Implicit conversion from float -81.96652719665272 to int loses precision