Most memcoins created on the Pump Fun platform were rugged before they could graduate to Raydium—and now to PumpSwap. Using information from just the first few transactions after token creation, it's possible to pinpoint tokens with a higher likelihood of being rugged.
This track challenges participants to predict whether a newly launched Pump Fun token on Solana will graduate—based on data available in the first 100 blocks post-mint. In practice, this means distinguishing tokens with a higher risk of being rugged from those more likely to succeed.
You will receive a prepared dataset including:
The objective is to leverage machine learning and blockchain analytics to generate a binary prediction:
Will this token graduate? (i.e., Will its liquidity reach at least 85 SOL?)
This task is designed to attract machine learning practitioners eager to apply their skills to real-world blockchain data, and to engage blockchain analytics experts interested in exploring machine learning methods. The goal is to foster cross-domain collaboration and innovation. It is also intended to be accessible to beginners in ML, while still offering depth for advanced participants. By learning what data matters and how to model it, success here could unlock more complex challenges—like real-time rug pull prediction.
This might not be obvious from the task description, but success also depends on participants’ ability to identify serial token deployers and snipers—actors who typically try to avoid detection but have a major impact on whether a token thrives or gets rugged.
https://www.kaggle.com/competitions/solana-skill-sprint-memcoin-graduation
To receive the prize, you should provide a reproducible solution that includes: