multi-strategy-orchestration

> **Skill:** Multi Strategy Orchestration | **Domain:** trading | **Category:** meta-intelligence | **Level:** expert > **Tags:** `trading`, `orchestration`, `strategy`, `pipeline`, `risk`, `meta` # Multi-Strategy Orchestration > The command layer: regime detection → strategy selection → allocation → **this file** → execution > For strategy weights → see `portfolio-optimization.md` > For execution mechanics → see `trading-brain.md` ## Orchestration Pipeline ``` ┌─────────────────────────────────────────────────────────┐ │ ORCHESTRATION PIPELINE │ ├─────────────────────────────────────────────────────────┤ │ │ │ 1. REGIME DETECTION │ │ market-regime-classifier.md │ │ → Output: TRENDING | RANGING | TRANSITIONING | VOL │ │ │ │ 2. STRATEGY SELECTION │ │ strategy-selection.md │ │ → Output: Active strategy set for current regime │ │ │ │ 3. PORTFOLIO ALLOCATION │ │ portfolio-optimization.md │ │ → Output: Weight per strategy, max positions │ │ │ │ 4. SIGNAL GENERATION (parallel) │ │ Each active strategy scans independently: │ │ ├── ICT MSS+FVG scanner │ │ ├── ORB detector │ │ ├── S&D zone fade scanner │ │ └── Liquidity trap scanner │ │ → Output: Raw signals with confluence scores │ │ │ │ 5. SIGNAL ARBITRATION ← THIS FILE │ │ Resolve conflicts, rank, filter │ │ → Output: Approved trade list │ │ │ │ 6. RISK CHECK │ │ risk-and-portfolio.md │ │ → Output: Position sizes, heat check │ │ │ │ 7. EXECUTION │ │ execute_trades.py │ │ → Output: Open positions │ │ │ │ 8. MONITORING & REBALANCE (loop) │ │ → Regime change? → Back to step 1 │ │ → Drawdown threshold? → Scale down │ │ → Session change? → Adjust weights │ │ │ └─────────────────────────────────────────────────────────┘ ``` ## Signal Arbitration ### When Strategies Agree (Confluence) ``` Multiple strategies signal SAME direction on SAME instrument: → STRONGEST setup — increase conviction Scoring: 1 strategy agrees: Base signal (normal size) 2 strategies agree: +25% confidence (can increase to 1.25% risk) 3+ strategies agree: +50% confidence (can increase to 1.5% risk) Example: ICT MSS+FVG signals LONG XAUUSDm (confluence: 7/10) ORB signals LONG XAUUSDm (confluence: 5/10) S&D zone supports LONG XAUUSDm (zone tested) → 3 strategies agree → 1.5% risk, highest priority entry ``` ### When Strategies Conflict (Opposing Signals) ``` Two strategies signal OPPOSITE direction on SAME instrument: Resolution rules (in priority order): 1. HTF WINS: Strategy aligned with daily/4H trend takes priority → If daily bullish + ICT signals long + ORB signals short → Take the ICT long, ignore ORB short 2. HIGHER CONFLUENCE WINS: Compare confluence scores → If S&D fade: 8/10 vs ICT trend: 5/10 → Take S&D fade (but at half size due to conflict) 3. HIGHER TIMEFRAME STRATEGY WINS: 4H signal > 1H signal > M5 signal → If 4H OB signals long + M5 ORB signals short → Take 4H OB long 4. IF TRULY EQUAL: NO TRADE on that instrument → Conflicting signals = market indecision → Move to next instrument with clearer signal NEVER take both sides of a conflict simultaneously. ``` ### Signal Priority Ranking ``` When multiple valid signals across different instruments: Priority 1: Multi-strategy confluence (3+ strategies agree) Priority 2: HTF-aligned + high confluence (7+/10) Priority 3: Killzone-aligned signals (correct session) Priority 4: Single strategy, high confluence (7+/10) Priority 5: Single strategy, moderate confluence (4-6/10) Execute in priority order until: - Max position count reached (from allocation table) - Portfolio heat limit reached (from risk rules) - No more valid signals above confluence threshold ``` ## Capital Allocation Per Strategy ### Equity Bucketing Model ``` Total trading equity: 100% ├── Strategy A (ICT trend): 40% of equity → max 2 positions × 1% ├── Strategy B (ORB): 20% of equity → max 1 position × 1% ├── Strategy C (S&D fade): 20% of equity → max 2 positions × 0.5% ├── Strategy D (liquidity trap):20% of equity → max 1 position × 0.5% └── Reserve: 0% (fully allocated when regime clear) Per-strategy equity is a SOFT LIMIT: - Strategy can exceed allocation by 25% if confluence > 8/10 - Strategy must reduce to 0% if regime changes away from it - Reserve builds to 20-40% during VOLATILE regime (cash is a position) ``` ### Kelly-Adjusted Strategy Sizing ``` For each strategy, calculate optimal fraction: Kelly % = (Win Rate × Avg Win) - (Loss Rate × Avg Loss) ─────────────────────────────────────────────── Avg Win Half-Kelly (conservative): Strategy | WR | Avg W | Avg L | Kelly | Half-K | Max Risk ────────────────┼──────┼───────┼───────┼───────┼────────┼───────── ICT MSS+FVG | 55% | 2.5R | 1.0R | 19.5% | 9.75% | 1.0% ORB Break/Rest | 60% | 2.0R | 1.0R | 20.0% | 10.0% | 1.0% S&D Zone Fade | 65% | 1.5R | 1.0R | 21.7% | 10.8% | 0.75% Liquidity Trap | 50% | 