Implicit conversion from float -2.6086956521739353 to int loses precision