Implicit conversion from float -0.5462184873949667 to int loses precision