Odin AI
Institutional-Style Trend Intelligence & Trade Decision Engine
Version 2.6.1 | Plazo Sullivan Roche Capital (PSRC)
Product category: Multi-Timeframe Technical Analysis, Trend Intelligence, Market Regime Classification and Trade Planning
Supported platform: TradingView — Pine Script v6
Recommended chart timeframe: 15 minutes
Developer: Plazo Sullivan Roche Capital
1. Executive Overview
Markets rarely move because of a single technical indicator. Price trends, market structure, liquidity conditions, volatility, momentum and trading participation interact continuously. A promising entry signal can fail when these forces disagree.
Odin AI was developed to address this challenge.
Odin AI is a multi-timeframe trading intelligence system that combines a refined MFTrend signal engine with six weighted institutional-style market factors.
Rather than presenting a simple BUY or SELL signal in isolation, Odin evaluates the market environment surrounding that signal and provides a structured trade plan with entry, stop-loss and three take-profit levels.
The system is designed for discretionary traders, systematic trading researchers, proprietary trading participants and quantitative trading teams seeking a more disciplined framework for evaluating directional opportunities.
Odin's central principle is straightforward:
A technical trigger identifies an opportunity. Market context evaluates its quality. Risk geometry determines whether the opportunity can become a trade.
The system does not claim to predict prices with certainty, access private institutional order flow or guarantee profitability.
Download the Technical Whitepaper
Python Multi Model Signal Confirmation Engine
Download the Confirmation Engine for ChatGPT, Claude, Grok Integration
The Python service will receive confirmed TradingView BUY/SELL alerts and independently ask ChatGPT (OpenAI), Claude (Anthropic), or Grok (xAI) to review the signal's six-factor evidence, stop-loss geometry, and risk conditions.
Crucially, the models will act as a secondary analytical review layer, not as a replacement for Odin's deterministic calculations. A model can approve, reject, or abstain from a proposed trade, but it cannot establish the current market price or independently verify Odin's readings without additional market data.
The engine will support individual model reviews and a three-model consensus, with configurable voting rules and a fail-closed default when evidence is insufficient.
How it Works:
Each model returns APPROVE, REJECT, or ABSTAIN, with reasoning and risk flags. The default consensus requires unanimous approval. API failures do not produce approval.
Quick start
Set the API keys in your environment, then launch:
python odin_confirm.py \
--mode webhook \
--providers openai,claude,grok \
--consensus unanimous \
--min-approvals 3 \
--port 8080
You'll also need to set ODIN_WEBHOOK_SECRET and expose the listener through a secure HTTPS reverse proxy. In TradingView, select Odin's Any alert() function call condition and point its webhook to your server.
You can run OpenAI, Claude, or Grok individually, or combine them. The model identifiers are configurable because API availability changes.
2. Core Architecture
Odin AI operates through three distinct analytical layers.
Layer One: MFTrend Signal Generation
The primary signal engine is derived from MFTrend, a volatility-adjusted trend-following framework.
It combines:
Average True Range (ATR) to establish adaptive price boundaries.
Money Flow Index (MFI) to incorporate price and reported volume conditions.
Relative Strength Index (RSI) as an alternative when reliable volume data are unavailable.
A trailing trend line that adapts to changes in directional conditions.
The MFTrend line advances according to the prevailing oscillator condition and ATR-derived boundaries.
The system compares the current MFTrend value against its value two bars earlier.
A bullish crossover creates a potential BUY trigger; a bearish crossunder creates a potential SELL trigger.
Odin also retains the original MFTrend signal-suppression logic to reduce certain repeated or conflicting signals.
Important: The MFTrend trigger is not itself a trade authorization. It is the starting point for the institutional decision and trade-geometry layers.
Layer Two: Omega Six-Factor Intelligence
The second layer evaluates six complementary dimensions of market behavior.
Each factor contributes a defined weight to the overall directional assessment.
Factor | Weight |
Institutional Daily Hypothesis | 30% |
H1 Market Structure | 20% |
H1 Auction and Liquidity Context | 15% |
Markov Market Regime | 15% |
Volume Pressure | 10% |
Momentum | 10% |
Total | 100% |
These factors produce a normalized directional score.
