SMT Divergence Strategy

SMT (Smart Money Technique) divergence is a multi-instrument price-action concept, popularised in the smart-money / ICT community, where two normally-correlated markets stop confirming each other at a swing: one makes a higher high (or lower low) while its correlated partner fails to. Traders read this non-confirmation as a sign that one of the moves may be a liquidity grab rather than a true continuation. This page explains the concept neutrally — how traders define swings, divergence, and how they tend to frame entries, exits, and risk — and provides a VALID Pine v6 illustrative detection/marker script. It does not reduce to a clean mechanical strategy, makes no profit or win-rate claims, and is for education only. Nothing here is predictive; trading involves the risk of loss.

How it works

SMT divergence ("Smart Money Technique" divergence) is built on the idea of inter-market correlation. Two instruments that usually move together — common examples discussed in the community include index futures such as ES / NQ / YM, currency pairs that share a quote currency (e.g. EUR/USD vs GBP/USD), or an asset and its inverse (e.g. EUR/USD vs DXY, where they tend to move opposite) — are expected to confirm each other's swing structure.

The concept says: at a meaningful swing point, compare the two correlated charts. - Bearish SMT divergence: Instrument A prints a HIGHER high than its prior swing high, but the correlated Instrument B prints a LOWER high (or fails to make a new high). The new high in A is unconfirmed. - Bullish SMT divergence: Instrument A prints a LOWER low than its prior swing low, but Instrument B prints a HIGHER low (or fails to make a new low). The new low in A is unconfirmed.

For inversely-correlated pairs (e.g. EUR/USD vs DXY) traders flip the comparison: a new high in one should coincide with a new low in the other; when that mirror breaks, it is read as divergence.

The narrative many traders attach to this is "liquidity": a new high/low that isn't confirmed by the partner market is interpreted as a possible stop-run or liquidity sweep above/below a prior swing, rather than a genuine break of structure. Importantly, this is an interpretation, not a guarantee — non-confirmation is a context cue, not a prediction. Correlations also drift and break down, so the comparison only carries meaning while the two instruments are actually correlated over the timeframe being traded.

Because SMT divergence inherently requires comparing TWO data feeds at swing points and then layering discretionary confirmation on top (structure shift, session timing, higher-timeframe bias), it does not reduce cleanly to a single mechanical entry/exit system. The Pine script below is therefore an ILLUSTRATIVE DETECTION / MARKER tool: it flags candidate SMT-divergence swings on the chart so you can study them. It is not a trade-signal generator and includes no profit logic.

Entry rules

This is how traders typically FRAME entries — not a recommendation or a claim that any of it is profitable. Confirm the two instruments are genuinely correlated (or inversely correlated) on your timeframe before reading anything into a divergence.
Identify a fresh swing on the traded instrument: a new swing high above the prior swing high (for shorts) or a new swing low below the prior swing low (for longs), using a consistent pivot/swing definition.
Check the correlated partner at the SAME swing: bearish SMT = partner makes a lower high (no new high) while your instrument makes a higher high; bullish SMT = partner makes a higher low (no new low) while your instrument makes a lower low. For inverse pairs, mirror the comparison.
Most discretionary traders do NOT enter on the divergence alone. They wait for a confirmation trigger — e.g. a shift in market structure / break of a short-term swing in the divergence's direction, often on a lower timeframe — before considering an entry.
Context filters traders commonly add: alignment with higher-timeframe bias, a specific session/time window, or proximity to a pre-identified level. These are discretionary and vary by trader.
Treat the divergence as confirmed only on a CLOSED bar / completed swing. Acting on an in-progress swing risks the 'divergence' disappearing as the bar develops.

