Implicit conversion from float -12.870813397129197 to int loses precision