Positive readings indicate bullish pressure, negative readings indicate bearish pressure, and readings near zero indicate mixed or neutral conditions.
The weighted score is an analytical measure, not a calibrated probability of winning.
Layer Three: Trade Qualification and Risk Geometry
The final layer determines whether a signal has a usable trade structure.
It evaluates:
Whether the stop-loss is on the correct side of the entry.
Whether the selected stop is within the maximum permitted ATR distance.
Whether confirmed structural levels are available.
Whether the take-profit configuration satisfies the selected requirements.
Whether the selected institutional confirmation mode permits the trade.
Only signals passing the applicable institutional and geometry conditions become trade-qualified BUY or SELL signals.
This separation helps distinguish a technically interesting setup from a trade with an acceptable risk structure.
3. The Six Institutional Intelligence Factors
Factor 1 — Institutional Daily Hypothesis (30%)
The Daily Hypothesis establishes the prevailing higher-timeframe directional framework.
It analyzes the previous completed daily candle using three components:
Price location: Where the previous daily close stands relative to the daily 50-period EMA, normalized by daily ATR.
Moving-average alignment: The relationship between the daily 20-period EMA and daily 50-period EMA, normalized by ATR.
Daily candle conviction: The previous candle's body direction and magnitude relative to its full trading range.
The model combines these measurements using internal weights of 50%, 30% and 20%, respectively.
The resulting daily score is classified as bullish, bearish or neutral using a configurable threshold.
The default threshold is 28.
Interpretation
Bullish: The completed daily evidence supports an upward directional hypothesis.
Bearish: The completed daily evidence supports a downward directional hypothesis.
Neutral: The evidence is insufficiently directional.
The Daily Hypothesis is the largest component of Odin's weighted intelligence model because higher-timeframe directional context can materially influence lower-timeframe trade behavior.
Best practice: Give greater attention to signals aligned with the daily hypothesis, particularly when the H1 structure and auction readings confirm that direction.
The daily calculation uses completed daily information rather than an unfinished daily candle.
Factor 2 — H1 Market Structure (20%)
Market structure examines the organization of price movement rather than relying solely on oscillator readings.
Odin evaluates H1 price behavior using confirmed structural events, breakout context and configurable lookback and expiry parameters.
The objective is to distinguish directional continuation from conflicting or inconclusive structure.
Bullish structure indicates that the H1 structural evidence favors upward movement.
Bearish structure indicates that the evidence favors downward movement.
Neutral structure indicates that a reliable directional structural conclusion is unavailable.
The default H1 breakout lookback is 20 bars.
Best practice: Favor signals where the H1 structural reading supports the proposed trade. When structure opposes the daily hypothesis, treat the environment as potentially transitional rather than automatically assuming the daily hypothesis is wrong.
Factor 3 — H1 Auction and Liquidity Context (15%)
Price movement reflects the interaction of buying and selling interest across different price levels.
Odin's auction component evaluates price acceptance and liquidity-related behavior using observable market-price evidence.
Its purpose is to assess whether the market is accepting movement in a particular direction or showing evidence of rejection and opposition.
The model includes configurable H1 acceptance requirements and structural context.
The default acceptance requirement is two H1 closes.
Bullish auction context suggests that price behavior supports upward acceptance.
Bearish auction context suggests that price behavior supports downward acceptance.
Neutral auction context indicates insufficient directional evidence.
This is an institutional-style auction proxy, not a direct measurement of exchange order-book liquidity or proprietary institutional transactions.
Best practice: Strong alignment between Daily Hypothesis, H1 Structure and H1 Auction provides a more coherent directional thesis than an isolated MFTrend crossover.
Factor 4 — Markov Market Regime (15%)
Not all trading environments behave alike.
A trend-following strategy may perform differently during persistent directional movement, consolidation, volatility expansion or transitions between market states.
Odin includes an empirical Markov transition model that estimates the behavior and persistence of observable market regimes.
The model distinguishes five states:
Uptrend
Downtrend
Range
High volatility
Transition
It evaluates historical state transitions using configurable decay, transition sample requirements, persistence and normalized entropy.
Persistence measures the tendency of a regime to continue.
Entropy measures the uncertainty or dispersion of estimated state transitions.
A stable classification requires sufficient supporting observations and acceptable persistence and entropy conditions.
