Proposal: Modernizing Robocode Restrictions for AI Development

11 views
Skip to first unread message

嗯嗯(窮苦肥宅一號)

unread,
Jun 14, 2026, 6:24:36 AM (11 days ago) Jun 14
to Robocode
## Description
Robocode’s current restrictions were designed in the early 2000s when hardware resources were limited. While these rules ensured fairness at the time, they now hinder creativity and prevent modern AI-driven robots from fully participating in competitions.

## Problem Statement
- File I/O limit is capped at 200 KB, which was reasonable in the past but is now outdated given modern storage capacities.
- Robots are forced to terminate immediately after matches, preventing background self-learning or post-match analysis.
- These restrictions no longer align with today’s hardware capabilities (even smartphones exceed the performance of early 2000s PCs).

## Proposed Reforms
1. **Increase File Limit**  
   - Raise the robot file I/O limit from 200 KB to at least 1 MB.  
   - This allows robots to store richer training data and strategies.

2. **Allow Background Training via API**  
   - Enable robots to request permission through an API from the Robocode engine to continue self-learning in the background after matches.  
   - Foreground matches remain unaffected, ensuring fairness.

3. **Resource Quota System**  
   - Define CPU/memory quotas for background training to prevent interference with other matches.

4. **Post-Match Analysis API**  
   - Provide full logs and snapshots for continuous strategy iteration and learning.

## Benefits
- Encourages creativity and innovation among learners.  
- Aligns Robocode with modern hardware environments.  
- Transforms Robocode from a teaching game into a true AI competition platform.

## Request
I would like the development team to consider these reforms and discuss their feasibility. This proposal aims to modernize Robocode while preserving fairness and accessibility.

fnl

unread,
Jun 14, 2026, 6:34:44 AM (11 days ago) Jun 14
to Robocode
Thank your for the proposals! 😊

I just realized that the GitHub repo for Robocode does not have the issue tracker. So I just added this here:
https://github.com/robo-code/robocode/issues

I should like you to create the feature request here instead, as this forum is more about developing bots in Robocode, and not fixing the game.
And you and others should be able to follow the progress and make more proposals and adjustments to the proposals. This is not the right forum for that.

You have 4 different feature proposals. Please create a feature request for each of those here marking those as feature request:
https://github.com/robo-code/robocode/issues/new

Thank you in advance. I look forward to follow up on the feature request at the GitHub repo for Robocode:
https://github.com/robo-code/robocode#

Pavel Šavara

unread,
Jun 14, 2026, 8:59:45 AM (11 days ago) Jun 14
to robo...@googlegroups.com
You can use `-DNOSECURITY=true` to disable many of the constraints when training on your own machine.

On the other hand, I'm very interested in allowing players to use their LLM to help with coding better robots.
I'm playing with such an idea myself. Long time ago I helped to write a battle recorder, so that we can learn from battles in offline learning.

To be able to run recorded battles in un-constrained environment, I'm building "experience replay" engine for last few weeks.


I'm not sure we want the robocode robot to become DNN or LLM with GPU requirements. That would be a very different game.

 
## Proposed Reforms
1. **Increase File Limit**  
   - Raise the robot file I/O limit from 200 KB to at least 1 MB.  
   - This allows robots to store richer training data and strategies.


I think this is a valuable constraint. Also code size is valuable in the same way.
 
2. **Allow Background Training via API**  
   - Enable robots to request permission through an API from the Robocode engine to continue self-learning in the background after matches.  
   - Foreground matches remain unaffected, ensuring fairness.

I agree we could make it possible to lift the thread killing feature of the game engine. For the local environment, not for tournaments.
Being able to learn from online battles under time constraints is part of the challenge.

my 2c
Pavel

Reply all
Reply to author
Forward
0 new messages