Strategy Philosophy¶
Core Principles¶
The V1 strategy is an ICT-based interpretation layer that reads four TradingView Pine indicators through the MCP protocol and produces structured decision-support flags. It does not trade. It does not override. It interprets.
1. ICT Foundation¶
The strategy is built on Inner Circle Trader concepts:
- Liquidity hunts: Institutions sweep resting orders above/below session ranges before reversing toward the true draw on liquidity.
- Order blocks (OB): Institutional footprints left by aggressive displacement candles. Price returns to these zones before continuing.
- Fair value gaps (FVG): Three-candle imbalances that act as magnets and entry refinement zones within order blocks.
- SMT divergences: When correlated instruments (ES/NQ) make structural highs or lows that diverge, it signals institutional distribution or accumulation.
- Session timing: ICT macro windows (e.g., 09:50--10:10, 10:50--11:10 ET) concentrate institutional activity. Entries outside these windows carry lower conviction.
2. TradingView-First Architecture¶
All market analysis is performed by four Pine Script indicators running on TradingView charts. The strategy layer never recalculates what the indicators already compute.
| Indicator | Responsibility |
|---|---|
| Bias AI | Daily/weekly directional bias, draw-on-liquidity targets |
| Session Hunt AI | Session range, hunt direction, sweep state, macro windows |
| OB AI | Order block zones, FVG levels, ATR stops, position sizing |
| SMT Divergences | Inter-market structural confirmation |
The MCP protocol reads indicator output (labels, lines, tables, boxes) and the strategy interprets those readings into structured flags. If an indicator does not signal something, the strategy does not infer it.
3. Decision Support, Not Automation¶
V1 is a decision-support system. It produces a StrategyFlags JSON object that a human trader reviews before acting. The system:
- Does: Identify valid setups, compute confluence scores, calculate risk/reward, flag no-trade conditions.
- Does not: Place orders, manage positions, or make autonomous trading decisions.
This constraint is intentional. Automated execution is a Phase 3 concern. V1 must prove the interpretation logic is correct before any automation is built on top of it.
4. One Directional Trade Per Session¶
The ICT hunt model identifies a single draw-on-liquidity target per session. Once a hunt direction is locked (one side of the session range is swept), the strategy commits to that direction until invalidation or target completion.
There is no flipping. If HUNT_HIGH locks (high swept), the draw is on the low side. If the setup invalidates, the result is no-trade, not a reversal.
5. Indicators Are the Source of Truth¶
The strategy layer is a pure consumer of indicator output. It applies conditional logic to determine whether a setup exists, but it never:
- Overrides an indicator reading with its own calculation.
- Applies filters that contradict what an indicator reports.
- Adjusts price levels produced by the indicators.
If Bias AI says neutral, the strategy says no-trade. If OB AI marks an order block as spent, the strategy treats it as spent. The indicators are authoritative.
6. Documentation-First Development¶
Every rule, condition, and threshold is specified in documentation before any code is written. This ensures:
- The strategy logic is reviewable by humans who do not read TypeScript.
- Changes to rules require a documentation update first, then a code change.
- Edge cases are identified during specification, not during debugging.
- The mapping from ICT concept to implementation is explicit and traceable.
Design Constraints¶
| Constraint | Rationale |
|---|---|
| No direct market data feeds | V1 reads everything through TradingView indicators via MCP |
| No order placement | V1 is decision support only |
| No indicator modification | Strategy interprets, never patches indicator code |
| No multi-session memory | Each scan is stateless; persistence is a Phase 3 concern |
| Minimum 2:1 risk-reward | Sub-2:1 setups are filtered as no-trade regardless of confluence |
What Success Looks Like¶
A successful V1 produces StrategyFlags that a skilled ICT trader would look at and say: "Yes, that is the same setup I see on the chart." The interpretation layer adds structure and speed, not judgment.