Default parameters include:
Minimum transition-row mass: 30
Transition decay: 0.995
Minimum persistence: 0.52
Maximum normalized entropy: 0.82
The model is an empirical state-transition framework, not a fitted hidden Markov model or an independently trained machine-learning predictor.
Best practice: Treat transition and high-uncertainty conditions cautiously. A directional vote can exist even when the underlying regime is not classified as stable.
Factor 5 — Volume Pressure (10%)
Volume provides useful information about market participation, although its interpretation depends on the instrument and data source.
Odin evaluates H1 volume-related pressure and relative participation using smoothed measurements.
The default volume-pressure smoothing period is eight H1 bars, with a minimum relative-volume parameter of 0.65.
A bullish reading indicates supportive upward participation pressure.
A bearish reading indicates supportive downward participation pressure.
A neutral reading indicates insufficient or mixed evidence.
Important: Forex and CFD feeds may provide tick volume rather than centralized exchange volume. Odin's volume factor must not be interpreted as verified institutional buying or selling.
Best practice: Use volume as supporting evidence, not as a standalone directional prediction.
Factor 6 — Momentum (10%)
Momentum measures the strength and direction of recent price movement.
Odin evaluates H1 momentum using a configurable lookback, defaulting to five bars.
The momentum component helps identify whether directional movement supports or contradicts the broader market hypothesis.
Bullish momentum: Recent directional movement favors buyers.
Bearish momentum: Recent directional movement favors sellers.
Neutral momentum: No sufficiently strong directional reading.
Momentum receives a smaller weight than Daily Hypothesis and Market Structure because short-term acceleration can reverse quickly and may overlap with other trend-related evidence.
Best practice: Momentum is most useful as confirmation of an existing directional thesis rather than as justification for entering against all other factors.
4. Weighted Decision Engine
Odin combines the six factors using the following structure:
Weighted Decision =
30% Daily Hypothesis
20% H1 Structure
15% H1 Auction
15% Markov Regime
10% Volume Pressure
10% Momentum
Each factor contributes a normalized signed score.
The final score is directional and typically interpreted on a scale from approximately -100 to +100.
The default final decision threshold is 45.
A score at or above +45 indicates a bullish weighted decision.
A score at or below -45 indicates a bearish weighted decision.
Values between these thresholds indicate a neutral or insufficiently directional weighted conclusion.
Evidence coverage is displayed separately from directional strength.
Coverage measures how much of the weighted evidence is available. It is not a confidence interval or a prediction accuracy statistic.
The default minimum evidence coverage for Omega qualification is 85%.
A strong weighted score does not automatically mean that Odin has granted full trading permission. The complete Omega permission logic also evaluates structural confirmation, higher-timeframe agreement and other conditions.
5. Four Selectable Confirmation Modes
Odin provides four modes so traders can choose how strictly the six-factor framework influences MFTrend signals.
Mode | Purpose | Behavior |
Informational | Research and discretionary analysis | Omega readings are displayed but do not veto MFTrend triggers |
Daily Veto | Higher-timeframe directional discipline | Requires valid Omega evidence, daily directional alignment and no daily invalidation |
Conflict Veto | Detect major H1 opposition | Rejects a signal when both H1 Structure and H1 Auction oppose its direction |
Full Omega Gate | Strict institutional-style permission | Requires complete Omega permission and alignment with the Daily Hypothesis |
Informational Mode — Default
This mode preserves the widest opportunity set.
MFTrend triggers are evaluated without an Omega directional veto, although stop-loss and target geometry must still qualify for a trade signal to appear.
Recommended for baseline research, discretionary traders and initial forward testing.
Daily Veto
Signals opposing the valid daily hypothesis are rejected.
This mode also requires available Omega evidence and checks the daily invalidation condition.
Recommended for traders who prioritize higher-timeframe alignment.
Conflict Veto
Signals are rejected when both H1 Market Structure and H1 Auction oppose the proposed trade direction.
This mode requires available Omega evidence but does not automatically reject every counter-daily signal.
Recommended for traders who want to avoid simultaneous H1 structural and auction opposition without imposing complete daily alignment.
Full Omega Gate
The strictest confirmation mode.
