Portfolio Analysis

Explore how strategies relate and what combining them means for portfolio characteristics.

What This Skill Does

Surfaces data to help understand portfolio dynamics:

• Health Check: One-call comprehensive portfolio assessment
• Correlation: How do strategies move relative to each other?
• Structure: Where does the portfolio have overlap or blind spots?
• Marginal Contribution: Which strategies help or hurt risk-adjusted returns?
• Similarity: Are any strategies redundant?
• What-If Scaling: How would changing allocations affect the portfolio?

Prerequisites

• TradeBlocks MCP server running
• Multiple strategy blocks or a multi-strategy block loaded
• Strategy profiles recommended for regime coverage and structure analysis

Process

Step 1: Portfolio Health Check

Start with portfolio_health_check for a comprehensive one-call assessment.

Key parameters:

• blockId: Block folder name
• correlationThreshold: Flag pairs above this (default: 0.5)
• tailDependenceThreshold: Flag tail pairs above this (default: 0.5)
• profitProbabilityThreshold: MC profit probability warning (default: 0.95)
• wfeThreshold: Walk-forward efficiency warning (default: -0.15)

Tool returns a layered report:

• Verdict: HEALTHY / MODERATE_CONCERNS / SIGNIFICANT_CONCERNS
• Grades: A-F across 9 dimensions:

• Flags: Specific warnings with details (high correlation pairs, tail dependence pairs, etc.)
• Key numbers: Sharpe, Sortino, max drawdown, avg correlation, avg tail dependence, MC stats, WFE

Present the verdict, flag any grades below B, and surface the specific warning details.

Step 2: Correlation Analysis

Use get_correlation_matrix for detailed pairwise correlation data.

Key parameters:

• blockId: Block folder name
• method: "kendall" (robust, rank-based, default), "spearman" (rank), "pearson" (linear)
• alignment: "shared" (only days both traded) or "zero-pad" (fill missing with 0)
• timePeriod: "daily", "weekly", or "monthly" aggregation
• normalization: "raw" (absolute P&L), "margin" (P&L/margin), "notional" (P&L/notional)
• minSamples: Minimum shared periods for valid calculation (default: 10)
• highlightThreshold: Flag pairs above this (default: 0.7)

Interpreting correlation values:

See references/correlation.md for why Kendall's tau is often more informative than Pearson for trading returns.

Step 3: Portfolio Structure Map

Use portfolio_structure_map to see a Vol_Regime x Trend_Direction matrix across all profiled strategies.

Key parameters:

• blockId: Optional — omit to aggregate across all blocks
• minTrades: Thin-data warning threshold (default: 10)

Tool returns:

• 18-cell matrix (6 Vol_Regimes x 3 Trend_Directions) with per-strategy stats
• Overlap detection: 2+ strategies active in the same cell
• Blind spots: Cells with zero trades across all strategies
• Thin-data warnings: Cells with fewer trades than threshold

This is the key tool for understanding portfolio construction — where the portfolio has coverage and where it doesn't.

Step 4: Marginal Contribution

Use marginal_contribution to see how each strategy affects portfolio risk-adjusted returns.

Key parameters:

• blockId: Block folder name
• targetStrategy: Calculate for specific strategy only (optional)
• topN: Number of top contributors to return (default: 5)

Tool returns:

• Baseline portfolio Sharpe and Sortino
• Per-strategy marginal Sharpe and Sortino impact
• Most beneficial and least beneficial strategies

Interpretation:

• Negative marginal Sharpe: Removing this strategy would LOWER the portfolio Sharpe — it's contributing positively
• Positive marginal Sharpe: Removing this strategy would RAISE the portfolio Sharpe — it may be hurting risk-adjusted returns
• Note: marginal contribution measures risk-adjusted impact, not absolute P&L. A profitable strategy can hurt Sharpe if it adds volatility.

Step 5: Regime Allocation Advisor

Use regime_allocation_advisor to cross-reference strategy profiles' expected regimes with actual performance.

