Google Cloud Big Data and Machine Learning Blog

Innovation in data processing and machine learning technology

Halite helps students and developers compete to build better AI on Google Cloud Platform

Monday, October 23, 2017

By Harikrishna Menon, Product Manager, Two Sigma

Looking to show off your AI programming chops? Halite is an annual open source artificial intelligence programming challenge, created by New York-based Two Sigma, where players build bots using the programming language of their choice to battle on a two-dimensional virtual game board. This year’s challenge, Halite II, is a space-themed game where players’ bots control ships that mine planets to grow their fleets and defeat their opponents. Two Sigma designed Halite as an open-source project for community outreach and education, and it’s deployed on Google Cloud Platform.

Introducing the new Halite challenge

The Halite competition is a fun way to learn and apply AI, machine learning, and other advanced algorithms by writing smart agents in a collaborative yet also competitive setting. Designed for coding enthusiasts of all experience levels and open to the public, Halite is an enticing game environment to learn, write, and visualize code in action.

The 2017-2018 season runs from October to January, and participants can join until the last week of the competition. Players submit bots that play games against other bots until the Halite ranking algorithm determines a player rank. After submissions are complete, a winner will be determined after a week of final matches. Learn how to get started.

Some background

Figure 1: Halite I game board screenshot.

Two high school interns at Two Sigma conceptualized and built Halite I in 2016. Two Sigma had already established a record of playful programming challenges for its mathematical and software-oriented teams (for example, see the Robotic Air Hockey Competition). In the summer of 2016, Two Sigma ran an internal Halite Challenge, and ultimately opened Halite up to coding enthusiasts worldwide. Halite I was a great success, and developed a flourishing community of bot builders from around the globe that represents more than 35 universities and 20 organizations.

While the original Halite involved moving pieces around a board with only up-down-left-right options, 2017’s Halite II allows bots to battle for control of a virtual continuous universe, where ships mine planets to grow larger fleets and defeat their opponents. And while last year’s game featured bots that were built with machine learning techniques to further encourage the community to explore deep learning, reinforcement learning, and other techniques, this year’s game goes a step further: the Two Sigma team, along with support from Google, will make GPU-enabled instances available for competitors who can put them to use. To encourage novices to experiment with machine learning, the Halite team has put together an ML starter bot  to help competitors get started on building ML bots.

Successful players will win based on the relative sophistication of their code as well as their gaming strategies. The best tactics for Halite II will require the use of sophisticated pathfinding and swarming techniques, using concepts such multi-agent pathfinding.

How to compete

Please read the full terms and conditions here.1

It’s easy to sign up for Halite II and download a starter bot from an initial set for various programming languages, but you are also encouraged to publish your own starter kits if you are passionate about an alternate programming language. You can download a starter bot, try out one of our tutorials to make updates to the bot, and/or improve the bot on your own. The Halite team has made a number of API’s that allow a beginner to jump right in, but we expect more experienced players to try machine learning and other sophisticated AI techniques, in many cases coding a bot from scratch.

As you improve and re-submit your bots, you should see your rankings on the leaderboard change. You can also join one of many private or public hackathons to see how you compare to your peers, and even create custom leaderboards to see how your rank compares to friends, classmates, or colleagues over time.
Players may have an accelerated Two Sigma recruiting process if they achieve the ‘gold’ tier or above by the end of the game, representing the top ~.8% of players.

Halite I was especially successful because of its robust forum community. Now this is your chance, whether you’re a beginner or advanced, to join the forum to ask questions or learn best practices — not to mention make some online friends.

Figure 2: Halite II gameplay screenshot.

May the best bot win!

1The competition may be subject to other terms and conditions issued by Two Sigma. Participant should read all such terms and conditions to understand any additional restrictions on participation in the competition.
Google makes no representations, warranties or guarantees with respect to the competition, including without limitation the conduct, dealings of Two Sigma or any other third parties.
  • Big Data Solutions

  • Product deep dives, technical comparisons, how-to's and tips and tricks for using the latest data processing and machine learning technologies.

  • Learn More

12 Months FREE TRIAL

Try BigQuery, Machine Learning and other cloud products and get $300 free credit to spend over 12 months.

TRY IT FREE