Implicit conversion from float -0.20408163265304324 to int loses precision