A signal requires valid Omega evidence, completed H1 permission, daily directional alignment and no applicable daily invalidation.
Recommended for research teams studying highly selective signal qualification.
Important: Stricter gates can reduce trade frequency without necessarily improving expectancy. No confirmation mode should be presumed superior without independent testing.
6. Entry, Stop-Loss and Take-Profit System
Odin generates a trade plan for qualified signals.
Entry
The signal reference entry is based on the confirmed signal-bar closing price.
This is a planning reference, not a guaranteed executable fill. Actual orders may be filled at different prices because of spread, latency, gaps or slippage.
Stop-Loss
Odin supports three stop policies.
Strict Structure
Uses the original confirmed structural pivot selection and applicable ATR-based fallback behavior.
The stop must satisfy the configured direction and maximum-distance requirements.
This is the default policy and the primary research candidate for EURUSD.
Nearest Valid Structure
Searches stored confirmed pivots for an eligible nearby structural stop.
This can increase signal eligibility but may create stops that are excessively close to current price.
It should be treated as experimental until independently validated.
ATR Alternative
Uses an ATR-derived alternative when the original structural stop fails the applicable distance requirements.
This can provide more consistent stop distances but may sacrifice the structural invalidation rationale.
It is a secondary research candidate for XBRUSD.
Take-Profit Levels
Odin provides three target levels:
Target | Default minimum reward multiple |
TP1 | 1.5R |
TP2 | 2.5R |
TP3 | 4.0R |
R represents the initial distance between entry and stop-loss.
TP1 can use a confirmed opposing structural level when it satisfies the minimum reward requirement. Otherwise, Odin can use an R-multiple projection.
TP2 and TP3 use progressively extended R-based targets, with spacing adjusted when the structural TP1 requires it.
An optional setting requires TP1 to be structural rather than allowing the projected fallback.
These are analytical price objectives. They do not guarantee that price will reach the targets.
7. The Decision Dashboard
Odin's forward-testing dashboard intentionally focuses on actionable information rather than research statistics.
It displays:
Active confirmation mode and stop policy.
Daily Hypothesis and its 30% weighting.
H1 Structure and its 20% weighting.
H1 Auction and its 15% weighting.
Markov direction and stability.
Volume Pressure and Momentum.
Weighted directional decision.
Omega permission and evidence coverage.
Most recent signal and qualification status.
Most recent decision or rejection reason.
How to read the dashboard
A bullish Daily Hypothesis, bullish H1 Structure, bullish H1 Auction and supportive momentum indicate a coherent directional environment.
A bullish MFTrend signal against bearish Daily, Structure and Auction readings indicates significant disagreement.
The dashboard is a decision-support interface, not an autonomous trade execution or account-risk management system.
8. Recommended Timeframes
Odin uses a multi-timeframe architecture.
Timeframe | Function |
Daily (D1) | Primary directional hypothesis |
Four-hour (H4) | Higher-timeframe trend context |
One-hour (H1) | Structure, auction, regime, volume, momentum and permission |
15-minute (M15) | Recommended entry-signal chart |
5-minute (M5) | Experimental faster entry chart |
30-minute (M30) | Alternative signal chart |
One-hour (H1) | Slower signal chart |
M15 is the recommended starting point because it provides a practical balance between signal frequency and higher-timeframe context.
The current source supports standard intraday chart timeframes up to H1. Nonstandard chart types and timeframes above H1 are not supported by the current permission framework.
H4 context offers two selectable configurations:
H4 EMA200
H4 EMA20 with Daily EMA50 confirmation
The default is H4 EMA200.
Higher-timeframe calculations are based on completed bars to reduce unfinished-candle instability. This does not eliminate market risk or every possible difference between historical and realtime behavior.
9. Supported Assets and Market Suitability
Odin is designed for instruments with adequate price history, reliable market data and sufficient liquidity.
Forex
Examples include EURUSD, GBPUSD, USDJPY, AUDUSD and USDCHF.
Research status: EURUSD M15 with Strict Structure has shown the strongest historical consistency among the configurations examined so far.
This does not establish a statistically verified live trading advantage.
Precious Metals
Examples include XAUUSD and XAGUSD.
Gold can experience abrupt volatility expansion, liquidity shifts and substantial news-driven movement.
