false. That is a thesis of the method, not a proven fact: there is no public statistic on SMT, neither a win rate nor a comparison with a control group.
That is why this guide treats it for what it is in practice: a confirmation layer inside a broader model. You will see which pairs to use in futures, how to read it step by step, the systems that have written rules, the usual mistakes and how it fits in a prop firm account.
Free tool: if you use NinjaTrader 8, you can download the SMT divergence indicator we built at El Trader Financiado for the community. See the indicator and download it.
What is SMT divergence
In the most widespread version of the method, the definition is simple:
- Bearish SMT: in a zone of highs, one asset takes out its previous high and the correlated one does not.
- Bullish SMT: in a zone of lows, one asset loses its previous low and the correlated one does not.
The swings being compared have to be coincident: around the same times and within the same session. Comparing Monday's NQ high with Tuesday's ES high is not an SMT.
The logic the method defends is this: if two indices rise and fall together almost always, and at an extreme only one of them takes it out, the strength of the move is not broad. The asset making the new extreme would be sweeping liquidity, not starting a trend. The first part (that the indices are highly correlated and sometimes split apart) is observable. The second (that this split anticipates a reversal) has not been measured.
Which pairs to use in futures
| Market | Usual pairs | Note |
|---|---|---|
| US indices | NQ, ES and YM | NQ/ES is the most used pair; YM acts as a third |
| Metals | GC and SI (gold and silver) | Same logic as in indices |
| Currencies | EUR/USD and GBP/USD | With the DXY the logic is inverted |
| Crude oil (CL) | None standard | No accepted SMT pair showed up |
With an inverse pair, such as EUR/USD against the DXY, the divergence is read the other way around: a new high in one should go with a new low in the other.
In futures there is a practical detail: compare micros with micros or full-size contracts with full-size contracts, and always the same expiration. We cover it in the mistakes.
How to read an SMT step by step
- Mark the level on the higher timeframe. SMT makes sense at a place the market was going to seek: previous day high or low, Asia or London high or low, or a session extreme. The levels and their times are in ICT killzones.
- Open both charts on the same timeframe. The most used are 15, 5 or 1 minute. Same timeframe on both.
- Wait for price to reach the level. Watch what each asset does over the same stretch of time.
- Compare the swings with tolerance. The two assets often drift apart by a few minutes. If you demand the extreme on the exact same candle, you will miss valid divergences and see others that are not.
- Identify which one sweeps and which one does not. The one that takes out the extreme is the one that "lies", according to the method. If NQ sweeps the high and ES does not, it is a bearish SMT.
- Look for confirmation. SMT triggers the search for displacement and a structure break on the lower timeframe. Without that confirmation, there is no entry.
- Discard it if the correlation has broken. If the two assets have been moving separately for hours (for example, because of news that only affects one sector), SMT loses its basis. Do not trade.
SMT systems with rules
SMT is almost never presented as a complete system. These are the models that have written rules in their sources:
| System | Entry | Stop | Target | Time filter | Management |
|---|---|---|---|---|---|
| Public script: SMT + IFVG in killzones | Only NQ or only ES sweeps the saved session extreme ("exclusive sweep"); confirmed when price closes beyond the edge of the IFVG | Not defined by the script | Not defined by the script | Session extremes within killzones | Alerts on bar close |
| Public script: multi-level SMT | Divergence if at least one asset takes the level but not all (chart versus two "sister" symbols) | Not defined | Not defined | Intraday cycles anchored at 20:00 NY and timeframes from 1h to monthly | Rejects the divergence if the laggard had already cleaned the level before |
| IFVG model with SMT | Clear sweep with SMT, clean FVG, close through it that turns it into an IFVG and entry on the return | Beyond the swept extreme (common ICT skeleton) | TP1 at the internal liquidity, final TP at the major swing | New York session | Two staggered targets |
The first two are public TradingView scripts; the third is the most widespread version of the IFVG model.
Note the stop column: SMT scripts do not set one. The stop and target are set by the model the SMT accompanies. The author of the multi-level script himself presents it as a context reading, not as a complete system. And the people who spread the IFVG model talk about "high probability" without providing data.
The rejection rule of the multi-level script is worth copying even if you do not use the script. If the lagging asset had already taken that level a while earlier, there is no real divergence: both swept it, just at different times.
Common SMT divergence mistakes
- Demanding the extreme on the same candle. Assets drift apart by minutes. You need tolerance between the two swings.
