Mission
Math, applied to things
people already care about
Pattern is a collection of simulation tools and data-driven investigations. It started as a way to understand how sector returns, audio spectra, NBA shot selection, and tumor growth curves behave as systems you can model.
Each investigation begins with a concrete question, applies a documented method to a public dataset, and shows the math behind the answer. Some of these hold up better than others, and the limitations section says so plainly.
About the Author
Aalay Shah
Aalay Shah is a high school researcher from Sammamish, Washington, with interests in mathematics and finance. He plays basketball and piano, and plans to study applied mathematics. His work has appeared in the National High School Journal of Science and the Oxford Journal of Student Scholarship.
Selected publications
- "ChatGPT as a Real-Time Sector Rotation Predictor: Accuracy, Bias, and Accessibility in Earnings-Driven Market Movements"National High School Journal of Science
- "Quantifying the Impact of Macroeconomic and Geopolitical Events on Stock Market Behavior Using Time-Series Deviation Modeling"Oxford Journal of Student Scholarship, Vol. 8, Issue 2 (June 2026)
Principles
No fabricated data
Every dataset links to a public source you can check yourself. When data is missing, we say so rather than filling the gap with something invented.
Documented methodology
Every investigation includes the method, the data dictionary, and enough detail that someone with the source data could redo the analysis from scratch.
Explicit limitations
Every investigation has a limitations section. Good analysis is honest about what the data does not show, and null results get published here too.
Core Disciplines
Applied Mathematics
Differential equations, linear algebra, real analysis, and discrete mathematics as the underlying language of quantitative reasoning.
- Ordinary & partial differential equations
- Linear algebra & matrix methods
- Optimization theory
- Real analysis
Data Science
Statistical methods for extracting knowledge from structured and unstructured data at scale, with rigorous attention to uncertainty quantification.
- Statistical estimation & inference
- Machine learning methodology
- Regression & classification
- Cross-validation & model selection
Signal Processing
The mathematics of transforming, analyzing, and extracting information from signals, from audio waveforms to financial time series.
- Fourier & wavelet transforms
- Spectral analysis
- Filter design
- Time-frequency representations
Optimization
Finding the best solution from a set of feasible alternatives, subject to constraints: the core mathematical framework for decision making.
- Linear & quadratic programming
- Convex optimization
- Stochastic optimization
- Game theory
Statistical Modeling
Building probabilistic descriptions of data-generating processes, quantifying uncertainty, and testing hypotheses rigorously.
- Bayesian & frequentist inference
- Generalized linear models
- Time series modeling
- Causal inference
Machine Learning
Algorithmic methods that allow systems to learn from data, applied at Pattern with attention to interpretability and validation.
- Supervised & unsupervised learning
- Neural networks
- Ensemble methods
- Feature engineering
Where the data comes from
Everything here pulls from public sources: government databases, university repositories, league APIs, and peer-reviewed open datasets. Nothing is purchased, fabricated, or synthetically generated without a clear label saying so.
SEC EDGAR
Full-text earnings filings
Federal Reserve (FRED)
Macroeconomic time series
JHU CSSE
COVID-19 outbreak data
NBA Stats API
Basketball shot & tracking data
Basketball Reference
Historical NBA statistics
Compmusic / UPF
Hindustani raga recordings
Univ. of Iowa EMS
Instrumental sound recordings
Benzekry et al. (2014)
Tumor growth datasets (PLOS CB)
Philadelphia Fed SPF
Economic forecast accuracy data
Good Judgment Project
Probabilistic forecasting data
Start exploring
Browse the investigations, download the datasets behind them, or open a tool and run the numbers yourself.