Implicit conversion from float -100.74074074074073 to int loses precision