Brain Libraries Suite: WorldQuant BRAIN Alpha Discovery & Meta-Synthesis

The Brain Libraries Suite (brain-libraries) is a published collection of 7 pip-installable Python libraries designed to decouple and scale quantitative research on the WorldQuant BRAIN platform. The suite provides mathematical type contracts, dual-mode persistence, an extensible 10-axis orthogonal decorrelation engine with AST motif mining and asymmetric veto memory, specialized domain engines (Options surfaces, PEAD sentiment, systematic risk models), and an apex meta-synthesis CLI with Hierarchical Risk Parity (HRP).





---

Suite Architecture & Dependency Graph

The suite cleanly separates domain modeling, execution state, and alpha transformation:

brain-libraries/
├── brain_core/          # Shared type contracts, AST deduplication, multi-tier LLM router
├── brain_store/         # Dual-mode relational persistence (PostgreSQL + flat-file JSON/CSV)
├── brain_decorrelator/  # 10-axis orthogonal decorrelation, AST-diff engine, Veto memory
├── brain_options/       # IV surface, term structure, VRP, and straddle breakeven archetypes
├── brain_sentiment/     # SUE surprises, net revision breadth acceleration, PEAD drift
├── brain_risk_model/    # Systematic factor premia, Betting-Against-Beta (BAB), idiosyncratic vol
├── brain_synthesis/     # Apex synthesis engine, Hierarchical Risk Parity (HRP), dynamic regime tuning
└── articles/            # 5-part quantitative econometric research masterclass

graph TD
    BC[brain-alpha-core] --> BS[brain-store]
    BC --> BD[brain-decorrelator]
    BC --> BO[brain-options]
    BC --> BSe[brain-sentiment]
    BC --> BRM[brain-risk-model]
    
    BS --> BO
    BS --> BSe
    BS --> BRM
    
    BD --> BO
    BD --> BSe
    BD --> BRM
    
    BO --> BSyn[brain-synthesis]
    BSe --> BSyn
    BRM --> BSyn

---

Published Package Index

All 7 libraries are independently installable from PyPI and maintain rigorous test suites (86 passing unit and integration tests):

| Library | PyPI Package | Version | Role | Key Capabilities |

|---|---|:---:|---|---|

| brain_core | `brain-alpha-core` | 0.1.0 | Core Foundation | Universal dataclasses, commutative AST deduplication, multi-LLM fallback router |

| brain_store | `brain-store` | 0.1.0 | Persistence | Connection-pooled PostgreSQL (psycopg_pool) with offline flat-file fallback |

| brain_decorrelator | `brain-decorrelator` | 0.2.0 | Correlation Gate | 10-axis orthogonal transforms, AST-diff motif mining (*AlphaMemo*), asymmetric veto memory |

| brain_options | `brain-options` | 0.1.0 | Options Domain | Implied volatility surfaces, skew, term structure, and straddle breakeven models |

| brain_sentiment | `brain-sentiment` | 0.1.0 | Sentiment Domain | Standardized Unexpected Earnings (SUE), revision diffusion, and PEAD half-life |

| brain_risk_model | `brain-risk-model` | 0.1.0 | Risk Premia | Betting Against Beta (BAB), idiosyncratic volatility, and factor crowding deflation |

| brain_synthesis | `brain-synthesis` | 0.2.0 | Apex Meta-Synthesis | 8 Golden Apex archetypes, Hierarchical Risk Parity (HRP), dynamic regime tuning |

---

10-Axis Orthogonal Decorrelation Engine (brain-decorrelator)

WorldQuant BRAIN strictly enforces a correlation gate ($|\rho| < 0.70$) against all previously accepted alphas on the platform. When mining signals within a single dataset family, variations naturally cluster with correlations $\rho \in [0.75, 0.95]$.

brain-decorrelator systematically transforms alpha candidates across 10 orthogonal mathematical axes to break platform correlation gates while retaining $\ge 85\text{–}95\%$ of predictive Sharpe ratio:

1. Velocity Axis: Temporal phase shifting via ts_decay_linear(ts_delta(expr, d1), d2) ($\rho \to 0.35\text{--}0.52$).

2. Neutralization Axis: Group neutralization removing sector and sub-industry common modes ($\rho \to 0.42\text{--}0.61$).

