Squeeze Labs / Research

Research

Current studies and experiments using Squeeze’s datasets and analysis tools. The directions below describe planned work, with results and methodology on each study’s page.

Short interest & short squeezes

How does reported short positioning relate to volatility and subsequent price extremes?

Explore short-interest research

Current work

A descriptive event study of the tracked universe, with days-to-cover and prior-volatility cohorts. Linked stock charts place adjusted prices, volatility, and short-interest reports on a shared timeline.

Planned directions

Improve historical float coverage, examine reporting delays, and test whether relationships persist across liquidity groups and market regimes. Separate short-covering explanations from other causes of large price moves.

Report dates are correct; exact publication timestamps are currently unknown. Current results describe associations rather than a tradable signal.

News impact · CatalystBERT-v1

Can the content of a news article help identify a significant subsequent market reaction?

View CatalystBERT-v1

Current work

ModernBERT-large experiments using news articles, with return-based labels across multiple horizons. The study page documents input and output tensors, the connection to returns, methodology, and evaluation results.

Planned directions

Investigate temporal robustness, probability calibration, and variation by article type and reaction horizon. Compare model results with simpler baselines before drawing conclusions about trading usefulness.

The v1 experiment uses news articles. Its evaluation does not establish a profitable directional trading strategy.

Strategy backtesting

How do widely used technical strategies behave across a broad universe?

Inspect strategy experiments

Current work

MACD, RSI, and TTM Squeeze backtests across tracked symbols. Interactive distributions and individual account histories expose returns, drawdowns, and other performance metrics.

Planned directions

Compare parameter sensitivity and stock characteristics, then evaluate transaction costs and performance outside the parameter-selection period. Use these experiments as baselines for more targeted research.

Default parameters provide a common comparison. Individual stocks may perform better with targeted parameters; tuning requires separate validation.