This repository provides a suite of Python packages and tools for time-series analysis, prediction, and optimization, targeting quantitative analysts, data scientists, and researchers interested in collective intelligence and market-inspired prediction mechanisms. It aims to advance online autonomous prediction through a collection of specialized libraries and a platform for benchmarking and contests.

How It Works

The core of the project revolves around the timemachines package, which enumerates online time-series methods and benchmarks univariate models against data streams from the microprediction platform. It defines a functional interface for prediction: f : (y_t, state; k) \mapsto ([\hat{y}(t+1), \dots, \hat{y}(t+k)], [\sigma(t+1), \dots, \sigma(t+k)], posterior\_state), where $\sigma$ represents prediction uncertainty. This approach emphasizes pure functions and online, incremental updates for prediction and state management.

Quick Start & Requirements

• Install via pip: pip install microprediction
• Requires Python 3.x.
• Official quick-start and documentation are available via links within the README.

Highlighted Details

• Benchmarks popular time-series packages using Elo ratings derived from microprediction data streams.
• Offers humpDay for derivative-free optimizer comparisons and precise for portfolio optimization.
• Includes a Monte Carlo simulation game (monteprediction) with a scoring mechanism and Slack community.
• Features a fast algorithm for inferring relative ability from win probabilities, applicable beyond sports analytics.

Maintenance & Community

The project is primarily driven by Peter, a career quant and applied mathematician. Community interaction and support are available via Slack, with links provided in the README.

Licensing & Compatibility

The README does not explicitly state a license. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

The real-time time-series platform previously maintained by the author is currently defunct but may be revived in the future. Some other repositories are described as speculative or quirky, indicating potential for instability or experimental features.