Updated
Genjutsu Kai - cracks in correlation (ICT's SMT)

Genjutsu Kai (幻術解 · "Genjutsu Release")
A run on one index can be an illusion: a high or low taken that its sister index refuses to confirm. Genjutsu Kai maps those cracks in correlation (ICT's SMT) between your chart and its sister market (NQ ↔ ES by default) during the regular session. It measures how far the trailing market, the "sick sister", is behind, and tracks whether each divergence holds or syncs.
A confluence map, not a signal.
What it draws
- Trendline from the reference high or low to the leader's new extreme
- Ghost: the sister's move, scaled by % onto your chart
- Lag bracket: the gap between them, in points and % ($ per contract in the tooltip)
- Tag: ▼/▲, the reference (30, 1H, OR) and its state: ● open · ◐ synced · ■ held
- Swing dots at both ends of the trendline; hover any mark for the full read
- Opening range lines (9:30–10:00 ET)
- Compact card: who leads since 9:30, each reference's high/low state, today's tally
How it works
- References: 30-minute ranges, hourly ranges anchored to 9:30, and the opening range checked 10:00–16:00
- SMTs are only checked in RTH, 9:30–16:00 ET
- A divergence becomes an SMT once the sister has failed to confirm for longer than the grace period (default 2 min) and the lag reaches the minimum (default 0.02%)
- Each SMT then syncs (the sister takes its level: continuation on both) or holds (the divergence stands to the end of its window)
- Confirmed bars only, so nothing repaints
- Pairs itself on continuous (NQ1! ↔ ES1!) and single-contract (NQZ2026 ↔ ESZ2026) charts, and MNQ ↔ MES; any symbol can be set by hand
HYDRA heads (NQ/ES 1-minute data: fit 2018–24, locked test 2025–26)
- Sync odds ↻: the chance the sister still takes its level before the window ends, from its distance, its volatility and the time left. Out of sample it separated syncs from holds with AUC 0.84 and read 25→21%, 45→40%, 65→67%, 93→93% (predicted→actual)
- Roll week flag: in the week before quarterly expiry week, fading an SMT won the 1σ race* 54% of the time in 2018–24 and 58% in 2025–26 (30 days), vs 49% in other weeks. It was the only one of 635 candidate features that survived the locked test. Treat it as a candidate
What the testing says
- On its own an SMT was a coin flip: fading one won the 1σ race 49.5% (2018–24) and 49.9% (2025–26), close to fading an ordinary sweep or simply being long
- Whether an SMT held or synced said nothing about direction afterwards
- Sync odds forecast the sister's reach, not where price goes next
*1σ race: did price move one 30-minute standard deviation in the fade direction before moving one against, by 16:14 ET
Settings
- Each reference on/off with its own colours and width (default: 30-minute + opening range)
- Grace period, minimum lag, synced SMTs keep/fade/remove, days kept
- Ghost style, lag bracket, swing dot size, tag content, size and backing, lag units
- Card position, size and theme; sync odds in card/tooltip/tag; roll week flag
- Alerts: new SMT, sister syncs, SMT held
Use
- Chart timeframes of 1–30 minutes that divide 30; 1-minute recommended (the odds were fitted on 1-minute data)
- NQ, ES, MNQ and MES pair automatically; set "Compare with" for anything else (e.g. RTY1!)
Credits
- SMT, sick sister and opening range concepts as taught by ICT (Michael J. Huddleston)
- Trendline presentation inspired by "SMT Divergences" by LuxAlgo (CC BY-NC-SA 4.0); written from scratch, no LuxAlgo code used
- Not affiliated with or endorsed by either
Open source under MPL 2.0. For education and research, not financial advice. Past results do not guarantee future results.
Release Notes
• Settings: a View section at the top with Theme and Card text, the same as the other indicators. The card's on/off and position stay in Card.Release Notes
v4: ETH hours, session opening ranges, YM and RTY pairs, and sensitivity tested per pair and session█ HOURS
• New Hours input under References:
– RTH 9:30–16:00, the default and as before;
– ETH 18:00–17:00, the whole CME day.
• In ETH every window is tagged by its session: Asia · Asia→London · London · London→NY · NY · post-close. The card header shows the session, and the alerts name it.
█ SESSION OPENING RANGES (ETH)
The opening-range track runs once per session, each opening on that session's hot block (the widest 30 minutes of the session, measured on NQ, ES, YM and RTY 2018–26):
• Asia: the Tokyo open, 09:00–09:30 Tokyo, checked to 00:00 ET. It was the widest Asia block on a third of days, at 1.31× the session's average 30 minutes.
• London: the London open, 08:00–08:30 London, checked to 10:00 London.
– That is 03:00 ET most of the year, and 04:00 ET in the weeks the US and UK clocks differ.
– It ran 1.24× the session's average; the 02:00 ET block ran 0.9×.
• New York: 9:30–10:00, checked to 16:00.
• The card row reads OR·AS / OR·LN / OR·NY, and the opening-range lines carry their name.
█ PAIRS
• Auto pairs now include YM → ES and RTY → ES (MYM / M2K → MES), alongside NQ ↔ ES. Compare with still overrides.
• NQ–YM diverges most. In NY hours it gave 9.8 30-minute SMTs a day (NQ–ES 6.3), and 64% of them held (NQ–ES 54%).
█ SENSITIVITY, TESTED PER PAIR AND SESSION
Tested on NQ–ES, NQ–YM, ES–YM and RTY–ES, 1-minute 2018–26.
• Grace stays at 2 minutes. The sister's chance of following is highest in the first minute after the leader and halves by the second. The shape is the same for every pair and session.
• Minimum lag: new Auto (default). The cut is k × the pair's spread noise, with k set per session.
– It cuts about 1 SMT in 9 as hairlines everywhere (8–13% on 2025–26).
– The old fixed 0.02% is still available. It cut 11% of NQ–ES SMTs in NY hours over 2018–26, but 21% in 2025–26 and 40–45% in Asia hours.
• Sync odds: refit on all four pairs and every session, with one term per session. Pair terms added nothing.
– Locked test 2025–26: AUC 0.85 in RTH and in ETH.
– Predicted → actual in ETH: 3→3%, 20→18%, 40→39%, 59→63%, 77→82%, 96→93%.
– The v3 fit read 2–3 points high in NY hours, 6–9 in Asia and 17–21 in the post-close hour.
█ WHAT DID NOT CHANGE
• What follows an SMT is still a coin flip, in ETH as in RTH.
• Fading the leader's take won 47–51% of 1σ races in every pair and session: 88 tests, none beyond chance against matched controls.
• An SMT describes the moment; it does not call the next move.
█ FIXES AND SETTINGS
• Every window now resolves on its own last bar. In ETH the 16:30–17:00 window has no next bar until 18:00.
• In ETH the chart keeps at most 3 days of drawings (TradingView's 500-line cap).
• Minimum lag is now a choice: Auto (default) or Fixed %. Pick Fixed % at 0.02 to keep v3's minimum lag.
• The leader row now runs from the session's open: 18:00, the London open, or 9:30.
Historical frequencies, not guarantees. Not financial advice.
Disclaimer
The information and publications are not meant to be, and do not constitute, financial, investment, trading, or other types of advice or recommendations supplied or endorsed by TradingView. Read more in the Terms of Use.
Disclaimer
The information and publications are not meant to be, and do not constitute, financial, investment, trading, or other types of advice or recommendations supplied or endorsed by TradingView. Read more in the Terms of Use.