FAMADOR

Section 01·

Executive Summary & Core Axioms

Modern portfolio management platforms frequently suffer from an institutional dilemma: either they obscure the allocation process inside proprietary black-box machine-learning models, or they rely on rudimentary heuristic asset allocations that neglect real-world covariance noise, liquidity constraints, and tail-risk dynamics.

FAMADOR resolves this through a mathematically explicit, closed-form quantitative architecture designed under three non-negotiable axioms:

Axiom 1: Determinism

Given identical market histories and investor constraints, the optimization manifold yields an exact, reproducible global minimum.

Axiom 2: Conditioned Stability

Sample covariance matrices must be asymptotically regularized to eliminate numerical singular matrices during high-volatility shifts.

Axiom 3: Sovereign Provenance

Zero automated trade execution. All proposals require explicit cryptographic acceptance logged to an append-only audit trail.

Section 02·

Security Universe & Liquidity Sieve

The FAMADOR master catalog monitors over 13,274 multi-asset instruments spanning equities, ETFs, ADRs, REITs, and sovereign fixed income. To safeguard optimization matrices from survivorship bias, illiquidity distortions, or missing observation vectors, the pipeline enforces a multi-tier filtration sieve:

Formal Universe Partition Function

U_{eligible} = \{ s \in \mathcal{C} \;\vert\; \mathcal{T}_{obs}(s) \ge 36 \land \text{ADTV}_{30}(s) \ge \tau_{min} \land \text{Tradable}(s) = 1 \}

Where:
• \mathcal{C}: Comprehensive catalog master
• \mathcal{T}_{obs}: Discrete monthly closing observations
• \text{ADTV}_{30}: 30-day Average Daily Trading Volume
• \tau_{min}: Institutional liquidity hurdle preventing execution slippage

Instruments failing the 36-month observation threshold remain searchable for research, benchmark analysis, and watchlist surveillance, but are programmatically excluded from the quadratic covariance solver to avoid corrupting matrix rank.

Section 03·

Ledoit-Wolf Covariance Regularization

A standard sample covariance matrix S calculated over p securities with T historical observations becomes ill-conditioned when p / T \to c > 0. Inverting an unregularized matrix creates severe weight instability and allocates extreme positive and negative corner positions.

FAMADOR implements the analytical optimal shrinkage framework developed by Olivier Ledoit and Michael Wolf (2004). The sample covariance matrix is asymptotically regularized toward a structured single-index target matrix F:

Shrunk Covariance Estimator

\boldsymbol{\Sigma}_{shrunk} = (1 - \delta^*) \mathbf{S} + \delta^* \mathbf{F}

Where the optimal asymptotic shrinkage intensity \delta^* \in [0, 1] is computed analytically:

\delta^* = \max \left( 0, \; \min \left( 1, \; \frac{\kappa}{T} \right) \right), \quad \kappa = \frac{\pi - \rho}{\gamma}

Where \pi measures asymptotic variance of sample covariance elements, \rho measures asymptotic covariance between sample and target elements, and \gamma measures misspecification of the structured target.

Mathematical Invariant: Positive Definiteness

Because the target matrix \mathbf{F} is constructed with strictly positive diagonal elements and uniform non-negative correlations, \boldsymbol{\Sigma}_{shrunk} is guaranteed strictly positive-definite:

\mathbf{z}^T \boldsymbol{\Sigma}_{shrunk} \mathbf{z} > 0, \quad \forall \; \mathbf{z} \ne \mathbf{0}

This eliminates numerical singular matrix errors in the solver during all market volatility regimes.

Section 04·

Constrained Active-Set Quadratic Solver

Portfolio allocation in FAMADOR is formulated as a strictly convex Quadratic Programming (QP) optimization problem. Using an active-set dual algorithm (Goldfarb & Idnani, 1983), the engine minimizes portfolio variance subject to investor return requirements and statutory allocation boundaries:

Primal Optimization Problem

\min_{\mathbf{w}} \; \left( \frac{1}{2} \mathbf{w}^T \boldsymbol{\Sigma}_{shrunk} \mathbf{w} - \lambda \mathbf{w}^T \boldsymbol{\mu} \right)

Subject to linear equality and inequality constraints:

