Implicit conversion from float -0.4903417533432446 to int loses precision