Implicit conversion from float -15.105105105105096 to int loses precision