Implicit conversion from float -44.0956340956341 to int loses precision