Implicit conversion from float -9.736842105263165 to int loses precision