SPY Research Analyst

Overview

This role involves conducting equity research focused on the SPY, the S&P 500 ETF, adhering to a rigorous shared standard that includes causal feature analysis, walk-forward testing, baseline comparisons, and an explicit willingness to halt projects lacking an edge. The live production process concentrates on short-horizon realized volatility, with experiments failing to meet these criteria documented as null results and excluded from production.

Research Vision

The objective is to develop a comprehensive SPY market-state suite composed of modular models each addressing specific questions such as predicting next-horizon volatility, session range expansion, identifying market regimes, and directional trends. This unified system relies on shared data inputs and operational protocols, offering transparent and verifiable track records rather than relying on a single black-box model.

The suite features:

• A consistent data foundation using SPY OHLCV data and optional VIX inputs.
• An operational framework including registries, candidate versus active model differentiation, promotion criteria, health monitoring, and ledger maintenance.
• A product interface presenting panels for volatility, range, regime, direction, and performance tracking.

Research culture mandates walk-forward testing exclusively, use of label-native baselines, separation of freeze and promotion phases, and public sharing of failures. Models predicting multiple factors simultaneously or reopening frozen models for different targets are excluded, as is passing research results directly into production without appropriate validation.

Live Production

• Next 5-day annualized realized volatility forecasts for SPY, compared against naive persistence and GARCH(1,1) benchmarks.
• Monthly LightGBM model architecture with nine key baseline features, monthly gated weight updates, and a stable modeling ladder without frequent architecture changes.
• Production and operational controls include strict freeze and promotion protocols, with a public-facing pipeline and data board.

Research Products and Infrastructure

• SPYForge: An application of the GraphForge system designed to encode product laws as typed state graphs and verify model pipelines through freeze and gating processes, with agent-based access controls. SPYForge serves to verify operational compliance rather than forecast skill, supporting the live vol pipeline before formal cutover.
• Additional modules under development or on hold include range expansion modeling, regime classification, and directional forecasting.

Null Results and Research Integrity

A documented null result exists for SPY open-to-close directional forecasting which found no predictive advantage over always-up baseline models after realistic cost considerations. Negative findings are valued and shared under the same stringent evaluation standards.

Disclaimer and Legal Information

This experimental research report is not financial advice and carries no guarantees regarding forecast accuracy or trading performance. The research does not endorse, sponsor, or affiliate with any ETF providers or index issuers mentioned, and all trademarks belong to their respective owners. Data are sourced from publicly available feeds and may be incomplete or delayed.

Martial Systems LLC is an independent entity unaffiliated with Kalshi Inc. or other named companies. Unauthorized reproduction of materials is strictly prohibited.