That's not really true. The selection of computational geometry problems on the ICPC tends to be limited because testing and implementing numerical algorithms

That's not really true. The selection of computational geometry problems on the ICPC tends to be limited because testing and implementing numerical algorithms is really difficult in a competition environment. When you do see geometry algorithms they tend to be "trick" problems or else don't use any really deep or hard results.