Implicit conversion from float -12.272727272727266 to int loses precision