In many multiplayer online games, players vote on which map to play before a match begins, and the game system selects a final map based on all players' choices. This competition provides a large dataset of game sessions from a fictional online game, where each session records 8 players' map selections (from 16 available maps, numbered 0–15) along with the timestamps of their votes, as well as the timestamp of the voting window. Your task is to predict the final map selected by the system for each session. The dataset is synthetically generated with realistic game mechanics, and the evaluation metric is Accuracy. Good luck! 🎮
In this competition, you are tasked with predicting which map a game system will ultimately select for a match, based on the voting behavior of 8 players. Each player chooses from 16 available maps (numbered 0 to 15), and each vote is accompanied by a precise timestamp. The dataset includes the timestamp of when the voting window opened, the timestamp of each player's vote, and each player's map choice.
Your objective is to build a model that learns the mapping from these inputs to the system's final selection. The training data contains the ground-truth target labels, while the test data does not. This is a classification task with 16 possible classes, evaluated using Accuracy.
The data is synthetically generated and designed to be both realistic and challenging. Success in this competition requires careful feature engineering, particularly around the temporal and ordering aspects of the data. Good luck! 🎮
Submissions are evaluated on Accuracy between the predicted map and the true target.
For each id in the test set, you must predict the final map selected by the system. The file should contain a header and have the following format:
id,target
0,3
1,7
2,3
3,12
id: The session identifier
target: Your predicted map (integer between 0 and 15)
The file must contain exactly one row for each test sample, with no missing values.