Key parameters:

• blockId: Optional — omit to aggregate across all profiled strategies
• minTrades: Minimum trades per regime cell for reliable stats (default: 5)

Tool returns per strategy per regime:

• Whether the regime was expected (from profile) or unexpected
• Win rate, P&L, trade count in that regime
• Classifications:
  • thesis_aligned: Strategy performs as expected in its target regimes
  • thesis_violation: Strategy underperforms in regimes it should handle
  • hidden_edge: Strategy performs well in regimes not marked as expected

Surface any thesis violations (strategies failing where they shouldn't) and hidden edges (opportunities the profile doesn't capture).

Step 6: Strategy Similarity

Use strategy_similarity to detect potentially redundant strategies.

Key parameters:

• blockId: Block folder name
• correlationThreshold: Min correlation to flag (default: 0.7)
• tailDependenceThreshold: Min tail dependence to flag (default: 0.5)
• minSharedDays: Minimum shared trading days (default: 30)
• topN: Number of most similar pairs (default: 5)

Tool returns:

• Most similar strategy pairs ranked by combined similarity score
• Correlation, tail dependence, and trading day overlap for each pair
• Flags for strategies that may be adding risk without diversification benefit

If two strategies are highly correlated AND have high tail dependence, they'll likely draw down together — the "diversification" between them is illusory.

Step 7: What-If Scaling

Use what_if_scaling to explore allocation changes.

Key parameters:

• blockId: Block folder name
• strategyWeights: Weight per strategy, e.g., {"5/7 17D": 0.5, "Pickle RIC": 1.5}. Unspecified default to 1.0. Weight 0 = exclude.
• strategies: Multi-strategy mode — array with per-strategy block source and scale factor
• showUncapped: Also show results without maxContractsPerTrade ceiling

Tool returns:

• Before/after comparison of portfolio metrics
• Per-strategy breakdown at new weights
• Profile-aware: respects maxContractsPerTrade ceilings from profiles
• Flags when ignoreMarginReq is set

Common scenarios:

• "What if I removed this strategy?" → set its weight to 0
• "What if I doubled this allocation?" → set weight to 2.0
• "What if I halved everything except my best performer?" → adjust weights accordingly

Step 8: Present Findings

Synthesize the data:

Portfolio Health:

• Verdict: [HEALTHY / MODERATE_CONCERNS / SIGNIFICANT_CONCERNS]
• Dimensions needing attention: [any grades below B]
• Key flags: [specific warnings]

Correlation Findings:

• Average correlation across pairs: [value]
• Highest correlation pair: [pair] at [value]
• Lowest correlation pair: [pair] at [value]

Structure:

• Regime blind spots: [cells with no coverage]
• Overlap areas: [cells with 2+ strategies]
• Concentration: [by underlying, DTE, structure type]

Marginal Contribution:

• Most beneficial: [strategy] (marginal Sharpe: [value])
• Least beneficial: [strategy] (marginal Sharpe: [value])

What stands out:

• [Notable patterns]
• [Any redundant strategy pairs]
• [Thesis violations or hidden edges]

Present these as insights from the historical data. The user can decide what fits their risk tolerance and portfolio goals.

Interpretation References

• references/correlation.md - Understanding correlation methods
• references/diversification.md - Diversification concepts and tail risks

Related Skills

After portfolio analysis:

• /tradeblocks:compare - Deep comparison of specific strategy pairs or blocks
• /tradeblocks:risk - Position sizing and Kelly analysis
• /tradeblocks:health-check - Full metrics on any individual strategy

Notes

• Correlation is measured on aggregated returns, not trade-by-trade
• Past correlation patterns may not persist in future market conditions
• Tail correlation (crisis behavior) is often higher than normal correlation
• Low correlation doesn't guarantee protection - both can lose for different reasons
• Sample size matters - 10 shared data points is minimum, more is better
• Strategy profiles are required for regime coverage, structure map, and scaling alignment analysis