Winners Announced for Dynamic Pricing Competition 2019

Learn more

Winners Announced for Dynamic Pricing Competition 2019

Learn more

announcing the winners of 2019

The Dynamic Pricing Competition 2019 has come to a close. We have two winners across three scenarios. We are very excited to declare the winners!!!

The winner of the Duopoly as well as Dynamic scenario is Kyle Maclean,  PhD, Assistant Professor at Ivey Business School at Western University, Canada.

The winner of Oligopoly scenario is Ruben van de Geer, Phd, Data Maestro at GoDataDriven, The Netherlands.

Congratulations to Kyle Maclean & Ruben van de Geer .

The dynamic pricing competition - Outlook 2020

Pricing is one of the most challenging topics in the business world, with huge impacts when it comes to making profit or loss. In today’s world, prices are changed more frequently than ever before. Everyone involved tries to make the best use of data for their pricing decisions. However, designing data-driven price algorithms that perform well in competitive markets with uncertainty about customer behavior remains a big challenge.

To address this challenge, we invite people from academia and industry from various backgrounds(Operations Research, Pricing & Revenue Management, Machine Learning & AI – Reinforcement Learning) to let their ideas and algorithms compete in various basic market settings. By participating you will receive valuable insights on how your pricing strategy is performing: where it is good and outperforming others, and where is it lacking behind and gets beaten. Get the ultimate proof that your pricing algorithm is rocket science and unbeatable!

Contest details: We test and evaluate the performance of data-driven pricing algorithms in three different market scenarios: a duopoly with inventory constraints, a multi-product oligopoly, and a large market with horizontal product differentiation (more details on the three market scenarios are provided below). Each participating team is allowed to submit at most one algorithm for each market scenario. It is thus possible (but not necessary) to submit an algorithm for all three market scenarios; one can also participate in just one or two market scenarios. Algorithms have to be written in Python; more details on the required input and output variables and computation time restrictions are provided below.

Begin April/May until begin June, 2020 we run a training period, during which we simulate the performance of the algorithms each night. The simulations are based on undisclosed demand-generating mechanisms that are used to determine the average total revenue obtained by each algorithm. Based on these simulations we will provide information on the algorithm’s performance to the participants, which can be used to improve the algorithms. Thus, during the test period it is possible to replace every day one’s algorithm by a (hopefully) improved version.

By begin June all participant needs to submit their final algorithms for the “real” contest simulation runs. Rankings are based on the obtained average revenue over all simulations. Based on these final rankings we will determine the winner of the contest, for each of the three market scenarios. For each market scenario, the best performing team will receive 500 euro. In case of unforeseen events, the organizing committee decides on the final ranking.

The submitted algorithm will always be your property, we will never publish the code, we only use it exclusively within the competition and for academic purposes (e.g. comparing different price algorithms), but never for commercial purposes. In case you would like to keep your algorithm private, it is possible to opt-out of an eventual future publication or other academic use by sending us an email to info@dynamic-pricing-competition.com before the competition starts. With the final submission – not the test phase – we only ask for a brief (e.g. 100 words) description of your approach. These descriptions are only used in the performance evaluation, similar to the below paper with the results of the 1st contest edition in 2017 at the Informs RM & Pricing Conference in Amsterdam.

 

Organized by:
Registration will open again early 2020!

important dates

Final definition of competition setup and deadlines

end of February 2020

Start of training period

April/ begin May 2020

Deadline for final algorithm submission

begin June 2020

Announcement of winners and publishing of results

mid June 2020

pricing scenarios - 2019 edition

(1) Airline / Hotel (Duopoly)

A duopoly inspired by airlines, with finite capacities, a finite booking horizon and unknown demand function.

(2) Retail (Oligopoly)

An oligopoly with unknown demand functions, where each seller has multiple (substitute or complementary) products on sale.

(3) E-Commerce (Dynamic)

A large competitive market where each seller sells only a single product, with horizontal product differentiation and unknown demand function.

general process

Signup

Simple registration: we only require your name and email address. Latest after 2-5 days of registration, you will receive a username, login-url and password. Each team can only submit one algorithm per pricing scenario. Note that we might close registration when the number of signups exceeds our capacity.

Development & Training

The training period starts in April/beginning of May, where you can at night upload your algorithms. We perform training runs in which the algorithms already compete against each other. In the morning we will email you the logs of the nightly run.

Final contest & evaluation

Submission deadline is midnight 12 a.m. 1st of June. The ultimate simulation runs are performed to identify and crown the winner. Full and detailed performance reports are generated and shared with participants.

WHY YOU SHOULD JOIN?

... apart from winning and proving you are best ...

Prize money

The winner of each pricing scenario will receive an economic prize of an amount of 500 EUR.

Detailed performance evaluation

You will receive a detailed performance evaluation of your algorithm and the competition.

Scoring & Benchmark

Get a scoring of your performance based on the benchmarks set by the other participants. Plus, an optional score entry of your algorithms on our (to come) "Leaderboard", similar to the netflixprize.

Similar to the 1st edition 2017, we may try to publish the results and findings of the 2019/2020 competition again in a publication, depending on the academic value of the obtained results.

Paper on the 1st edition - Dynamic Pricing Challenge at the Informs RM & Pricing Conference in Amsterdam 2017

published in the Journal of Revenue and Pricing Management 2018

Organizer

 

For more information please contact:  info@dynamic-pricing-competition.com

Newsletter