Just ignore these PTR pseudo-stats. A better tool is for example Leakbuster with HEM. And even then, you need quite a large sample to have meaningful stats on the turn and river.