OPEN-SOURCE SCRIPT
SAO RUBIQ Regime v2
# SAO · RUBIQ — Regime Visualizer (v2)
**by SNP420 · Jarvis Claudos · Finexus s.r.o.**
*Pine Script v6 · Build 2026-05-26*
---
## What is it
A research-grade market regime overlay that classifies every confirmed
bar into one of **five regimes** and paints the chart accordingly.
Born out of a simple insight:
> *"The right algorithm in the wrong phase of the market still fails —
> even when it is 100% correct under ideal conditions."*
This is the central thesis of the **SAO · RUBIQ** project: every
strategy lives or dies inside a specific regime × timeframe cell.
Without active regime perception, every trading system is necessarily
under-performing on the bars it was not designed for. This indicator
makes those cells visible.
## What's new in v2
v2 throws out the v1 AND-stack of ADX / BB-width / drift cutoffs and
replaces it with four orthogonal signals that vote together:
- **Fractal pivots** (3-bar centered) labeled **HH / LH / HL / LL**
- **Trend state machine** with **re-anchoring**:
UP / DOWN / RANGE / REVERSAL_UP / REVERSAL_DOWN
- **Multi-TF RSI** on M30 / H1 / H4 / D1 / W1 with per-TF thresholds
and a **≥3 of 5 agreement** rule
- **S/R level clustering** (±0.3 ATR tolerance) with **double-bounce**
reversal flag that persists for 10 bars
Measured improvements over v1.1 on EUR/USD M30, 2024-2025 (24,863 bars):
| Metric | v1.1 calibrated | v2 |
| ----------------------- | --------------- | ----------- |
| Confident coverage | 44.87% | **53.96%** |
| Transitions (less whipsaw) | 1,567 | **855** |
| BULL_CALM Sharpe (h=5) | +0.085 | **+0.271** |
| BEAR_CALM direction | +0.04 ⚠ (wrong)| **−0.19 ✅** |
| Shuffle F-stat (h=5) | 0.0027 | **0.0068** |
| Strategy diagonal score | 1 / 5 | **2 / 5** |
All validated by 1000-iter shuffle Monte Carlo (p < 0.001) and a
strategy × regime P&L matrix.
## The five regimes
| Regime | Meaning |
| ------------- | -------------------------------------------------------- |
| **BULL_CALM** | Trend UP + ≥3 of 5 RSI TFs agree bullish |
| **BEAR_CALM** | Trend DOWN + ≥3 of 5 RSI TFs agree bearish |
| **RANGE** | Mixed last pivots (no clean HH+HL or LH+LL) |
| **CHOPPY** | Trend UP/DOWN without RSI agreement, or active reversal |
| **STRESS** | Volatility spike: rv20_norm > 2 OR atr_z > 2 |
| *UNCERTAIN* | Fallback — pre-warmup or no labels yet (no box drawn) |
Priority order (highest wins): **STRESS > BULL > BEAR > CHOPPY > RANGE.**
## How it works
For every confirmed bar the indicator:
1. Computes ATR(14), realized vol (20), z-scored ATR — used by stress
detection and S/R cluster tolerance.
2. Requests RSI(14) on five timeframes via `request.security`
(M30 / H1 / H4 / D1 / W1) and counts bull / bear agreements.
3. Detects fractal pivots with `ta.pivothigh / pivotlow(3, 3)` and
labels each as HH / LH / HL / LL versus the previous same-type
pivot.
4. Adds each pivot to a greedy S/R level cluster (±0.3 ATR). On the
second touch within 500 bars it fires a **double-bounce reversal**
flag (type H → REVERSAL_DOWN, type L → REVERSAL_UP) that persists
for 10 bars.
5. Maintains a trend state machine: **UP** when last H=HH and last L=HL,
**DOWN** when last H=LH and last L=LL, **RANGE** otherwise.
6. Combines trend × RSI × stress into the final label, then applies
a hysteresis smoother (`min_run = 5 bars`) so single-bar flips
never get committed.
## Visualization (FX-Sessions-style)
- **Dashed segment box** per confirmed regime run, sized to that
segment's high/low.
- **Background tint** — semi-transparent regime color over the span.
- **Bar color** (off by default) — paints OHLC bars with regime color.
- **Segment label** anchored to the top of each box.
- **Pivot markers** — HH / LH / HL / LL drawn at every confirmed pivot
with bull/bear tint.
- **Reversal arrows** — ▲ green at support double-bounces, ▼ red at
resistance double-bounces.
- **Info table** (top-right) — current state, trend, last H and L labels,
RSI per TF, bull/bear agree counts, rv20_norm, atr_z.
- **Stats table** (bottom-right) — N bars and % share per regime across
the visible history.
UNCERTAIN bars deliberately render no box and no tint (clean chart).
