Implicit conversion from float -51.21019108280254 to int loses precision