General

This repo contains an oTree application customized for conducting real-time financial market experiments. It originally started as an attempt to build an oTree implementation of the experiment in [Aldrich_LopezVargas]. It now supports a large set of financial market environments.

This application connects to a remote exchange server, and human participants operate as traders. See more details in [Aldrich_Demirci_LopezVargas].

In this paradigm, each oTree subsession corresponds to a trading day and one webpage where subjects participate in a market by interacting with the components on the user interface.

This architecture is inteded for creating experimental environments to study algorithmic and high-frequency trading.

Demo

There is a simple demo available for the most recent implementation called 'Exogenous Limit Orders' environment, which can be accessed through this link. To run a session in this environment, select one of the configurations presented on the landing page and follow the instructions. You will participate in a trading session, equipped with a tunable trading bot.

Docker

clone this repository.

git clone https://github.com/Leeps-Lab/high_frequency_trading.git
cd high_frequency_trading

run a multicontainer docker application that includes an otree, continuous double auction and frequent batch auction servers along with other components.

docker-compose up

it will take a while until all services start, then go to localhost on your browser.

Server Setup

Redis is used as the primary data storage during a trade session for quick read/writes, and experiment data is written to the Postgres in background.

We use Huey for this purpose. Both Redis and Huey are already required for oTree. The interface server is used to connect to exchange server.

Prerequisites:

The tutorial below assumes that you have Python 3.6 (thus, pip3) installed in your computer as well as an up-to-date Google Chrome browser.

Redis and Postgres databases should be running and oTree configured to talk with them. See oTree docs documentation for details. You can also find instructions to install and run Redis here. Similarly, can download and install Postgres here. We were able to run a couple of successful tests using SQLite (the default development database for otree) instead of Postgres.

Easy install: See the vagrant based setup.

OR

Step-by-step tutorial to run a simple test

0. Open four terminals.

1. Create a virtual environment, you will install a slightly modified version of oTree in this new environment. A virtual environment will keep this version separate from the oTree version you might be already using. Warning: If you have a version of oTree installed in your computer and do not use a virtual environment to do Step 1, you will overwrite your current oTree installation.

In terminal #1, make sure to have virtualenv installed by checking the version.

virtualenv --version

The version for virtualenv should be printed on console, else install virtualenv

pip3 install virtualenv

Then run

virtualenv -p python3.6 otree_hft_env

2. Activate the virtual environment (still in terminal #1).

For mac and linux

source otree_hft_env/bin/activate

For windows

otree_hft_env/Scripts/activate

3. Using these commands, clone this repository, cd into the folder and install dependencies (Terminal #1).

git clone https://github.com/Leeps-Lab/high_frequency_trading.git
cd high_frequency_trading
pip3 install -r requirements.txt

4. In Terminal #2, follow the exchange server instructions and run a CDA exchange instance. Expect to see three timestamped lines that look like this

[14:45:00.803] Using selector: KqueueSelector
[14:45:00.803] DEBUG [root.__init__:35] Initializing exchange
[14:45:00.803] INFO [root.register_listener:112] added listener 0

5. Go back to Terminal #1, reset the database and copy static files by running these commands.

otree resetdb
otree collectstatic

6. Then, in the same Terminal #1, run oTree server.

otree runhftserver

Note: this step requires Redis to be running either in the background or in a separate Terminal (run 'redis-server' in Terminal #4)

7. In Terminal #3, go to the folder that contains 'otree_hft_env' and do Step 2 (activate the virtual environment). Then, cd into the 'high_frequency_trading' folder and start the following background process.

cd high_frequency_trading
otree run_huey

8. Open your Chrome browser and go to localhost. Click on the 'demo session' and follow the screen instructions to launch clients' (traders') screens as tabs in the same browser.

Final notes

Here, we have four terminals running four processes that conform to our financial market environment. These processes are talking to each other during a trading session.

In production mode, you should run each of these as a 'service'. The method above is only intended for testing on your personal computer.

[Aldrich_LopezVargas]

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3154070

[Aldrich_Demirci_LopezVargas]

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3354426