Implicit conversion from float -33.59550561797754 to int loses precision