1. Full Investment: \sum_{i=1}^{N} w_i = 1

2. Long-Only Mandate: w_i \ge 0, \quad \forall i

3. Asset-Class Caps: w_{min, k} \le \sum_{i \in \text{Class}_k} w_i \le w_{max, k}

4. Single-Asset Limit: w_i \le c_{max} \quad (\text{e.g.,} \; 15\%)

The parameter \lambda \ge 0 reflects the investor’s risk-tolerance coefficient calibrated from their suitability questionnaire score. The solver satisfies the first-order Karush-Kuhn-Tucker (KKT) optimality conditions:

\boldsymbol{\Sigma}_{shrunk} \mathbf{w}^* - \lambda \boldsymbol{\mu} + \mathbf{A}^T \boldsymbol{\nu}^* + \mathbf{C}^T \boldsymbol{\mu}_{kkt}^* = \mathbf{0}

Section 05·

Dynamic CAPM & Multi-Stage DCF

Expected asset returns \boldsymbol{\mu} are synthesized through a blended fundamental-quantitative architecture combining the Capital Asset Pricing Model (Sharpe, 1964) with two-stage Discounted Cash Flow valuation:

Cost of Capital (CAPM + WACC)

K_e = R_f + \beta_i (R_m - R_f)

\text{WACC} = \frac{E}{V} K_e + \frac{D}{V} K_d (1 - T_c)

Anchored dynamically by the 10-Year US Treasury yield (^TNX) as the risk-free benchmark R_f.

Intrinsic Two-Stage DCF

V_0 = \sum_{t=1}^{N} \frac{\text{FCF}_t}{(1 + \text{WACC})^t} + \frac{\text{TV}}{(1 + \text{WACC})^N}

\text{TV} = \frac{\text{FCF}_N (1 + g)}{\text{WACC} - g}

Projects discrete free cash flows over N = 5 years converging to long-term economic growth g.

Section 06·

Intraday Session Curve & Execution Flow

For institutional execution timing and tactical rebalancing, FAMADOR tracks tick-level microstructure indicators to mitigate liquidity impact:

Volume-Weighted Average Price (VWAP) & Variance Bands

\text{VWAP} = \frac{\sum_{i} P_i \cdot V_i}{\sum_{i} V_i}, \quad \text{Band}_k = \text{VWAP} \pm k \cdot \sigma_{vw} \; (k \in \{1, 2, 3\})

Dynamic statistical dispersion envelopes identify mean-reverting execution boundaries across session profiles.

Cumulative Volume Delta (CVD) Order Flow

\text{CVD}_t = \sum_{\tau=1}^{t} (V_{buy,\tau} - V_{sell,\tau})

Isolates buyer vs seller initiative at the bid-ask spread to uncover institutional absorption and price-delta divergences.

Section 07·

Explainability & SHA-256 Audit Chain

Unlike traditional robo-advisors that present static portfolio scores, FAMADOR mathematically decomposes every recommendation into additive feature contributions using SHAP attribution (Lundberg & Lee, 2017):

Shapley Additive Feature Attribution

\phi_i(v) = \sum_{S \subseteq N \setminus \{i\}} \frac{|S|! (|N| - |S| - 1)!}{|N|!} \left[ v(S \cup \{i\}) - v(S) \right]

Proves exactly how your investment horizon, drawdown tolerance, and cash reserves shifted the final asset class weights.

Tamper-Evident SHA-256 Governance Trail

Every portfolio draft, risk scoring event, model proposal, and user acceptance is hashed and anchored into a sequential, append-only cryptographic log on the server:

\text{BlockHash}_k = \text{SHA256}(\text{BlockHash}_{k-1} \;\parallel\; \text{AccountId} \;\parallel\; \text{Event} \;\parallel\; \text{PayloadJSON} \;\parallel\; \text{Timestamp})

Guarantees that no model recommendation or account history can be silently altered post-facto.

Section 08·

Peer-Reviewed References & Foundation Literature

Markowitz, H. (1952). “Portfolio Selection.” The Journal of Finance, 7(1), 77–91.
Ledoit, O., & Wolf, M. (2004). “A well-conditioned estimator for large-dimensional covariance matrices.” Journal of Multivariate Analysis, 88(2), 365–411.
Goldfarb, D., & Idnani, A. (1983). “A numerically stable dual method for solving strictly convex quadratic programs.” Mathematical Programming, 27(1), 1–33.
Sharpe, W. F. (1964). “Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk.” The Journal of Finance, 19(3), 425–442.
Lundberg, S. M., & Lee, S.-I. (2017). “A Unified Approach to Interpreting Model Predictions.” Advances in Neural Information Processing Systems (NeurIPS 30), 4765–4774.

FAMADOR Quantitative Engine · Cryptographically Verified