Implicit conversion from float -37.431192660550465 to int loses precision