Exit rules

Again, descriptive only — how traders tend to manage the idea, with no implication of profitability. Define a stop BEFORE entering, since the premise can simply be wrong.
Invalidation level: many traders place the protective stop just beyond the swing extreme that formed the divergence (above the unconfirmed high for a short, below the unconfirmed low for a long). If price reclaims that extreme, the 'liquidity grab' read has failed.
Targets are usually defined by structure, not by the divergence itself — e.g. an opposing swing point, a prior level, or a session high/low. Some traders use a fixed risk-to-reward framing; this does not make any outcome more likely.
Time-based exit: if the expected structure shift does not develop within the trader's chosen window, some flatten regardless, treating the setup as failed.
Risk sizing is independent of the signal: position size per trade is typically capped to a small fixed percentage of account equity so that any single failed setup is survivable. Risk management does not turn a non-predictive cue into a positive expectancy.

Risk notes

Educational content only — nothing here is financial advice or a recommendation to trade.
No method described claims or implies profit, positive returns, or any win-rate. SMT divergence is non-predictive.
Trading carries a real risk of loss, including loss exceeding deposits on leveraged products. Only risk capital you can afford to lose.
Any backtest or marker output is illustrative; past behaviour does not indicate future results.
References to the smart-money / ICT community are neutral context; trademarks belong to their owners and no endorsement or affiliation is implied.

Pine v6 example

//@version=6
// SMT Divergence — ILLUSTRATIVE DETECTION / MARKER (educational only)
// SMT divergence does NOT reduce to a clean mechanical strategy: it needs a
// second correlated feed + discretionary confirmation. This script only MARKS
// candidate divergence swings between the chart symbol and a correlated symbol.
// It is NOT a trade-signal generator and contains NO profit logic.
// Nothing here is predictive. Trading involves risk of loss; past != future.
indicator("SMT Divergence (Illustrative Marker)", overlay = true)

// --- Inputs ---
corrSym = input.symbol("CME_MINI:NQ1!", "Correlated symbol")
inverse = input.bool(false, "Correlated symbol is INVERSELY correlated")
pivLen  = input.int(5, "Pivot lookback (left = right bars)", minval = 1)

// --- Pull the correlated symbol's OHLC on the chart timeframe (no look-ahead) ---
// barmerge.lookahead_off + only reading historical/confirmed offsets avoids
// look-ahead bias and repainting.
[cHigh, cLow] = request.security(corrSym, timeframe.period, [high, low], lookahead = barmerge.lookahead_off)

// --- Swing detection on the chart symbol (confirmed pivots only) ---
ph = ta.pivothigh(high, pivLen, pivLen)
pl = ta.pivotlow(low,  pivLen, pivLen)

// A non-na pivot value refers to the bar pivLen bars back. Read both feeds at
// that same offset so the chart symbol and the partner are compared on one bar.
selfHighAtPivot = high[pivLen]
selfLowAtPivot  = low[pivLen]
corrHighAtPivot = cHigh[pivLen]
corrLowAtPivot  = cLow[pivLen]

// Track the previous confirmed pivot of each kind to compare "new" vs "prior".
var float prevSelfHigh = na
var float prevCorrHigh = na
var float prevSelfLow  = na
var float prevCorrLow  = na

bearSMT = false
bullSMT = false

if not na(ph)
    if not na(prevSelfHigh) and not na(prevCorrHigh)
        selfHigherHigh = selfHighAtPivot > prevSelfHigh
        // Direct correlation: partner should ALSO make a higher high to confirm.
        // Inverse correlation: partner mirrors down, so it should make a LOWER
        // high to confirm. Divergence = the partner does NOT confirm.
        corrConfirms = inverse ? (corrHighAtPivot < prevCorrHigh) : (corrHighAtPivot > prevCorrHigh)
        bearSMT := selfHigherHigh and not corrConfirms
    prevSelfHigh := selfHighAtPivot
    prevCorrHigh := corrHighAtPivot

if not na(pl)
    if not na(prevSelfLow) and not na(prevCorrLow)
        selfLowerLow = selfLowAtPivot < prevSelfLow
        // Direct: partner should also make a lower low to confirm.
        // Inverse: partner mirrors up, so it should make a HIGHER low to confirm.
        corrConfirms = inverse ? (corrLowAtPivot > prevCorrLow) : (corrLowAtPivot < prevCorrLow)
        bullSMT := selfLowerLow and not corrConfirms
    prevSelfLow := selfLowAtPivot
    prevCorrLow := corrLowAtPivot

