As for the less than 1% stuff, I think the sample size may be too small (4 only, 2-up, 2-down) to draw a conclusive pattern.