## Settings worth knowing
- **Hysteresis min_run** (5) — bars of consistent raw state before
commit. Bigger = less flicker, more boundary lag.
- **Pivot left/right** (3) — fractal pivot window. Smaller catches
more pivots but more noise.
- **Per-TF RSI thresholds** — defaults are 65/35 (M30), 62/38 (H1),
60/40 (H4), 55/45 (D1), 50/50 (W1). Overridable per pair / TF.
- **RSI agree min** (3) — TFs that must agree to qualify as
BULL_CALM / BEAR_CALM. Lower = more sensitive.
- **S/R cluster tolerance** (0.3 × ATR) — width of an S/R level zone.
- **S/R max age** (500 bars) — oldest first-touch still eligible for
double-bounce reversal.
- **Reversal persistence** (10 bars) — how long after the second touch
the REVERSAL state stays active.
- **Max active S/R levels** (200) — FIFO ceiling on level memory.
- **All six regime colors + reversal arrow colors** — fully overridable.
## Sanity-check expectation (EUR/USD M30, 2 years)
If the v2 baseline distribution holds on your data window:
| State | Share |
| --------- | ------ |
| BULL_CALM | ~2.6% |
| BEAR_CALM | ~1.8% |
| RANGE | ~2.4% |
| CHOPPY | ~38.7% |
| STRESS | ~8.4% |
| UNCERTAIN | ~46.0% |
v2 deliberately has stricter BULL / BEAR (needs trend state + 3-of-5
RSI agreement) and a wider CHOPPY (catches trend bars without RSI
agreement + all active reversals). Cleaner regime blocks, better
direction mapping.
## Honest limitations — please read
- **Tuned to EUR/USD M30 vol scale.** Features and rules are
TF-agnostic in math, but the thresholds were calibrated on M30.
On H1 / D1 / crypto / equities the distribution will be approximate
unless you re-tune.
- **v2 direction mapping is improved but not perfect.** Diagonal
score = 2 / 5 (vs 1 / 5 in v1.1). BULL_CALM and BEAR_CALM now point
the right way; RANGE drifts slightly up; CHOPPY and STRESS still
show counter-intuitive mean-reversion edge. **Treat the indicator
as a labeled regime overlay for research, not a stand-alone
trade-direction signal.**
- **Streaming lags exist.** Pivot detection lag ≈ 3 bars,
hysteresis commit lag ≈ `min_run − 1` bars, multi-TF RSI uses
`lookahead=barmerge.lookahead_off` so higher-TF RSI updates only on
higher-TF bar close.
- **S/R level memory capped at 200 (FIFO).** The Python reference
keeps levels unbounded; on very long charts you may see different
bounce decisions than the offline version near the cap.
## Alerts
Seven alert conditions ship in:
- Regime → STRESS
- Regime → BULL_CALM
- Regime → BEAR_CALM
- Regime → RANGE
- Regime → CHOPPY
- Double-bounce UP (support held, second touch)
- Double-bounce DOWN (resistance held, second touch)
Regime alerts fire on transition (state differs from previous bar).
Double-bounce alerts fire at the second-touch bar of any cluster.
## Credits & attribution
- **Visual style** inspired by *FX Market Sessions* by **boitoki**
(Mozilla Public License 2.0). The segment-box-per-run pattern is
borrowed from that script; all classifier logic, feature math,
pivot state machine and S/R clustering is original to SAO · RUBIQ.
- **RUBIQ thesis** — *Rubik's-cube model of the market*: the right
algorithm in the wrong market phase still fails. Distilled from
100+ failed variants across the SAO portfolio.
- **Built by** SNP420 · Jarvis Claudos · Finexus s.r.o.
## License
Same as the parent SAO_RUBIQ project. Use freely, modify freely,
attribute when republishing.
---
*"Trh je proměnlivé prostředí. RUBIQ je centrální nervová soustava,
která to řeší pro všechny SAO strategie."*
— SNP420
**by SNP420 · Jarvis Claudos · Finexus s.r.o.**
*Pine Script v6 · Build 2026-05-26*
---
## What is it
A research-grade market regime overlay that classifies every confirmed
bar into one of **five regimes** and paints the chart accordingly.
Born out of a simple insight:
> *"The right algorithm in the wrong phase of the market still fails —
> even when it is 100% correct under ideal conditions."*
This is the central thesis of the **SAO · RUBIQ** project: every
strategy lives or dies inside a specific regime × timeframe cell.
Without active regime perception, every trading system is necessarily
under-performing on the bars it was not designed for. This indicator
makes those cells visible.
