Implicit conversion from float -6.122448979591837 to int loses precision