OPEN-SOURCE SCRIPT
Güncellendi Multi-Axis Confluence Matrix Adaptive MTF & Forward Calibrated
Multi-Axis Confluence Matrix — Adaptive MTF & Forward-Calibrated
What it is
A confluence oscillator that scores agreement across five independent information axes, each measured on three adaptively-resolved timeframes, then weights every axis by its own forward-measured edge on the current symbol so the score self-corrects instead of treating all inputs as equally reliable. It separates the two kinds of multi-timeframe agreement that most stacks wrongly merge, gates direction with a higher-timeframe bias, and reports the whole state in one plain-language verdict panel.
The plotted line reads like a conventional oscillator (−100 to +100, with strong-zone bands and price-chart BUY/SELL signals), so a non-technical user can interpret it at a glance — while the engine underneath is doing edge-weighted, multi-timeframe, forward-calibrated confluence.
Why these components are combined (mashup rationale)
Stacking indicators usually produces false confluence: five momentum tools in different clothes vote the same bet five times, and it looks like five confirmations when it is really one. This script is built specifically to avoid that, and every layer removes a distinct failure of naive stacking. They ship as one engine because direction, certainty, and agreement are one decision — splitting them across separate scripts would lose the cross-checks below.
1. Five orthogonal axes, each from a different data source. Agreement only carries information if the axes are independent. Each axis reads a different thing:
Momentum — price (normalized MACD-histogram sign)
Stretch — price vs an adaptive fair value (z-score of the deviation)
Order flow — signed volume (sign of a bounded cumulative volume-delta z-score)
Fear — an implied-volatility index (direction of a negated IV trend)
Cross-market — a correlated leader instrument (signed momentum), off by default
Because no two axes share an input, four of them agreeing is four independent witnesses, not one witness repeated.
2. Vote / gate split. Directional axes vote long or short. Context does not vote — it scales conviction or vetoes. A volatility-regime measure and a trend-efficiency measure only shrink conviction; the higher-timeframe bias only decides which side is allowed. High volatility is neither bullish nor bearish, so folding it into the direction sum would inject bias. Keeping context as a multiplier rather than an addend is the single thing that stops the stack collapsing into "momentum with extra steps."
3. Two separate confluence counts, never summed.
Method confluence = how many independent axes agree on the signal timeframe (the rare, high-information agreement).
Scale confluence = how many timeframes one axis agrees on (robustness/timing only — the same signal at different resolutions, not a new vote).
Merging them lets three timeframes of one axis drown out a contradicting axis. They are reported on separate rows so you can see which kind of agreement you actually have.
4. Adaptive multi-timeframe resolution. The bias / signal / trigger timeframes are derived as multiples of the chart timeframe, clamped and snapped to real frames. The "bias-above, vote-between, trigger-at" relationship is preserved on any chart, instead of fixed frames that are only correct on the one chart they were tuned for.
5. Per-axis forward calibration (the core original component). Each axis is calibrated on each timeframe by a triple-barrier outcome (profit barrier / stop barrier / time horizon), with sample-uniqueness weighting and recency decay, reported as an edge over a barrier-matched base rate with a Wilson confidence interval. Each axis then votes weighted by its own edge lower-bound, so an axis that is not paying on the current symbol contributes little and cannot bias the score merely by being present. A small weight floor keeps the engine responsive (edge tilts the score rather than fully gating it); set the floor to zero for strict gating.
6. Higher-timeframe bias gate. The slow frame sets which side is allowed and does not vote, which blocks the counter-trend entries that punish mean-reversion approaches.
Remove any one layer and the stack loses a check it cannot recover. That is the justification for combining them.
How to use it
The line lives in −100..+100. In the green zone = strong bullish confluence; red zone = strong bearish; mid = no edge. Read it like any oscillator.
BUY / SELL triangles print on the price chart only when the gated, edge-weighted verdict fires (method confluence + conviction + bias gate all pass). Triangles with text = high conviction; without text = weaker. A latch + cooldown gives one clean mark per swing, not a cluster.
Bull / Bear divergence is drawn on the score line, spaced by a cooldown, and suppressed only when it directly contradicts a high-conviction opposite verdict (early-reversal divergences against a weak score are kept).
The verdict panel (top-right, on by default) states the action in plain language: tier, what to do, method/scale confluence, conviction, calibration warm-state, and which side the higher-TF bias allows.
The axis × timeframe matrix is an advanced view, off by default — turn it on to inspect each axis's arrows across timeframes and its measured edge. A Key-info table (Compact by default, Pro for full detail) summarizes the live state on the left.
Any market: the price source, volume source (with a borrow-symbol option for instruments that report no volume), the implied-volatility symbol, and the cross-market leader are all inputs in the settings. Defaults suit a major index future; change the symbols for any market, or clear the fear / cross-market symbols to drop those axes (the score re-weights over the axes that remain).