## What's new in v2
v2 throws out the v1 AND-stack of ADX / BB-width / drift cutoffs and
replaces it with four orthogonal signals that vote together:
- **Fractal pivots** (3-bar centered) labeled **HH / LH / HL / LL**
- **Trend state machine** with **re-anchoring**:
UP / DOWN / RANGE / REVERSAL_UP / REVERSAL_DOWN
- **Multi-TF RSI** on M30 / H1 / H4 / D1 / W1 with per-TF thresholds
and a **≥3 of 5 agreement** rule
- **S/R level clustering** (±0.3 ATR tolerance) with **double-bounce**
reversal flag that persists for 10 bars
Measured improvements over v1.1 on EUR/USD M30, 2024-2025 (24,863 bars):
| Metric | v1.1 calibrated | v2 |
| ----------------------- | --------------- | ----------- |
| Confident coverage | 44.87% | **53.96%** |
| Transitions (less whipsaw) | 1,567 | **855** |
| BULL_CALM Sharpe (h=5) | +0.085 | **+0.271** |
| BEAR_CALM direction | +0.04 ⚠ (wrong)| **−0.19 ✅** |
| Shuffle F-stat (h=5) | 0.0027 | **0.0068** |
| Strategy diagonal score | 1 / 5 | **2 / 5** |
All validated by 1000-iter shuffle Monte Carlo (p < 0.001) and a
strategy × regime P&L matrix.
## The five regimes
| Regime | Meaning |
| ------------- | -------------------------------------------------------- |
| **BULL_CALM** | Trend UP + ≥3 of 5 RSI TFs agree bullish |
| **BEAR_CALM** | Trend DOWN + ≥3 of 5 RSI TFs agree bearish |
| **RANGE** | Mixed last pivots (no clean HH+HL or LH+LL) |
| **CHOPPY** | Trend UP/DOWN without RSI agreement, or active reversal |
| **STRESS** | Volatility spike: rv20_norm > 2 OR atr_z > 2 |
| *UNCERTAIN* | Fallback — pre-warmup or no labels yet (no box drawn) |
Priority order (highest wins): **STRESS > BULL > BEAR > CHOPPY > RANGE.**
## How it works
For every confirmed bar the indicator:
1. Computes ATR(14), realized vol (20), z-scored ATR — used by stress
detection and S/R cluster tolerance.
2. Requests RSI(14) on five timeframes via `request.security`
(M30 / H1 / H4 / D1 / W1) and counts bull / bear agreements.
3. Detects fractal pivots with `ta.pivothigh / pivotlow(3, 3)` and
labels each as HH / LH / HL / LL versus the previous same-type
pivot.
4. Adds each pivot to a greedy S/R level cluster (±0.3 ATR). On the
second touch within 500 bars it fires a **double-bounce reversal**
flag (type H → REVERSAL_DOWN, type L → REVERSAL_UP) that persists
for 10 bars.
5. Maintains a trend state machine: **UP** when last H=HH and last L=HL,
**DOWN** when last H=LH and last L=LL, **RANGE** otherwise.
6. Combines trend × RSI × stress into the final label, then applies
a hysteresis smoother (`min_run = 5 bars`) so single-bar flips
never get committed.
## Visualization (FX-Sessions-style)
- **Dashed segment box** per confirmed regime run, sized to that
segment's high/low.
- **Background tint** — semi-transparent regime color over the span.
- **Bar color** (off by default) — paints OHLC bars with regime color.
- **Segment label** anchored to the top of each box.
- **Pivot markers** — HH / LH / HL / LL drawn at every confirmed pivot
with bull/bear tint.
- **Reversal arrows** — ▲ green at support double-bounces, ▼ red at
resistance double-bounces.
- **Info table** (top-right) — current state, trend, last H and L labels,
RSI per TF, bull/bear agree counts, rv20_norm, atr_z.
- **Stats table** (bottom-right) — N bars and % share per regime across
the visible history.
UNCERTAIN bars deliberately render no box and no tint (clean chart).
## Settings worth knowing
- **Hysteresis min_run** (5) — bars of consistent raw state before
commit. Bigger = less flicker, more boundary lag.
- **Pivot left/right** (3) — fractal pivot window. Smaller catches
more pivots but more noise.
- **Per-TF RSI thresholds** — defaults are 65/35 (M30), 62/38 (H1),
60/40 (H4), 55/45 (D1), 50/50 (W1). Overridable per pair / TF.
- **RSI agree min** (3) — TFs that must agree to qualify as
BULL_CALM / BEAR_CALM. Lower = more sensitive.
- **S/R cluster tolerance** (0.3 × ATR) — width of an S/R level zone.
- **S/R max age** (500 bars) — oldest first-touch still eligible for
double-bounce reversal.
- **Reversal persistence** (10 bars) — how long after the second touch
the REVERSAL state stays active.
- **Max active S/R levels** (200) — FIFO ceiling on level memory.
