MITSUI&CO. Commodity Prediction Challenge

Overview

This competition challenges you to develop models that predict future commodity returns using historical data from various markets like LME, JPX, US Stocks, and Forex. The goal is to create stable, long-term forecasts that can optimize trading strategies and manage risk.

Requirements

You need to build models that predict price-difference series derived from historical data. Submissions must be made through Kaggle Notebooks with a run-time limit of 8 hours for CPU and GPU during the training phase, extended to 9 hours for the forecasting phase. Internet access must be disabled. The metric used is a variant of the Sharpe ratio based on Spearman rank correlation.

Prizes

The total prize pool is $100,000. The top three teams will receive $20,000, $10,000, and $10,000 respectively. Teams placing 4th through 15th will each receive $5,000. The competition also awards Kaggle points and medals.