Research status: XAUUSD has not demonstrated a sufficiently consistent positive advantage in the completed research.
Equity Indices
Examples include NAS100, USTEC, US30 and SPX500.
Index instruments can be sensitive to cash-session openings, macroeconomic announcements and broker-specific CFD pricing.
Research status: USTEC has not yet demonstrated a convincing validated edge.
Energy Markets
Examples include XBRUSD and USOIL.
Energy markets can exhibit large directional movements and volatility shocks.
Research status: XBRUSD M15 using ATR Alternative has produced encouraging historical observations, but performance has varied materially between chronological periods.
Cryptocurrency
Liquid cryptocurrency instruments may be studied where TradingView provides adequate price and volume data.
However, the existing Odin research does not validate a cryptocurrency-specific trading advantage.
Asset compatibility is not the same as proven profitability. Each instrument requires separate forward testing and execution-cost calibration.
10. Recommended Starting Settings
The following settings reproduce the principal default framework.
Parameter | Starting setting |
Chart timeframe | M15 |
MFTrend ATR multiplier | 1.0 |
MFTrend common period | 14 |
Oscillator | MFI; RSI when volume is unreliable |
Legacy signal suppression | Enabled |
Confirmation mode | Informational |
Daily fast EMA | 20 |
Daily slow EMA | 50 |
Daily hypothesis threshold | 28 |
H4 context | EMA200 |
H1 breakout lookback | 20 |
H1 acceptance closes | 2 |
Final decision threshold | 45 |
H1 consecutive confirmations | 2 |
Require H4 agreement | Enabled |
Minimum evidence coverage | 85% |
Structural pivot strength | 5 |
Stop policy | Strict Structure |
Stop ATR buffer | 0.2 |
Maximum stop distance | 4.0 ATR |
TP1 / TP2 / TP3 | 1.5R / 2.5R / 4.0R |
Recent displayed signals | 2 |
Unified alerts | Enabled |
These settings are a research baseline, not universally optimized parameters.
Avoid adjusting several parameters simultaneously. Changes should be evaluated against a frozen baseline using the same instruments, dates and transaction-cost assumptions.
11. Best Practices for Traders
Trade with a defined directional thesis
Before acting on a signal, examine the Daily Hypothesis, H1 Structure and H1 Auction readings.
Understand whether the signal is aligned with or opposed to the prevailing market context.
Distinguish a signal from permission
An MFTrend trigger is not equivalent to Omega approval.
A favorable Omega score is not equivalent to full permission.
A permitted signal is not equivalent to acceptable stop-loss geometry.
Respect stop-loss invalidation
Avoid widening stops simply to keep a trade alive.
When the structural stop is excessively distant, consider whether the entry is poorly positioned rather than assuming the filter is defective.
Manage account risk independently
Odin does not calculate or enforce broker-level position sizing, portfolio exposure, account drawdown limits or automatic order execution.
Traders should establish independent account-risk limits, position-sizing rules and maximum concurrent exposure.
A conservative initial risk budget, such as 0.25%–0.50% of account equity per trade, may be appropriate for controlled paper or small-size testing, but must be adapted to individual circumstances.
Account for economic announcements
Major central-bank decisions, inflation releases, employment data and unexpected geopolitical events can produce rapid moves that overwhelm technical models.
Consider reducing exposure or suspending discretionary entries around major scheduled events.
Evaluate instruments separately
A stop policy that appears useful on EURUSD may not behave similarly on gold, oil or equity indices.
Do not assume that positive results on one asset transfer to another.
Record every trade
Maintain a journal containing the signal timestamp, instrument, confirmation mode, entry, stop, targets, actual fill, realized outcome and execution costs.
Evaluate performance using expectancy, profit factor, drawdown, sample size and chronological consistency—not win rate alone.
12. Unified TradingView Alerts
Odin includes a unified dynamic alert system for qualified BUY and SELL signals.
A single TradingView alert can monitor both directions.
Setup
Add Odin AI to a supported TradingView chart.
Enable Unified Dynamic Alerts in the indicator settings.
Select the desired confirmation mode and stop policy.
Create a TradingView alert.
Under Condition, select Odin AI and Any alert() function call.
Set the alert frequency to Once Per Bar Close.