What makes it original
The multi-axis matrix, the vote/gate split, the dual (method vs scale) confluence accounting, the adaptive timeframe resolution, and the per-axis-per-timeframe forward edge-weighting are this script's own construction. It is not a re-skin of a single classic indicator — it is an engine for organizing several independent reads into one honest, self-correcting confluence score.
Outputs (Data Window)
Generic, identifiable names for use by other scripts: direction, conviction, method confluence, scale confluence, higher-TF bias, regime direction, edge, edge lower-bound, and sample count.
Non-repainting
Every off-chart-timeframe read uses confirmed previous bars with lookahead disabled; votes confirm on bar close; every calibrated statistic is forward-measured at fixed barriers on confirmed bars. The live bar is provisional, as with any indicator.
Concept credits (techniques operationalized here — not third-party code)
MACD: Gerald Appel. Bands / %B style stretch: John Bollinger. Cumulative volume delta and divergence: standard order-flow practice. Implied-volatility index methodology: CBOE. Trade-side from standardized returns (Bulk Volume Classification): Easley, López de Prado & O'Hara. Triple-barrier labeling and sample uniqueness: López de Prado. Wilson score interval: Edwin B. Wilson. Efficiency-ratio regime: Perry Kaufman. The matrix architecture and the calibration/weighting scheme are the author's own.
Disclaimer
Research and education only. This is not financial advice, not a recommendation, and not a guarantee. The axes are deliberately compact reads of their concepts. Calibration is in-sample, forward-measured at fixed barriers, with no costs, slippage, or guaranteed fills. Indicators describe past behavior; they do not predict the future. Trading carries risk of loss. Test out-of-sample and make your own decisions.
What it is
A confluence oscillator that scores agreement across five independent information axes, each measured on three adaptively-resolved timeframes, then weights every axis by its own forward-measured edge on the current symbol so the score self-corrects instead of treating all inputs as equally reliable. It separates the two kinds of multi-timeframe agreement that most stacks wrongly merge, gates direction with a higher-timeframe bias, and reports the whole state in one plain-language verdict panel.
The plotted line reads like a conventional oscillator (−100 to +100, with strong-zone bands and price-chart BUY/SELL signals), so a non-technical user can interpret it at a glance — while the engine underneath is doing edge-weighted, multi-timeframe, forward-calibrated confluence.
Why these components are combined (mashup rationale)
Stacking indicators usually produces false confluence: five momentum tools in different clothes vote the same bet five times, and it looks like five confirmations when it is really one. This script is built specifically to avoid that, and every layer removes a distinct failure of naive stacking. They ship as one engine because direction, certainty, and agreement are one decision — splitting them across separate scripts would lose the cross-checks below.
1. Five orthogonal axes, each from a different data source. Agreement only carries information if the axes are independent. Each axis reads a different thing:
Momentum — price (normalized MACD-histogram sign)
Stretch — price vs an adaptive fair value (z-score of the deviation)
Order flow — signed volume (sign of a bounded cumulative volume-delta z-score)
Fear — an implied-volatility index (direction of a negated IV trend)
Cross-market — a correlated leader instrument (signed momentum), off by default
Because no two axes share an input, four of them agreeing is four independent witnesses, not one witness repeated.
2. Vote / gate split. Directional axes vote long or short. Context does not vote — it scales conviction or vetoes. A volatility-regime measure and a trend-efficiency measure only shrink conviction; the higher-timeframe bias only decides which side is allowed. High volatility is neither bullish nor bearish, so folding it into the direction sum would inject bias. Keeping context as a multiplier rather than an addend is the single thing that stops the stack collapsing into "momentum with extra steps."
3. Two separate confluence counts, never summed.
Method confluence = how many independent axes agree on the signal timeframe (the rare, high-information agreement).
Scale confluence = how many timeframes one axis agrees on (robustness/timing only — the same signal at different resolutions, not a new vote).
Merging them lets three timeframes of one axis drown out a contradicting axis. They are reported on separate rows so you can see which kind of agreement you actually have.
4. Adaptive multi-timeframe resolution. The bias / signal / trigger timeframes are derived as multiples of the chart timeframe, clamped and snapped to real frames. The "bias-above, vote-between, trigger-at" relationship is preserved on any chart, instead of fixed frames that are only correct on the one chart they were tuned for.
5. Per-axis forward calibration (the core original component). Each axis is calibrated on each timeframe by a triple-barrier outcome (profit barrier / stop barrier / time horizon), with sample-uniqueness weighting and recency decay, reported as an edge over a barrier-matched base rate with a Wilson confidence interval. Each axis then votes weighted by its own edge lower-bound, so an axis that is not paying on the current symbol contributes little and cannot bias the score merely by being present. A small weight floor keeps the engine responsive (edge tilts the score rather than fully gating it); set the floor to zero for strict gating.
6. Higher-timeframe bias gate. The slow frame sets which side is allowed and does not vote, which blocks the counter-trend entries that punish mean-reversion approaches.
Remove any one layer and the stack loses a check it cannot recover. That is the justification for combining them.
How to use it
The line lives in −100..+100. In the green zone = strong bullish confluence; red zone = strong bearish; mid = no edge. Read it like any oscillator.
