Implicit conversion from float -92.46753246753246 to int loses precision