Implicit conversion from float -20.13011152416358 to int loses precision