ITMO Quants Challenge

Predict the next market state vector from historical sequences. Build a model, submit your solution, and compete on the leaderboard.

Get Started

Sign in to submit your solutions and track your progress on the leaderboard.

1. Download

Get the competition package with training data, utils, and an example solution.

2. Build

Implement your PredictionModel in solution.py. Use LSTMs, Transformers, or any approach you like.

3. Submit

Upload your solution as a .zip file. It will be evaluated in a sandboxed environment and scored by R².

Competition Rules

  • Submit a .zip containing solution.py at its root
  • Your PredictionModel.predict() receives DataPoints sequentially
  • Return None when need_prediction is False, otherwise return an array of N features
  • Evaluation runs in a sandbox: 6 CPUs, 16GB RAM, no network, 10-minute timeout
  • Scored by mean R² across all features
  • Select up to 2 submissions for the finale before the deadline
  • After the finale deadline, no new submissions are accepted