Implicit conversion from float -12.726063829787222 to int loses precision