This repository contains the source code for AuctionGym: a simulation environment that enables reproducible offline evaluation of bandit and reinforcement learning approaches to ad allocation and bidding in online advertising auctions.
AuctionGym was released in the context of our "Off-Policy Learning to Bid with AuctionGym" publication in the Applied Data Science Track of the 2023 ACM SIGKDD Conference. An earlier version of our work was presented at the AdKDD '22 workshop, where it received a Best Paper Award.
Offline evaluation of "learning to bid" approaches is not straightforward, because of multiple reasons: (1) observational data suffers from unobserved confounding and experimental data with broad interventions is costly to obtain, (2) offline experiments suffer from Goodhart's Law: " when a measure becomes a target, it ceases to be a good measure ", and (3) at the time of writing and to the best of our knowledge -- there are no publicly available datasets to researchers that can be used for this purpose. As a result, reliable and reproducible validation of novel "learning to bid" methods is hindered, and so is open scientific progress in this field.
AuctionGym aims to mitigate this problem, by providing a unified framework that practitioners and research can use to benchmark novel methods and gain insights into their inner workings.
We provide two introductory and exploratory notebooks. To open them, run jupyter notebook in the main directory and navigate to src.
" Getting Started with AuctionGym (1. Effects of Competition) " simulates second-price auctions with varying levels of competition, visualising the effects on advertiser welfare and surplus, and revenue for the auctioneer. Analogosuly, " Getting Started with AuctionGym (2. Effects of Bid Shading) " simulates first-price auctions where bidders bid truthfully vs. when they shade their bids in a value-based manner.
This section provides instructions to reproduce the results reported in our paper.
We provide a script that takes as input a configuration file detailing the environment and bidders (in JSON format), and outputs raw logged metrics over repeated auction rounds in .csv-files, along with visualisations. To reproduce the results for truthful bidders in a second-price auction reported in Fig. 1 in the paper, run:
python src/main.py config/SP_Oracle.json
A results-directory will be created, with a subdirectory per configuration file that was ran. This subdirectory will contain .csv-files with raw metrics, and .pdf-files with general visualisations.
Other configuration files will generate results for other environments, and other bidder behaviour.
See configuration for more detail on the structure of the configuration files.
Please cite the accompanying research paper if you use AuctionGym in your work:
@inproceedings{10.1145/3580305.3599877,
author = {Jeunen, Olivier and Murphy, Sean and Allison, Ben},
title = {Off-Policy Learning-to-Bid with AuctionGym},
year = {2023},
isbn = {9798400701030},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3580305.3599877},
doi = {10.1145/3580305.3599877},
booktitle = {Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining},
pages = {4219–4228},
numpages = {10},
keywords = {online advertising, counterfactual inference, off-policy learning},
location = {Long Beach, CA, USA},
series = {KDD '23}
}See CONTRIBUTING for more information.
This project is licensed under the Apache-2.0 License.