ML Competition Platforms Guide

Kaggle — kaggle.com/competitions

Best for: Beginners AND experts. Biggest community, most resources.
Prize range: $0–$1,000,000+
Rank system: Novice → Contributor → Expert → Master → Grandmaster
Tips: Start with "Getting Started" competitions (Titanic, House Prices). Read Discussion tab obsessively.
How to win: Feature engineering + ensembling + CV discipline

Best for: Social-impact ML (health, climate, NLP for good)
Prize range: $5,000–$50,000 typically
Competition count: ~$4.97M total prizes given out
Tips: Less competitors = better odds. Partners: NASA, Microsoft, World Bank.

Best for: Stock market prediction. Runs a real hedge fund using your models.
Format: Weekly rounds. Submit predictions every Saturday.
Reward: Stake NMR (crypto token) on your model. Positive score = more NMR. Negative = burned.
Key fact: Data is fully obfuscated — you don't need to know finance.
Why enter: Only competition where your model runs in production in a real hedge fund.
Tips: Start without staking to learn the format. Use tree-based models. Read era-boosting techniques.

Best for: Marathon Matches (long-form optimization problems) + AI/ML challenges
Format: Single Round Matches (competitive coding) + Marathon Matches (ML/optimization)
Tips: Top Marathon Match performers get invited to TCO (Topcoder Open) Finals in USA
Prize range: $100–$100,000+

Best for: AGI research. The hardest ML competition in the world.
2026 prize pool: $2,000,000
Tracks:
- ARC-AGI-2: $1M (open source your solution to ARC reasoning benchmark)
- ARC-AGI-3: $75K in milestone prizes (build agents that play ARC games)
Hosted on: Kaggle
Key dates 2026: Jun 30 Milestone 1, Sep 30 Milestone 2, Nov 2 final submission
Tips: Requires open-sourcing your solution. Even partial progress = publications + recognition.