Implicit conversion from float -67.69230769230768 to int loses precision