Video summary
UNSW Battlecode Grand Finals
Main summary
Key takeaways
Event and Game Overview
The UNSW Battlecode Grand Finals brought together 972 teams from 41 countries. Organizers reported 2 million games played during the 25-day competition, with bots uploaded about 23,000 times. Teams programmed bots in Python, C, or C++ to compete in a multiplayer, Snake-like game featuring sea dragons.
Each team used the same program for all its dragons. Games could include up to 64 dragons per team and last up to 500 rounds. Dragons collect pellets and powers to grow, but die if they hit kelp. Hidden-destination portals, restricted vision, and the need for many identical bots to cooperate made navigation and coordination difficult.
Important Mechanics and Strategies
- Queen survival matters: Each team has a designated queen dragon. Keeping it alive is often essential; if all queens die, the result depends on other criteria, such as the longest remaining dragon.
- Explore and consolidate: Teams must discover resources with only a 7-by-7 view around each dragon, then coordinate to turn scattered resources and dragon length into a stronger queen or other long dragons.
- Choose when to split or merge: Splitting creates more explorers and can help a dragon escape danger, but teams need to consolidate effectively—especially near the end of a game.
- Adapt to the map: Maps varied in size, portal layout, resource density, and hazards. Teams that relied on fixed assumptions, such as always heading for the center, could be disadvantaged on maps where resources were elsewhere.
- Balance safety and aggression: Some bots protected their queens in safe areas; others used them to explore or bait opponents. Aggressive incursions could eliminate an enemy queen, but exposed a team’s own dragons to risk.
- Use replays and testing: Finalists described tuning heuristics on generated maps, reviewing replays for ideas, and adjusting strategies to map size. Several teams said the 64-dragon limit became a major constraint on larger maps.
The organizers also ran an estimate-a-thon, where spectators updated estimates about the finals as matches unfolded.
Tournament Highlights and Results
The finals used a large selection of unfamiliar maps, including Manhattan, Twin Ramparts, Strange Queens, Summoner’s Rift, Hydra, and Invisible Cities. The unfamiliar layouts tested whether bots could generalize rather than rely on hard-coded routes.
Several matches were especially close. Test 4 upset teams during the early bracket, while Aha and Forgot to Mention went to a tiebreaker after a tied series. The second-chance bracket also produced a close contest between Test 4 and Forgot to Mention.
The announced final standings were:
- Chad GDP — champion and $8,000 prize
- Aha
- Lop
- Calorie
- We Have Quizzes
- Forgot to Mention
- Test 4 and Sponge (tied)
- SSS and TTT Is All You Need (tied)
Chad GDP defeated Aha in the grand final. Chad’s team described using a two-stage search: a deeper, aggressively pruned search to identify promising moves, followed by a shallower, more detailed search to choose between them. Aha emphasized keeping its queen alive while trying to eliminate the opponent’s queen.
Other awards included an HRT bug bounty for a team that found a serious judge-compilation vulnerability, and Jump Trading’s fastest-algorithm award, which went to Lop.
Talks and Acknowledgements
Jump Trading speakers Matthew Fulcher and Ali discussed the firm’s evolution toward automated, research-driven trading, the importance of speed and careful handling of edge cases, and how competitive-programming skills can transfer to industry. HRT speaker Noni talked about FPGA engineering, workplace culture, and the company’s Explore, internship, and full-time pathways. Dr. Henry Hickman of UNSW Computer Science and Engineering encouraged students to explore research and keep the joy and creativity in programming.
The event acknowledged the Bidjigal people, Traditional Custodians of the land, and thanked sponsors and partners Jump Trading, HRT, and UNSW Computer Science and Engineering, alongside organizer UNSW Competitive Programming and Mathematics Society.
Teams and People Featured
Competition teams mentioned: Lop Stock (also referred to as Lop), Chad GDP (also called Chad GDT/Chad GPT during the event), Cutlery/Calorie, R, Aha, We Have Quizzes, SSS, Sponge Albert and Bob, TTT Is All You Need, Test 4, and Forgot to Mention. Special-award teams included FundaGong, Hieroglyphics, and the HRT bug-bounty team of Adam, Daniel, Eric, and Matthew.
Other people mentioned: Finalists Victor, Jeremy, and Liam, the primary-school competitors; Sebastian Miles and Milan, recognized as strong non-APAC competitors; and speakers Matthew Fulcher, Ali, Noni, and Dr. Henry Hickman.
Source: UNSW CPMSoc, UNSW Battlecode Grand Finals.
Rate this summary
Your feedback will help improve summaries.
Improve this summary
Reprocess with a stronger model when the summary feels incomplete or inaccurate.
Translate summary in another language
Ask questions to this video
Chat for follow-up questions, clarifications, and source-backed answers.