// --- Markers (offset back to the pivot bar so the marker sits on the swing) ---
plotshape(bearSMT, title = "Bearish SMT", style = shape.triangledown, location = location.abovebar, color = color.red, size = size.small, offset = -pivLen, text = "SMT")
plotshape(bullSMT, title = "Bullish SMT", style = shape.triangleup, location = location.belowbar, color = color.teal, size = size.small, offset = -pivLen, text = "SMT")

// Educational marker only. Any backtest layered on top would be ILLUSTRATIVE;
// results are not predictive and past behaviour does not indicate future results.

Pitfalls

Correlation is assumed, not guaranteed. SMT only carries meaning while the two instruments are actually correlated on your timeframe; correlations drift and break, producing meaningless 'divergences'. Always verify the relationship is live before reading anything into it.
Look-ahead / repainting risk. Pulling a second symbol with request.security can leak future data if lookahead is misused. Use barmerge.lookahead_off and only act on CLOSED bars / completed pivots — an in-progress swing can change. ForexCodes' Validator is designed to catch exactly this kind of look-ahead bias and repainting.
Pivot lag. ta.pivothigh / ta.pivotlow only confirm a swing after 'right-bars' have closed, so detection is inherently delayed — the marker appears bars after the actual high/low. This is honest, not a bug, but it means you cannot act at the exact extreme.
Partner sampled at the chart's swing bars only. This illustrative script reads the partner's value at the chart symbol's pivots rather than detecting the partner's own independent swings; that is a deliberate simplification, and a stricter implementation would track each feed's swings separately.
Timeframe and timestamp mismatch. The two feeds must be aligned on the same timeframe and session; comparing markets with different trading hours or holidays can create false non-confirmations.
It is not mechanical. SMT divergence on its own is a context cue, not an entry. Treating the marker as a buy/sell signal ignores the discretionary confirmation (structure shift, bias, timing) that practitioners layer on top.
Confirmation bias. Because you can pick from many correlated partners, it is easy to find SOME chart that 'diverges' after the fact. Decide the partner and rules in advance to avoid cherry-picking.
No edge is implied. Detecting a pattern says nothing about outcome. Nothing here predicts price, and any setup can fail; trading risks loss of capital.

FAQ

Does SMT divergence predict reversals?

No. It is a non-confirmation cue, not a prediction. When one instrument makes a new swing extreme that its correlated partner does not confirm, some traders interpret it as a possible liquidity grab — but it can equally precede continuation. Nothing about the pattern guarantees a reversal or any particular outcome, and trading it involves the risk of loss.

Which instruments are used for SMT divergence?

Any two markets that are normally correlated (or inversely correlated) on the timeframe you trade. Examples commonly discussed in the smart-money / ICT community include correlated index futures, currency pairs sharing a common currency, or an asset versus an inverse instrument like a dollar index. The key requirement is that the correlation is genuinely present — verify it before drawing conclusions, since correlations break down.

Why is your Pine script a marker instead of a full strategy?

SMT divergence inherently needs a second correlated data feed AND discretionary confirmation (structure shift, higher-timeframe bias, timing), so it doesn't reduce to a clean mechanical entry/exit system. We provide a valid Pine v6 illustrative detection script that flags candidate divergence swings for study. It uses barmerge.lookahead_off and confirmed pivots to avoid look-ahead bias, and it contains no profit logic. ForexCodes' Validator exists precisely to check that scripts like this don't repaint or peek at future data.

Want this as a clean, validated Pine script? run it through the Validator → or convert a trading strategy to Pine Script →

Educational & software only — not financial advice, not a recommendation to trade. Backtests are illustrative; past performance does not predict future results. Trading involves substantial risk of loss.