Choose the desired notification or webhook destination.
The alert payload includes the instrument, timeframe, BUY/SELL direction, entry, SL, TP1–TP3, confirmation mode, stop policy, six-factor directional readings, weighted score, evidence coverage and signal timestamps.
The alert also identifies relevant stop and target classifications.
Alerts are generated for trade-qualified signals, not every raw MFTrend crossover.
Important: TradingView alerts are notifications. Automated execution requires a separately configured and tested execution system. Webhook receivers must validate incoming data, prevent duplicate orders and apply independent risk controls.
13. Research Findings and Current Validation Status
Odin has undergone iterative development and historical shadow-trade evaluation across several instruments.
The research compared Strict Structure, Nearest Valid Structure and ATR Alternative stop policies.
The strongest observed candidates were:
EURUSD M15 — Strict Structure
Demonstrated positive modeled expectancy across the examined Development, Validation and Holdout date partitions.
XBRUSD M15 — ATR Alternative
Demonstrated positive modeled expectancy in certain periods, with weaker results during other market conditions.
These findings remain provisional.
The historical evaluation involved modeled entries and exits, overlapping shadow trades, configurable execution costs and research decisions informed by earlier historical results.
Consequently, the reported historical partitions cannot automatically be considered pristine, independent out-of-sample evidence.
Odin's current classification is:
Forward-testing research system — not production-certified.
The next validation stage requires independently reconstructed trade records, realistic execution assumptions and new forward observations.
14. Why Odin Uses an Institutional-Style Framework
Odin is described as institutional-style because it separates market analysis into distinct functions commonly found in systematic trading research:
Directional research: Establishing a higher-timeframe market hypothesis.
Market structure analysis: Evaluating structural continuation and opposition.
Auction analysis: Studying price acceptance and liquidity-related behavior.
Regime classification: Identifying changing market conditions.
Participation and momentum analysis: Evaluating supporting or contradictory evidence.
Risk qualification: Rejecting trades with unacceptable stop-loss or target geometry.
Observability: Providing transparent factor readings, permission status and decision reasons.
This architecture is more comprehensive than a standalone crossover indicator.
However, institutional-style architecture does not imply access to proprietary exchange data, institutional counterparties, superior execution infrastructure or a verified trading advantage.
15. Frequently Asked Questions
Does Odin repaint?
Its higher-timeframe framework is designed around completed Daily and H1 observations, and trade signals require confirmed chart bars. Users should still verify realtime behavior and alert consistency on their specific instruments and TradingView data feeds.
Can Odin execute trades automatically?
No. The TradingView indicator produces analytical signals and alerts. Execution requires separate automation infrastructure.
Can I change the six-factor weights?
Yes. The default weighting is 30/20/15/15/10/10. Changing the weights alters the research model and should be treated as a new configuration requiring validation.
Which confirmation mode is best?
There is no universally established best mode. Informational is the default research baseline. The stricter modes can reduce trade frequency but have not been independently established as superior across all instruments.
Which stop policy should I use?
Strict Structure is the default. EURUSD Strict Structure and XBRUSD ATR Alternative are the two principal historical research candidates. Neither is certified for live deployment.
Is Odin suitable for beginners?
The interface can assist developing traders, but its concepts—market structure, volatility-normalized risk, regime transitions and multi-timeframe analysis—require an understanding of trading risk.
16. Risk Disclosure
Odin AI is a technical analysis and trading research product developed by Plazo Sullivan Roche Capital.
It is provided for informational, educational and research purposes.
Trading foreign exchange, contracts for difference, commodities, indices, cryptocurrencies and leveraged financial products involves substantial risk of loss.
Historical simulations, shadow-trade results and forward-testing observations do not guarantee future performance.
Market gaps, spreads, slippage, liquidity limitations, broker execution policies and data-feed differences can cause actual trading outcomes to differ materially from displayed signals or modeled results.
Users remain responsible for independent analysis, position sizing, risk management and compliance with applicable laws and trading regulations.
Odin AI provides structured trading intelligence. It does not eliminate uncertainty.
ODIN AI v2.6.1Plazo Sullivan Roche CapitalMulti-Timeframe Intelligence. Market Context. Disciplined Trade Decisions.
Comments