The cost function currently evaluates the error in seconds between actual vs predicted. Instead this should be changed to use the relative error between actual and predicted.
Using the absolute levels overweights the longer races at the expense of the shorter ones.
The cost function currently evaluates the error in seconds between actual vs predicted. Instead this should be changed to use the relative error between actual and predicted.
Using the absolute levels overweights the longer races at the expense of the shorter ones.