Implicit conversion from float -19.644670050761434 to int loses precision