In the readme position_history is described as
Previous positions for this player in the round, sampled every 5 seconds (but not in freezetime)
A few samples in the dataset have a position_history with more than 50 values, which would imply a round time of more than 4 minutes... How is this possible?
Edit: There are 425 samples with round_status_time_left=-1, will we have to predict rounds that have already finished, or does -1 mean something else?
In the readme position_history is described as
Previous positions for this player in the round, sampled every 5 seconds (but not in freezetime)
A few samples in the dataset have a position_history with more than 50 values, which would imply a round time of more than 4 minutes... How is this possible?Edit: There are 425 samples with
round_status_time_left=-1
, will we have to predict rounds that have already finished, or does -1 mean something else?