First congrats to the winners👏👏👏. It was a great competition, congrats to my teammate also @koleshjr.

Our approach is:-

• First, we build an action prediction model, which will add 7 features related to goals scored and goals conceded to the original Train.csv and Test.csv. These features are the key featues for our model. here is the approach for this prediction model link.
• Second, we use the modified data from the above notebook and the main game statistics files to generate more features like pass accuracy, total shots, team attacking and defense strength from the previous seasons of goals scored and goals conceded. Then train a catboost and lightgbm models on 10 StratifiedKFold.

The final was great 0.216 on the private leaderboard and 0.17 on the public leaderboard yet we didn't choose it😔.

Please upvote the notebook, if you find it useful.