- Using SMT alone. Correlated instruments briefly desync all the time. A loose SMT may only indicate a short pullback.
- Trusting a lower-timeframe SMT without higher-timeframe backing. It is more likely to be an inducement, a trap.
- Comparing assets that are not correlated. They produce meaningless divergences. That is why CL has no standard pair.
- Using tiny swings on 1 minute. Normal desync generates constant "SMTs". The smaller the swing, the more noise.
- Comparing different expirations during the roll. If you compare December NQ with March ES, or one adjusted continuous contract with an unadjusted one, the extremes shift and you see divergences that do not exist. No source documents this mistake; it is deduced from how continuous contracts are built.
What to combine SMT with
SMT fits inside the sequence that almost all ICT models repeat: liquidity level, sweep in a time window, change of structure and entry.
- Level and window: the sweep of a session extreme inside a killzone. We explain it in ICT killzones and the session levels indicator draws it.
- Change of structure: the MSS after the SMT. It is in order block and market structure.
- Entry: an FVG or an IFVG of the leg that breaks. You have it in fair value gap.
- CRT: a 4h candle that sweeps its range on NQ but not on ES is the cross-asset version of the same argument. We cover it in CRT trading.
VWAP barely appears in the SMT material. If you use it as a location filter, that is your decision, not part of the method.
What the evidence says about SMT
It is short to tell: there is no public statistic on SMT divergence. No win rate, no comparison with random divergences, no backtest with costs. It is the piece with the least evidence in the whole ICT repertoire.
What can be said is that divergence between correlated indices exists and can be measured. What is not proven is that it anticipates reversals.
In its favor, it is also the piece that is least sold as a system. Its most defensible use is as a filter that reduces trades: you only enter when, in addition to your model, there is an SMT at a higher-timeframe level.
How SMT fits a prop firm
SMT adds no risk of its own, because it does not set the stop. The risk comes from the model it accompanies. Even so, there are four points worth having clear.
The stop, in dollars per contract. If you enter after an SMT with the stop beyond the swept extreme, measure that distance before entering. An example:
| Contract | Point value | 25-point stop | 10-point stop |
|---|---|---|---|
| NQ | $20 | $500 | $200 |
| MNQ | $2 | $50 | $20 |
| ES | $50 | $1,250 | $500 |
| MES | $5 | $125 | $50 |
With a daily loss limit of, say, $1,000, two 25-point stops on NQ eat the day. On micros, a tenth of that.
News. New York morning SMTs coincide with the 08:30 and 10:00 data releases. News can move one index more than another and create a divergence that has nothing to do with liquidity. Many prop firms restrict trading around news such as CPI or NFP; check the rules of yours.
Data. You need the feed of both assets at once. If your platform loads only one, you cannot read the SMT.
Daily loss and overtrading. The real risk of SMT in a funded account is using it as permission to enter every time NQ and ES split for a minute. That multiplies trades and brings the daily limit closer. Use it to filter, not to add entries.
If you want to see which firms have the roomiest daily loss limit, use the prop firm comparator.
Free NinjaTrader SMT indicator
We built an SMT divergence indicator for NinjaTrader 8 that automatically picks the pair with the same expiration (NQ with ES, gold with silver, micros with micros) and marks each divergence on your chart, with tolerance between the two swings. It is free, built by El Trader Financiado for the community and available in English and Spanish. Download it and see its settings on the SMT indicator for NinjaTrader page.
SMT divergence FAQ
What is SMT divergence?
It is a non-confirmation between two correlated assets: one takes out its previous high or low and the other does not. In the ICT method it is read as a sign that the extreme may be a liquidity sweep.
Which pairs should I use for SMT in futures?
NQ with ES (and YM as a third) in indices, and gold with silver in metals. For crude oil there is no standard pair.
On which timeframe do you look for SMT?
On the same timeframe on both charts, usually 15, 5 or 1 minute, and at a higher-timeframe level. A lower-timeframe SMT without higher-timeframe backing is more likely to be a trap.
Can SMT be used on its own?
It is not recommended. Correlated assets often desync, and there is no public statistic showing that SMT works as an isolated signal. Use it as confirmation of a sweep.
What do I do during contract rollover?
Always compare the same expiration on both assets. Mixing the expiring contract with the next one, or an adjusted continuous contract with an unadjusted one, shifts the extremes.
