Implicit conversion from float -37.90262172284645 to int loses precision