Implicit conversion from float -7.368421052631575 to int loses precision