3.0R | 1.0R | 16.7% | 8.3% | 0.75% Cap at 1% regardless of Kelly — protect against overfitting. ``` ## Risk Aggregation ### Portfolio Heat Dashboard ``` ┌─────────────────────────────────────────────────┐ │ PORTFOLIO HEAT CHECK │ ├─────────────────────────────────────────────────┤ │ Strategy │ Positions │ Risk Each │ Heat │ │─────────────────┼───────────┼───────────┼────────│ │ ICT MSS+FVG │ 2 │ 1.0% │ 2.0% │ │ ORB NY │ 1 │ 1.0% │ 1.0% │ │ S&D Fade │ 1 │ 0.5% │ 0.5% │ │ Liquidity Trap │ 0 │ — │ 0.0% │ ├─────────────────┼───────────┼───────────┼────────│ │ TOTAL │ 4 │ │ 3.5% │ ├─────────────────────────────────────────────────┤ │ Max heat: 6% (trending) │ Status: ✅ UNDER LIMIT │ │ Correlated heat: 2.0% (2 gold positions) │ │ Corr limit: 3% │ Status: ✅ UNDER LIMIT │ └─────────────────────────────────────────────────┘ ``` ### Heat Limits by Regime ``` Regime | Max Total Heat | Max Per Strategy | Max Correlated ───────────────┼────────────────┼──────────────────┼─────────────── TRENDING | 6% | 3% | 3% RANGING | 4% | 2% | 2% TRANSITIONING | 3% | 1.5% | 1.5% VOLATILE | 1% | 1% | 1% ``` ## Regime Switchover Logic ### Detection → Action Timeline ``` Regime change detected (e.g., TRENDING → RANGING): T+0 min: DETECT — VP shape changed, ATR dropped below threshold T+0 min: FREEZE — No new entries from old regime strategies T+5 min: ASSESS — Are existing positions still valid? ├── Position WITH new regime → KEEP (adjust SL/TP) └── Position AGAINST new regime → CLOSE or TIGHTEN T+10 min: SWITCH — New strategy allocation active T+15 min: SCAN — New regime strategies begin scanning for signals T+30 min: EXECUTE — First entries under new regime allocation Total switchover time: ~30 minutes During switchover: maximum 2% portfolio heat (protective) ``` ### Regime Confidence Scoring ``` Don't switch on weak signals. Require confirmation: Confidence Score (0-10): +2 VP shape confirms new regime +2 ATR confirms (above/below 20-day avg) +2 Price structure confirms (HH/HL or LH/LL established) +2 Cumulative delta confirms +1 Session timing supports (London open often triggers transitions) +1 News catalyst present (fundamental driver for regime change) Score 7+ → CONFIRMED regime change → full switchover Score 4-6 → POSSIBLE change → reduce allocation 50%, wait for confirmation Score <4 → NOISE → maintain current regime allocation ``` ## Conflict Resolution Matrix ``` Scenario | Resolution ──────────────────────────────────┼─────────────────────────────────── Same signal, same instrument | Combine → highest confluence entry Opposing signals, same instrument | HTF wins, or skip if truly equal Same signal, correlated pairs | Take strongest setup only Opposing signals, correlated pairs| Take neither — market unclear Strategy says trade, regime says no| Regime overrides strategy always Multiple instruments, limited heat| Rank by priority, fill top-down News event approaching | Freeze all signals until T+30 min Drawdown >4% active | Half all allocations immediately ``` ## Monitoring Loop ```python # Conceptual orchestration loop (runs every 5 min during session) def orchestration_loop(): # 1. Check regime regime = detect_regime() # TRENDING | RANGING | TRANSITIONING | VOLATILE # 2. Get active strategies for regime strategies = select_strategies(regime) # 3. Get allocation weights weights = get_allocation(regime, current_session(), current_drawdown()) # 4. Generate signals (parallel) signals = [] for strategy in strategies: sigs = strategy.scan(watched_symbols) signals.extend(sigs) # 5. Arbitrate conflicts approved = arbitrate_signals(signals, weights) # 6. Risk check approved = risk_filter(approved, current_positions(), heat_limits[regime]) # 7. Execute approved signals for signal in approved: execute_trade(signal) # 8. Monitor existing positions for position in current_positions(): if not valid_for_regime(position, regime): tighten_or_close(position) ``` ## Strategy Performance Tracking ``` Track per-strategy metrics to adjust future allocation: Weekly Review: Strategy | Trades | WR | Avg R | Total R | Sharpe ────────────────┼────────┼───────┼────────┼─────────┼─────── ICT MSS+FVG | 8 | 62% | +1.8R | +6.4R | 2.1 ORB | 5 | 60% | +1.5R | +3.0R | 1.8 S&D Fade | 6 | 67% | +1.2R | +2.4R | 1.6 Liquidity Trap | 3 | 33% | +2.5R | -0.5R | 0.3 Action: If strategy Sharpe < 1.0 for 2 consecutive weeks: → Reduce allocation by 50% → Review: is the strategy broken, or is regime wrong for it? → If regime-appropriate but still underperforming → bench for 1 week ``` --- ## Related Skills - [Portfolio Strategy Allocation](../portfolio-optimization.md) - [Strategy Selection](../strategy-selection.md) - [Market Regime Detection](../market-regime-classifier.md) - [Risk And Portfolio](../risk-and-portfolio.md) - [Trading Brain Orchestrator](../trading-brain.md)