The paper describes some possible ways how to improve Olympiads in Informatics. Tasks in Olympiads are small models of programming tasks in software industry and in the limited amount of competition time contestants need to complete several software production phases - coding, testing and debugging. Currently, only coding effort is adequately graded, but grading of other activities may be improved. Ways to involve contestants in overall testing process are investigated and ways to improve solution debugging process are described. Possible scoring schemas are discussed. In International Olympiads tasks with real numbers are quite rare. Possible reasons are investigated and a way how to return such tasks back to competition arena is suggested.