How to Modify the Starter Bot in Halite II 2017 Artificial Intelligence Competition, Part 2

TL;DR

Improve the Halite II starter bot by studying losses, changing one or two behaviors at a time, and comparing the results. The demonstrated update sends ships toward unique planets instead of one shared target and removes the starter script’s max-speed division by 2. The bot begins at rank 22,346, giving these changes a concrete baseline for evaluation; read on for the exact modification and testing workflow.

Transcript

what's going on everybody welcome to part two of the highlight series in this video what rumi going over is how to go about improving your AI from this point so I'm gonna kind of just go through the pipeline process that I've been doing to incrementally improve my AI so first of all like I was saying in the previous video it's fun to watch the game... Read More

Key Insights

• 😉 Watching both wins and losses can provide valuable insights on gameplay and AI performance.
• 🚤 Directing ships to different planets and adjusting travel speed can significantly improve AI performance.
• 🥺 Attacking enemy ships can lead to the acquisition of new planets and increased production capabilities.
• 👾 Regular monitoring and adjustments are necessary to achieve better rankings in the game.
• ✋ Improvements in AI strategy can result in higher rankings and improved gameplay.
• 👾 Analyzing opponents' games can help identify common mistakes and strategies to avoid.
• 💱 Incremental changes are recommended to evaluate the impact of each modification.

Questions & Answers

Q: How do you improve the starter bot in Halite II 2017?

Review lost games to identify specific failures, then change only one or two related behaviors before evaluating the result. The demonstrated improvements send ships toward unique planets and stop limiting their travel speed to half of the maximum.

Q: Why does the speaker focus on games the bot lost?

Lost games make it easier to identify why the bot failed, such as choosing poor destinations or losing battles after becoming isolated. The speaker also studies wins to determine why the opponent lost and watches stronger players’ games for additional lessons.

Q: What problems are identified in the starter bot’s gameplay?

All the ships travel toward the same planet, pass other useful planets, and sometimes pursue a destination they cannot reach first. This leaves the fleet poorly positioned, contributes to lost battles, and can result in total elimination.

Q: Why should ships travel toward unique planets?

Sending every ship toward the same planet creates unnecessary concentration and causes the bot to ignore other planets along the route. The new logic skips a planet when it is already included in the planned-planets list, allowing another destination to be considered.

Q: How is the bot’s travel speed changed?

The starter script sets travel speed to the maximum speed divided by 2. The modification removes that division so ships no longer always travel at half speed, with the speaker noting that ships moving in different directions are less likely to run into one another.

Q: How does the planned-planets list work?

The code creates an empty list named planned_planets. If a planet is already in that list, the current ship skips it and continues to the next planet; once navigation is added to the command queue, that planet is appended to the list.

Q: How does the speaker manage different bot versions?

The speaker copies and pastes the old version, renames it as a versioned bot, and edits the new version. The submission bot still has to use the title-case name “MyBot.py,” and the command or batch file must be edited to run the intended versions.

Q: How are the bot modifications evaluated?

The speaker records the current rank of 22,346 before making the edits, then plans to observe how the rank changes. This incremental approach helps connect performance changes to the small set of modifications just introduced.

Summary & Key Takeaways

• The content discusses the process of improving AI in a game by analyzing wins and losses to identify issues and make necessary changes.

• The speaker emphasizes the importance of learning from both their own games and other players' games.

• The video demonstrates specific changes made to improve AI performance, such as directing ships to different planets and adjusting travel speed.