Our Research

A closer look at the questions our research teams explore.

Explore five projects from our Spring 2026 Research Showcase, spanning systematic strategies, market interconnectedness, and research infrastructure.

Quantitative Research / Spring 2026

Systematic Mean Reversion in Growth-Value Equity Pairs

We investigate short-term reversals in technology stocks through the lens of growth-value rotation. Rolling, exponentially weighted, and adaptive z-scores identify price overextensions; offsetting retail-stock positions serve as value-oriented hedges, without correlation or cointegration screening.

The study combines selected configurations into a diversified strategy index and tests transaction and financing costs across five scenarios. The reported in-sample results suggest that conditional mean reversion can survive trading frictions. Out-of-sample testing and evaluation across different market regimes remain next steps.

Read presentation PDF

If the preview is unavailable on your device, use the link above.

Quantitative Development / Spring 2026

Alternative Data Scraping

Our team built a serverless pipeline to collect news prominence and company hiring data for future trading-signal research. The April 2026 showcase reports coverage of 30 news sources and successful job-board scraping for 470 of 503 companies tested, using 15 applicant-tracking providers and timestamped PostgreSQL snapshots.

The dataset supports questions about whether headline placement, story convergence, hiring velocity, and wage changes anticipate market behavior. Those signals remain hypotheses: longer data histories and further validation are needed, while sentiment analysis and topic extraction are planned extensions.

Read presentation PDF

If the preview is unavailable on your device, use the link above.

Quantitative Research / Spring 2026

Cross-Country Sovereign Yield Spillovers

We adapt the Diebold-Yilmaz spillover framework to government bond yields across 11 countries and four maturities, using vector autoregression and forecast-error variance decomposition to measure cross-market connections. The study finds stronger spillovers at longer maturities and sharp increases during global stress events.

Historical analysis links lower return-spillover environments to stronger emerging-market versus safe-haven spread performance, with the pattern appearing in both pre-2020 and post-2022 subsamples. The model identifies predictive relationships rather than causality; omitted global factors, country-selection bias, and daily time-zone alignment remain limitations.

Read presentation PDF

If the preview is unavailable on your device, use the link above.

Quantitative Research / Spring 2026

Event-Driven Implied Volatility Mispricing

We study how options price uncertainty around earnings, FOMC decisions, CPI and unemployment releases, geopolitical shocks, and product launches. The framework compares long and short at-the-money straddles across event-relative entry and exit windows, using individual equities and liquid ETFs.

Results vary substantially by sector, event, and holding period: the broad pre-earnings long-straddle edge is weak, while sector-specific strategies and selected product-launch windows show stronger historical results. Limited test samples, missing historical liquidity measures, and unmodeled early assignment constrain interpretation; the technology earnings backtest also assumes negligible transaction costs.

Read presentation PDF

If the preview is unavailable on your device, use the link above.

Quantitative Development / Spring 2026

Agentic Backtest Evaluator

We built an AI-assisted evaluator that turns backtest CSVs and strategy descriptions into structured reports and follow-up conversations. It combines performance metrics, drawdown analysis, transaction-cost sensitivity, market news, and retrieval from a curated research knowledge base to help explain strategy vulnerabilities.

Across 10 benchmark scenarios scored by an LLM judge, the evaluator achieved quality scores comparable to ChatGPT and Claude while examining more time periods and using more tools. The results point to broader analysis, with efficiency still to improve. News-coverage gaps, simple VIX-based regime rules, and occasional fabricated headlines remain priorities for refinement.

Read presentation PDF

If the preview is unavailable on your device, use the link above.