Implicit conversion from float -31.746575342465746 to int loses precision