BUY / SELL triangles print on the price chart only when the gated, edge-weighted verdict fires (method confluence + conviction + bias gate all pass). Triangles with text = high conviction; without text = weaker. A latch + cooldown gives one clean mark per swing, not a cluster.
Bull / Bear divergence is drawn on the score line, spaced by a cooldown, and suppressed only when it directly contradicts a high-conviction opposite verdict (early-reversal divergences against a weak score are kept).
The verdict panel (top-right, on by default) states the action in plain language: tier, what to do, method/scale confluence, conviction, calibration warm-state, and which side the higher-TF bias allows.
The axis × timeframe matrix is an advanced view, off by default — turn it on to inspect each axis's arrows across timeframes and its measured edge. A Key-info table (Compact by default, Pro for full detail) summarizes the live state on the left.
Any market: the price source, volume source (with a borrow-symbol option for instruments that report no volume), the implied-volatility symbol, and the cross-market leader are all inputs in the settings. Defaults suit a major index future; change the symbols for any market, or clear the fear / cross-market symbols to drop those axes (the score re-weights over the axes that remain).
What makes it original
The multi-axis matrix, the vote/gate split, the dual (method vs scale) confluence accounting, the adaptive timeframe resolution, and the per-axis-per-timeframe forward edge-weighting are this script's own construction. It is not a re-skin of a single classic indicator — it is an engine for organizing several independent reads into one honest, self-correcting confluence score.
Outputs (Data Window)
Generic, identifiable names for use by other scripts: direction, conviction, method confluence, scale confluence, higher-TF bias, regime direction, edge, edge lower-bound, and sample count.
Non-repainting
Every off-chart-timeframe read uses confirmed previous bars with lookahead disabled; votes confirm on bar close; every calibrated statistic is forward-measured at fixed barriers on confirmed bars. The live bar is provisional, as with any indicator.
Concept credits (techniques operationalized here — not third-party code)
MACD: Gerald Appel. Bands / %B style stretch: John Bollinger. Cumulative volume delta and divergence: standard order-flow practice. Implied-volatility index methodology: CBOE. Trade-side from standardized returns (Bulk Volume Classification): Easley, López de Prado & O'Hara. Triple-barrier labeling and sample uniqueness: López de Prado. Wilson score interval: Edwin B. Wilson. Efficiency-ratio regime: Perry Kaufman. The matrix architecture and the calibration/weighting scheme are the author's own.
Disclaimer
Research and education only. This is not financial advice, not a recommendation, and not a guarantee. The axes are deliberately compact reads of their concepts. Calibration is in-sample, forward-measured at fixed barriers, with no costs, slippage, or guaranteed fills. Indicators describe past behavior; they do not predict the future. Trading carries risk of loss. Test out-of-sample and make your own decisions.
Sürüm Notları
v1.1 — Correctness + licence pass- Correctness: on the fear axis, ta.atr was called from inside a "skip if no fear-symbol data" check,
so it could be skipped on some bars and go inconsistent (CW10004). It now runs every bar; when the
implied-volatility symbol is present the output is unchanged.
- Cleared a second benign warning by hoisting the score-line smoothing average out of a constant
guard. Output is identical.
- Added the MPL-2.0 licence header.
- The five axes remain self-contained by design — each is re-evaluated on multiple timeframes inside
request.security, which an external connected source cannot be — so the multi-timeframe matrix is
preserved. The engine's own EXP_ outputs are unchanged.
Descriptive research tooling, in-sample per-axis calibration, not investment advice.
Açık kaynak kodlu komut dosyası
Gerçek TradingView ruhuyla, bu komut dosyasının mimarı, yatırımcıların işlevselliğini inceleyip doğrulayabilmesi için onu açık kaynaklı hale getirdi. Yazarı tebrik ederiz! Ücretsiz olarak kullanabilseniz de, kodu yeniden yayınlamanın Topluluk Kurallarımıza tabi olduğunu unutmayın.
Feragatname
Bilgiler ve yayınlar, TradingView tarafından sağlanan veya onaylanan finansal, yatırım, alım satım veya diğer türden tavsiye veya öneriler anlamına gelmez ve teşkil etmez. Kullanım Koşulları bölümünde daha fazlasını okuyun.
Açık kaynak kodlu komut dosyası
Gerçek TradingView ruhuyla, bu komut dosyasının mimarı, yatırımcıların işlevselliğini inceleyip doğrulayabilmesi için onu açık kaynaklı hale getirdi. Yazarı tebrik ederiz! Ücretsiz olarak kullanabilseniz de, kodu yeniden yayınlamanın Topluluk Kurallarımıza tabi olduğunu unutmayın.
Feragatname
Bilgiler ve yayınlar, TradingView tarafından sağlanan veya onaylanan finansal, yatırım, alım satım veya diğer türden tavsiye veya öneriler anlamına gelmez ve teşkil etmez. Kullanım Koşulları bölümünde daha fazlasını okuyun.