Implicit conversion from float -0.9051724137930819 to int loses precision