Abstract

The financial world of equity trading relies on centralised exchanges. These exchanges are disseminated around the globe. Thus, it is sometimes possible to trade a share of one given company in multiple regions. Arbitrage pricing theory tells us that the price on those exchanges should roughly be equivalent. However, because information speed is bounded, shocks on a given share price might take some time to reach all exchanges and be incorporated into the new stock prices.

General

This repository contains all the files used to analyse the price propagation delay for Shell's stock (RDSA) across multiple exchanges. It is organised into multiple folders and notebooks.

Folders

The repository contains the following folders :

figures: this folder contains HTML version of all the plotly figures produced during the analysis.
helpers: contains multiple python scripts used to process, load and extract information from the raw data. It also contains the config.yml file. More details about these files here.
results: folder containing results of big computations such as the lags time series.
scripts: contains a script used to efficiently push figures on the desired GitHub branch.

Notebooks

Most of the analysis is performed through notebooks. They all used the helper function defined in ./helper and are organised as follows :

data_exploration.ipynb: first glance at the raw data. It contains the experiments that we conducted on the data to understand how to approach the problem. We also developed the peak_finding algorithm in that notebook.
lag_computer_dask.ipynb: combines the peak_finding algorithm with Dask helpers to compute and save (into the results folder) the daily lags.
liquidity_computer.ipynb: compute the liquidity metric (daily median of the period between two consecutive trades) using Dask.
mean_price_and_ret_extractor.ipynb: extracts (using dask) the daily mean price and returns from the data files (one file per day) into time series.
results_exploration.ipynb : in this notebook, we study the obtained results (lags) and generate multiple plots out of them. These plots are saved in the figures folder.
viz.ipynb: contains interactive visualisation of the data.
viz_globe.ipynb: contains a visual comparison of the lags with the distance separating exchanges using a world globe.
simple_stats.ipynb: contains some simple statistics (average number of trades per day...)

Extra files

runtime.txt: file created by Heroku. Do not modify it.
requirements.txt: text file containing the requirements to run the code.

How to run the code

To run the analysis by yourself follow the described steps :

run the command pip install -r requirements.txt to obtain the needed libraries to run the code.
edit the helpers/config.yml file to choose the signal to work with : trade of bbo.
Download the data (here) into a folder named data. Note that you need access rights to use the mentioned link.
Run the "computational" notebooks : lag_computer_dask.ipynb, liquidity_computer.ipynb and mean_price_and_ret_extractor.ipynb.
Run the analysis and visualisation notebooks : data_exploration.ipynb, results_exploration.ipynb,viz.ipynb,simple_stats.ipynb and viz_globe.ipynb.

giordano-lucas / exchange-price-propagation