Implicit conversion from float -36.20689655172413 to int loses precision