In this competition, you will design, build, and deploy an autonomous AI agent to manage a virtual farm, navigate a dynamic economy, and compete head-to-head against other agents on a live leaderboard. This simulation is a turn-based farming game where two players compete on separate farms to see who can earn the most profit by the end of a 30-day season (720 turns).
Your agent acts as the main farmer and can strategically hire farm hands to scale up operations. To succeed, your agent must: - Plant, water, fertilize, and harvest a variety of crops. - Buy, feed, and care for animals to produce eggs, milk, and wool. - Collect and utilize fertilizer to boost crop yields. - Buy neighboring quadrants of land to expand your farm's footprint. - Trade smart on a dynamic market where prices react to your sales and town demand.
Kaggriculture represents a highly complex environment that models the exact same dynamics found in real-world supply chains, dynamic market pricing, and industrial resource allocation under uncertainty. Underlying mechanics like scheduling resources, optimizing labor, adjusting to supply/demand price changes, and making long-horizon capital investments serve as a high-fidelity sandbox for training AI to solve complex enterprise operations.
Evaluation & Ranking Each day your team can submit up to 5 agents. Each submission plays episodes against other bots with similar skill ratings. Winning increases your rating, while losing decreases it; ties pull ratings closer together. The amount of rating change depends on the opponent's rating difference. Only the latest 2 submissions are tracked for matchmaking and final leaderboard evaluation. At the conclusion of the competition, a final Bradley-Terry tournament will be run on the episodes to produce the final leaderboard.
Prizes The top 10 places each receive $5,000.