Implicit conversion from float -0.851063829787222 to int loses precision