This repository contains the code for our team's submision in Optiver's 2023 Ready Trader Go contest: a trading competition where students learn all the fundamentals of trading and truly test their coding and analytical skills. The aim is to build and optimise a trading algorithm that outperforms all others on a virtual exchange.
Our submission placed us in the top 17-32 of teams out of 2800 accepted, placing us in the top ~1%, unfortunately just one match win away from a spot in the top 16 and on the leaderboard. Nevertheless we found participating immensely fun as we got to quickly iterate on a solution to an interesting technical problem.
More information about the contest is available at readytradergo.optiver.com.
The algorithm we use for our trader is very simple but highly optmized. We act as market makers and try to constantly make small, profitable trades. We estimate the true price of the instrument we're trading by taking the average of prices it's traded at weighted by volume and then make a bid or ask order, depending on whether the current price is below/above the true price. We set this spread dynamically according to how are recent trades have been performing. We keep multiple orders alive and cancel them if they've existed for a given number of ticks or if they've become unprofitable. Additionally, to minimize risk we deploy a (delayed) hedging strategy, making simultaenous trades in two instruments so as to keep our position close to neutral.
We considered many other features and improvements such as adjusting spread based on recent market volatility but ultimately rejected them as they performed more poorly than our relatively simple approach. To minimize latency, in the last days of the contest we also rewrote our submission in C++, which gave an appreciable performance increase.
To run Ready Trader Go, you'll need Python version 3.11 or above and PySide6. To install PySide6, run:
pip3 install PySide6in your Python virtual environment.
To run a Ready Trader Go match with one or more autotraders, simply run:
python3 rtg.py run [AUTOTRADER FILENAME [AUTOTRADER FILENAME]]For example:
python3 rtg.py run autotrader.pyEach autotrader must have a corresponding JSON configuration file as described below.
This archive contains everything needed to run a Ready Trader Go match in which multiple autotraders compete against each other in a simulated market. For the exact definition of a match, see the competition terms and conditions.
The archive contains:
- autotrader.json - configuration file for an example autotrader
- autotrader.py - an example autotrader
- data - sample market data to use for testing
- exchange.json - configuration file for the exchange simulator
- ready_trader_go - the Ready Trader Go source code
- rtg.py - Use this with Python to run Ready Trader Go
Each autotrader is configured with a JSON file like this:
{
"Execution": {
"Host": "127.0.0.1",
"Port": 12345
},
"Information": {
"Type": "mmap",
"Name": "info.dat"
},
"TeamName": "TraderOne",
"Secret": "secret"
}
The elements of the autotrader configuration are:
- Execution - network address for sending execution requests (e.g. to place an order)
- Information - details of a memory-mapped file for information messages broadcast by the exchange simulator
- TeamName - name of the team for this autotrader (each autotrader in a match must have a unique name)
- Secret - password for this autotrader
The market simulator is configured with a JSON file called "exchange.json". Here is an example:
{
"Engine": {
"MarketDataFile": "data/market_data.csv",
"MarketEventInterval": 0.05,
"MarketOpenDelay": 5.0,
"MatchEventsFile": "match_events.csv",
"ScoreBoardFile": "score_board.csv",
"Speed": 1.0,
"TickInterval": 0.25
},
"Execution": {
"host": "127.0.0.1",
"Port": 12345
},
"Fees": {
"Maker": -0.0001,
"Taker": 0.0002
},
"Information": {
"Type": "mmap",
"Name": "info.dat"
},
"Instrument": {
"EtfClamp": 0.002,
"TickSize": 1.00
},
"Limits": {
"ActiveOrderCountLimit": 10,
"ActiveVolumeLimit": 200,
"MessageFrequencyInterval": 1.0,
"MessageFrequencyLimit": 50,
"PositionLimit": 100
},
"Traders": {
"TraderOne": "secret",
"ExampleOne": "qwerty",
"ExampleTwo": "12345"
}
}
The elements of the autotrader configuration are:
- Engine - source data file, output filename, simulation speed and tick interval
- Execution - network address to listen for autotrader connections
- Fees - details of the fee structure
- Information - details of a memory-mapped file used to broadcast information messages to autotraders
- Instrument - details of the instrument to be traded
- Limits - details of the limits by which autotraders must abide
- Traders - team names and secrets of the autotraders
Important: Each autotrader must have a unique team name and password
listed in the 'Traders' section of the exchange.json file.
The Ready Trader Go command line utility, rtg.py, can be used to run or
replay a match. For help, run:
python3 rtg.py --helpTo run a match, use the "run" command and specify the autotraders you wish to participate in the match:
python3 rtg.py run [AUTOTRADER FILENAME [AUTOTRADER FILENAME]]Each autotrader must have a corresponding JSON file (with the same filename,
but ending in ".json" instead of ".py") which contains a unique team name
and the team name and secret must be listed in the exchange.json file.
It will take approximately 60 minutes for the match to complete and several files will be produced:
autotrader.log- log file for an autotraderexchange.log- log file for the simulatormatch_events.csv- a record of events during the matchscore_board.csv- a record of each autotrader's score over time
To aid testing, you can speed up the match by modifying the "Speed" setting in the "exchange.json" configuration file - for example, setting the speed to 2.0 will halve the time it takes to run a match. Note, however, that increasing the speed may change the results.
To replay a match, use the "replay" command and specify the name of the match events file you wish to replay:
python3 rtg.py replay match_events.csvAutotraders in Ready Trader Go will be run in the following environment:
- Operating system: Linux
- Python version: 3.11.2
- Available libraries: numpy 1.24.2; pandas 1.5.3; scipy 1.10.1
- Memory limit: 2GB
- Total disk usage limit: 100MB (including the log file)
- Maximum number of autotraders per match: 8
- Autotraders may not create sub-processes but may have multiple threads
- Autotraders may not access the internet