GENEVA ALGORITHMIC RISK ADVISORY (GARA)

Independent Quantitative Due Diligence & Epistemic Audit · Geneva, Switzerland

STRICTLY CONFIDENTIAL · SOVEREIGN SFO

AUDIT REF: GARA-2026-CH-0842

DATE: SEPTEMBER 2026

GENEVA NATIONS BUSINESS CENTER

Algorithmic Epistemic Concentration & Fragility Diagnostic

Evaluating Structural Model Convergence, Order-Book Vulnerabilities, and the Salih Index (SIAD)

Audited Portfolio: SFO-GENEVA-ALPHA · Audited Capital Allocation: USD 500,000,000

⚠️ Executive Systemic Verdict

CRITICAL RISK ADVISORY: Independent quantitative due diligence conducted by GARA reveals that despite legal diversification across 5 tier-one quantitative hedge funds, the portfolio suffers from acute Algorithmic Cognitive Homogeneity. Under market stress, 82% of invested capital is governed by correlated exit heuristics and identical open-weight transformer embeddings. During an unexpected liquidity shock, this cognitive convergence precipitates Order-Book Evaporation, yielding an estimated self-induced slippage and capital drawdown of up to 14.8% within 420 milliseconds.

Salih Index Score (SIAD)

0.31

Critical Vulnerability Zone (< 0.40)

Audited Capital (AUM)

$500M

5 External Quantitative Desks

Model Lineage Overlap

80%

Shared Transformer Embedding Lineage

Weighted Gross Leverage

2.88x

Aggregate Gross Notional: $1.44B USD

The Salih Index for Algorithmic Diversity (SIAD) — Resilience Gauge

Portfolio Diagnostic Score: 0.312 / 1.000 (Acute Epistemic Deficit)

The Salih Index (SIAD) evaluates architectural entropy, data oracle cross-correlation, temporal horizons, and exit heuristic independence. A score below 0.40 establishes that the portfolio functions as a systemic shock amplifier rather than an insulated shock absorber.

Portfolio: 0.31

Critical Epistemic Fragility (0.00 – 0.39) Moderate Diversity (0.40 – 0.69) Sovereign Algorithmic Resilience (0.70 – 1.00)

Manager Ingestion & Parameter Forensic Breakdown

Allocation Code	Strategy Classification	AUM ($M)	Gross Lev.	Model Lineage & Backbone	Data Oracle Feed	Homogeneity Risk
MGR-01 (Zurich Alpha)	Statistical Arbitrage	$150M	3.5x	Temporal Transformer / FinGPT	Refinitiv Tick History	Critical (0.24)
MGR-02 (London Quant)	Trend Momentum & CTA	$100M	2.2x	Multi-Head Attention LSTM	Bloomberg B-PIPE	Moderate (0.48)
MGR-03 (New York Macro)	Systematic FX Arbitrage	$100M	4.0x	BERT Financial Fine-tuned	Refinitiv Real-Time	Critical (0.19)
MGR-04 (Singapore Equity)	Equity Market Neutral	$75M	1.8x	Gradient Boosted + Transformer	Bloomberg B-PIPE	Moderate (0.52)
MGR-05 (Geneva Multi-Strat)	Cross-Asset Volatility	$75M	2.5x	Deep RL + Foundation Model	S&P Global Alternative	Resilient (0.68)

The TAFC Six-Pillar Forensic Assessment

TAFC Audit Pillar	Weight	Pillar Score	GARA Forensic Quantitative Diagnostic
1. Data Provenance & Oracles	20%	28 / 100	80% of audited assets rely on two centralized market feeds (Bloomberg & Refinitiv), establishing a catastrophic single point of failure during oracle anomaly events.
2. Model Lineage & Backbones	25%	24 / 100	Four of five managers rely on shared open-weight transformer embeddings, guaranteeing identical semantic interpretation of macroeconomic and earnings releases.
3. Cloud Topology & Latency	15%	42 / 100	70% of execution engines are co-located in AWS US-East and Equinix LD4, generating cross-manager network bottlenecks during volatile market spikes.
4. Temporal Horizon Skewness	15%	35 / 100	65% of capital ($325M) competes within sub-second execution horizons, creating an absence of stabilizing patient capital during flash sell-offs.
5. Reflexive Liquidity & Exits	15%	18 / 100	Critical Hazard: Synchronized volatility trailing stop-losses calibrated to historical 20-day ATR, ensuring simultaneous market sell orders collide during shocks.
6. Human Governance Override	10%	40 / 100	Average human risk committee response time is 15 minutes—45,000 times slower than the algorithmic trading loop, rendering manual oversight structurally illusory.

Reconstructed Stress Simulation: The Order-Book Evaporation Cascade

T + 0.000s [Anomaly Dissemination]: Unexpected macro announcement paired with a 3.5% sudden drop in market depth; shared data oracles disseminate distorted quote spikes across all desks in 12ms.

T + 0.120s [Model Convergence]: Transformer pipelines across 4 managers output concurrent de-leveraging recommendations to curtail balance-sheet exposure simultaneously.

T + 0.280s [Order-Book Evaporation]: Bid-side liquidity in core instruments collapses by 84% as competing market makers and internal algorithms pull passive bids concurrently.

T + 0.420s [Cascading Drawdown]: Portfolio absorbs extreme execution slippage exceeding mandate risk limits by 610%, delivering a net liquidating equity drawdown of -14.8% in under 1 second.

Strategic Remediation & Algorithmic Decoupling Roadmap

Phase 1: Immediate De-Risking (14 Days)

Mandate MGR-01 and MGR-03 to decouple identical ATR trailing stops and install asynchronous execution bands.
Impose a temporary 2.0x gross leverage ceiling on transformer-correlated strategies.

Phase 2: Structural Diversification (30–60 Days)

Reallocate at least 25% of quantitative capital to non-transformer, evolutionary or Bayesian mechanistic algorithms.
Integrate decentralized, non-correlated alternative data oracles for commodities and FX desks.

Phase 3: Sovereign SIAD Alignment (90 Days)

Elevate aggregate portfolio resilience to target: SIAD ≥ 0.72.
Schedule follow-up quantitative verification audit with GARA Geneva for Q4 2026.