File size: 4,325 Bytes
5373c26 6483d0f 3e9fa58 6483d0f 3e9fa58 5373c26 3e9fa58 5373c26 6483d0f 3e9fa58 6483d0f 3e9fa58 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 | # SYNTHRA Glossary
This document provides concise definitions for essential WorldQuant BRAIN and SYNTHRA terms used throughout the codebase and documentation.
---
## 🌍 WorldQuant BRAIN Concepts
- **Alpha**: A mathematical expression or algorithm designed to predict price movements of assets and generate risk-adjusted returns.
- **ACE (Alpha Calculation Engine)**: The official WorldQuant backtesting and simulation platform accessed via the API.
- **Fast Expression**: A single-line mathematical formula using platform-defined operators (e.g., `ts_mean`, `rank`) to represent alpha logic.
- **Python Alpha**: An alpha strategy implemented as a multi-line Python function conforming to the official template, allowing custom logic and pandas operations.
- **Sharpe Ratio**: A metric representing the risk-adjusted return of an alpha, calculated as average return divided by volatility.
- **Fitness**: A composite platform metric evaluating alpha quality, combining Sharpe ratio, turnover, and margin.
- **Turnover**: The rate at which the alpha portfolio changes its asset holdings, representing the transaction volume.
- **Margin**: The ratio of expected book returns to total transaction volume.
- **Delay**: The time lag between the collection of data and the execution of trades based on that data (e.g., Delay 1, Delay 2).
- **Neutralization**: A statistical adjustment subtracting market, industry, or sub-industry averages from alpha values to isolate idiosyncratic returns.
- **Decay**: The rate at which an alpha's predictive power and returns degrade over time.
- **Coverage**: The percentage of assets in the target universe for which the alpha generates non-zero predictions.
---
## 🤖 SYNTHRA Architecture Concepts
- **Campaign (`CMP-XXXX`)**: The primary Research Program. It serves as the bounding context for all hypotheses, experiments, and learnings, restricted by specific assets, regions, and themes.
- **Hypothesis (`HYP-XXXX`)**: A structured document defining an economic rationale explaining why a specific relationship should yield predictability.
- **Experiment (`EXP-XXXX`)**: A single backtest execution mapping to a hypothesis, generating specific logs and performance metrics.
- **Research Asset (`AST-XXXX`)**: Any technical output produced during research (including Hypotheses, Expressions, Simulations, Visualizations, Correlation Reports, Notebooks, Parameter Sweeps, Summaries, Failure Analyses, Dataset Profiles).
- **Knowledge Record (`KNW-XXXX`)**: A record capturing research discoveries, observations, evidence, conclusions, and confidence scores.
- **Research Memory**: A persistence system caching historical campaign configurations, simulation logs, and performance metrics.
- **Knowledge Graph**: A conceptual representation linking datasets, mathematical operators, economic themes, and their empirical performance.
- **Experiment Graph**: A parent-child relational tree tracking the lineage of hypotheses, expression variations, and backtest results.
- **Portfolio Intelligence**: The layer that evaluates cross-strategy correlation, information ratio contribution, and overall portfolio diversification.
- **Planner**: The orchestration engine scheduling campaign tasks, managing dependencies, and routing messages.
- **Governor**: The highest system authority ("CEO") enforcing security, compliance boundaries, and resource budgets.
- **LLM Router**: A reasoning utility selecting the optimal model size, context limit, and cost profile for a given agent task.
- **Department**: An organizational structure grouping related agents (e.g., Research Department, Execution Department).
- **Agent**: A modular, single-responsibility AI actor executing tasks inside a department.
- **ADR (`ADR-XXXX`)**: Architecture Decision Record; a document capturing an engineering choice, its context, and its trade-offs.
- **RFC (`RFC-XXXX`)**: Request for Comments; a proposal detailing a new feature or design specification, circulated for review before writing code.
- **Design Paper (`DESIGN-XXXX`)**: A document analyzing technical options (pros/cons) and outlining the recommended implementation path before specifications are written.
- **Specification (`SPEC-XXXX`)**: A document detailing the precise technical requirements and interface designs for a system component.
|