Implicit conversion from float -0.19736842105263008 to int loses precision