3. Volatility Axis: Inverse realized return variance scaling to condition signals against market regimes ($\rho \to 0.48\text{--}0.65$).

4. Rank Axis: Non-linear power transformations emphasizing signal tails ($\rho \to 0.55\text{--}0.68$).

5. Volume Axis: Marginal order-flow liquidity weighting scaling trade size to participation capacity ($\rho \to 0.38\text{--}0.58$).

6. Calendar Axis: Intra-week and turn-of-month institutional rebalancing modulation ($\rho \to 0.60\text{--}0.68$).

7. Sector Demean Axis: Cross-sectional demean operator stripping broad sector beta drift.

8. Volatility Regime Scaling Axis: Normalizing signals dynamically during volatility expansions.

9. Gram-Schmidt Beta Deflation: Deflating alpha expressions against the rolling market benchmark.

10. Decay Dilation Axis: Extending signal half-life to reduce turnover below $4.5\%$ and elevate margin above $35\text{ bps}$.

AST-Diff Engine & Asymmetric Veto Memory

• AST Structural Diffing: Parses WorldQuant Fast Expressions into structured syntax trees (ASTNode), computing motif signatures to identify structural overlaps based on *AlphaMemo* (Yu et al., 2026).
• Asymmetric Veto Memory: Caches platform correlation-gate collisions and automatically suppresses repetitive dead-end expression motifs before submitting to the platform simulator.

---

Tri-Factor Apex Meta-Synthesis (brain-synthesis)

brain-synthesis fuses disparate domain signals across Options surfaces, Analyst Sentiment, and Systematic Risk into high-capacity composite meta-alphas:

• Hierarchical Risk Parity (HRP): Implements Marcos López de Prado's tree-clustering allocation. Computes correlation distance metric space $D_{i,j} = \sqrt{0.5 \times (1 - \rho_{i,j})}$, performs single-linkage agglomerative clustering, and allocates risk weights via recursive bisection without inverting ill-conditioned covariance matrices.
• Dynamic Macro Regime Tuning: Incorporates the AQuA framework (Guo et al., 2026) to detect HIGH_VOLATILITY, EARNINGS_SEASON, and TRENDING regimes, dynamically recalibrating domain weights.
• 8 Golden Apex Production Alphas: Pre-configured mathematical formulations yielding over 1,600+ platform points across TOP3000, TOP2000, TOP1000, TOP500, and TOPSP500 universes.

---

Installation & CLI Quickstart

# Install apex meta-synthesis suite and global 'brain' CLI
pip install brain-synthesis

# Or install individual libraries as needed
pip install brain-decorrelator brain-options brain-sentiment brain-risk-model

# Generate cross-domain candidates
brain generate --domain all --count 5

# Apply 10-axis decorrelation to clear correlation gate
brain decorrelate "rank(implied_volatility_call_30)" --base-sharpe 1.55

# Inspect storage health and alpha repository
brain store

Python API Example

from brain_synthesis import SynthesisEngine
from brain_decorrelator import DecorrelationEngine

engine = SynthesisEngine()
golden_alphas = engine.get_golden_candidates()

decorrelator = DecorrelationEngine()
results = decorrelator.decorrelate_all("rank(snt1_d1_netearningsrevision)")
for r in results:
    print(f"Axis: {r.axis_name} | Expr: {r.transformed_expression}")

---

5-Part Quantitative Masterclass Series

The suite is accompanied by a 5-part technical masterclass series detailing the underlying econometric and statistical mechanics:

1. Beating the Correlation Gatekeeper ($|\rho| < 0.70$): 10 orthogonal axes, AST motifs, and asymmetric veto memory.

2. Fixing `LOW_SUB_UNIVERSE_SHARPE`: Sub-industry vector neutralization eliminating unintended factor bets.

3. Taming the Turnover Trap ($< 12\%$): PEAD drift half-life matching and SUE surprise normalization.

4. Escaping the Price-Volume Trap: High-value options implied volatility surface extraction.

5. The 1,500+ Point Blueprint: Cross-domain signal fusion via Hierarchical Risk Parity (HRP).