This is a solution for the The Winton Stock Market Challenge Kaggle competition.
Since predicting stock market returns is notoriously hard, we will expect it to be difficult to beat a naive baseline model of just predicting mean returns of the training period. Also, since this is a hiring competition we will suspect the data set has been created in such a way, that dealing with important factors such as missing value imputation, numerical vs categorical features, colinearity, normalization, grouping and model robustness will be essential for competitors to demonstrate adequate skillsets in. Also, we note that we are here to win a Kaggle competition - not to make an actually viable trading algorithm.
There where 832 teams participating in the Kaggle competition. This solution has a Private Score of 1727.86647 and a Public Score of 1769.92199 and would have ranked as number 4 (top 0.5%) on the Private Leaderboard.