Statistical experiments with stock data

The purpose of this repository is to hold and document a series of statistical experiments with real financial data, primarily over the US stock market daily price data from 1998 to mid-2022. The code and the necessary configurations are provided, so the experiments can be repeated by anyone with access to the data and some basic understanding of how to compile and run a command-line app.

A few random conventions throughout the study:

• Even though I'm a single person writing the study, I use the pronoun "we" to refer to myself and my readers, implying that "we're doing this together" (and not "We, Louis XIV"). Whenever I make a personal choice (without the reader), I use the pronoun "I" indicating that it's me who's doing it, it's my personal choice. The distinction is somewhat arbitrary, but it explains why sometimes I write "we" and sometimes "I".

• Important definitions or key moments will be in italics, while math and code fragments will be either an inlined code or in a

multi-line
code block.

Most plots and other types of experimental results are accompanied by a corresponding config to reproduce the plot or experiment. See the usage section for details on how to use such configs with the tool.

Table of Contents

• Methodology
• Market, Random Walks and Log-Profits
• Distribution of Log-Profits
• Hypothesis Testing, Confidence Intervals and Monte Carlo
  • Theory
  • Normal distributinon
  • Student's t-distribution
  • The Tale of Fat Tails - finding a (alpha)
• History and Market Timing: Auto-Correlation of Log-Profits

Installation

Requirements:

• Google Go 1.19 or higher;
• parfait-sharadar app to download data; see stockparfait/stockparfait for installation instructions;
• Subscription to Sharadar Equities Prices on Nasdaq Data Link

git clone https://github.com/stockparfait/experiments.git
cd experiments
make init
make install

This installs an executable experiments in your ${GOPATH}/bin, where GOPATH=`go env GOPATH`.

Quick start

• Subscribe to the data source on Nasdaq Data Link; most of these experiments use only the equities prices.
• Download the data by running parfait-sharadar - see stockparfait/stockparfait for details.
• Copy stockparfait/stockparfait/js folder contents to a separate working directory, for example ~/plots/; here I'll refer to it as ${PLOTS}.
• Run an experiment.

An example of these steps together:

${GOPATH}/bin/parfait-sharadar
mkdir -p ${PLOTS}
cp stockparfait/stockparfait/js/* ${PLOTS}
experiments -conf ${CONFIG}.json -js ${PLOTS}/data.js

where ${CONFIG}.json is the config of your choice from one of the experiments, or your own. The schema of such configs is in config/config.go.

• Open ${PLOTS}/plot.html in your browser to see the resulting plots.

Contributing to Stock Parfait Experiments

Pull requests are welcome. We suggest to contact us beforehand to coordinate your code contributions.

Having said that, this repository serves primarily as the documentation of my own research into the behavior of the US stock market, and also as an example and the purpose and motivation behind the core libraries in stockparfait/stockparfait.