Hello, Codeforces!
We are excited to announce the second Cambridge Battlecode event this year with £15,000 in prizes!
Cambridge Battlecode is a 4 week long programming competition, where players compete to write autonomous bots that battle head to head in a strategy game.
This time we are trying something new and have taken inspiration from MOBAs such as League of Legends, Dota 2 and Heroes of the Storm and are planning to release with 12 unique champions! We've revealed 7 of them so far with more to come over the upcoming weeks!
The competition kicks off in under a month, on the 12th of October, and will run across three weekly sprints, culminating in our in-person finals on the 18th of November. Finalists will also have the opportunity for an office visit at HRT for recruitment and networking, with flights and 4 days of accommodation fully covered by us!
We're also incredibly grateful to the sponsors supporting us this year as this event would not be possible without them:
- Platinum: HRT
- Gold: Citadel and Citadel Securities, G-Research, Jane Street and QRT
- Silver: QuantCo
- Bronze: Anthropic and Pareto
Registrations are now live! You can sign up and find out more on our website:
As well as on our discord server: https://discord.gg/KaNZWaHtxX
To add, we are supporting C++, Python, Rust, OCaml and Kotlin this year!
We're really excited for what we've got planned, please do spread the word, and we hope to see you competing!
Auto comment: topic has been updated by awu (previous revision, new revision, compare).
Auto comment: topic has been updated by awu (previous revision, new revision, compare).
This is actually so interesting, looking forward to it!
What are the rules on LLM usage? I'm assuming that they are allowed, so I'm curious to know, have the authors confirmed that they are able to build something stronger than the frontier models are able to build with no human input?
LLMs are allowed, but based on my experience with the past edition, they are not very useful on their own. They can speed up implementation for sure and even write algorithms that you would not otherwise know how to write. But they are not very useful in coming up with ideas that would make your bot significantly better. I may be wrong (especially now as AI is getting better), but they are allowed and it is a skill to be able to use AI to make your bot better. So use it as you wish!
Ok thanks for letting me know! I find it quite surprising tbh, since this seems like exactly the type of task that LLMs excel at (try many ideas and there is a clear way to evaluate performance).
In particular, from my testing, the llms struggle with the "try many ideas" part. I see the "problem of importance" manifest, as they tunnel vision on trying irrelevant things instead of having good research taste (I believe this is quite a common issue with ai automated research in general)
I can confirm this is a very fun competition (especially with the discussions on the Discord server!)