- **All six regime colors + reversal arrow colors** — fully overridable.
## Sanity-check expectation (EUR/USD M30, 2 years)
If the v2 baseline distribution holds on your data window:
| State | Share |
| --------- | ------ |
| BULL_CALM | ~2.6% |
| BEAR_CALM | ~1.8% |
| RANGE | ~2.4% |
| CHOPPY | ~38.7% |
| STRESS | ~8.4% |
| UNCERTAIN | ~46.0% |
v2 deliberately has stricter BULL / BEAR (needs trend state + 3-of-5
RSI agreement) and a wider CHOPPY (catches trend bars without RSI
agreement + all active reversals). Cleaner regime blocks, better
direction mapping.
## Honest limitations — please read
- **Tuned to EUR/USD M30 vol scale.** Features and rules are
TF-agnostic in math, but the thresholds were calibrated on M30.
On H1 / D1 / crypto / equities the distribution will be approximate
unless you re-tune.
- **v2 direction mapping is improved but not perfect.** Diagonal
score = 2 / 5 (vs 1 / 5 in v1.1). BULL_CALM and BEAR_CALM now point
the right way; RANGE drifts slightly up; CHOPPY and STRESS still
show counter-intuitive mean-reversion edge. **Treat the indicator
as a labeled regime overlay for research, not a stand-alone
trade-direction signal.**
- **Streaming lags exist.** Pivot detection lag ≈ 3 bars,
hysteresis commit lag ≈ `min_run − 1` bars, multi-TF RSI uses
`lookahead=barmerge.lookahead_off` so higher-TF RSI updates only on
higher-TF bar close.
- **S/R level memory capped at 200 (FIFO).** The Python reference
keeps levels unbounded; on very long charts you may see different
bounce decisions than the offline version near the cap.
## Alerts
Seven alert conditions ship in:
- Regime → STRESS
- Regime → BULL_CALM
- Regime → BEAR_CALM
- Regime → RANGE
- Regime → CHOPPY
- Double-bounce UP (support held, second touch)
- Double-bounce DOWN (resistance held, second touch)
Regime alerts fire on transition (state differs from previous bar).
Double-bounce alerts fire at the second-touch bar of any cluster.
## Credits & attribution
- **Visual style** inspired by *FX Market Sessions* by **boitoki**
(Mozilla Public License 2.0). The segment-box-per-run pattern is
borrowed from that script; all classifier logic, feature math,
pivot state machine and S/R clustering is original to SAO · RUBIQ.
- **RUBIQ thesis** — *Rubik's-cube model of the market*: the right
algorithm in the wrong market phase still fails. Distilled from
100+ failed variants across the SAO portfolio.
- **Built by** SNP420 · Jarvis Claudos · Finexus s.r.o.
## License
Same as the parent SAO_RUBIQ project. Use freely, modify freely,
attribute when republishing.
---
*"Trh je proměnlivé prostředí. RUBIQ je centrální nervová soustava,
která to řeší pro všechny SAO strategie."*
— SNP420
Skrip open-source
Dengan semangat TradingView yang sesungguhnya, pembuat skrip ini telah menjadikannya sebagai sumber terbuka, sehingga para trader dapat meninjau dan memverifikasi fungsinya. Salut untuk penulisnya! Meskipun Anda dapat menggunakannya secara gratis, perlu diingat bahwa penerbitan ulang kode ini tunduk pada Tata Tertib kami.
24 yoooo!
SNP_420 + Jarvis
5 in investing and 4,5 in traiding
Czechia
SNP_420 + Jarvis
5 in investing and 4,5 in traiding
Czechia
Pernyataan Penyangkalan
Informasi dan publikasi ini tidak dimaksudkan, dan bukan merupakan, saran atau rekomendasi keuangan, investasi, trading, atau jenis lainnya yang diberikan atau didukung oleh TradingView. Baca selengkapnya di Ketentuan Penggunaan.
Skrip open-source
Dengan semangat TradingView yang sesungguhnya, pembuat skrip ini telah menjadikannya sebagai sumber terbuka, sehingga para trader dapat meninjau dan memverifikasi fungsinya. Salut untuk penulisnya! Meskipun Anda dapat menggunakannya secara gratis, perlu diingat bahwa penerbitan ulang kode ini tunduk pada Tata Tertib kami.
24 yoooo!
SNP_420 + Jarvis
5 in investing and 4,5 in traiding
Czechia
SNP_420 + Jarvis
5 in investing and 4,5 in traiding
Czechia
Pernyataan Penyangkalan
Informasi dan publikasi ini tidak dimaksudkan, dan bukan merupakan, saran atau rekomendasi keuangan, investasi, trading, atau jenis lainnya yang diberikan atau didukung oleh TradingView. Baca selengkapnya di Ketentuan Penggunaan.