I actually tracked dusk shield once upon a time with over 30 rounds of attacking with multiple attackers. I think it was somewhere in the 80-90 attacks total in those rounds (the sheet I tracked it with got thrown away some time ago). But to the point, the percentage was about 60% miss, though it was a small sample to test on. Ultimately I forewent anymore testing because I believe over time the hit/miss averages out in the run.
It's the long stretches of misses then one big hitting round that frustrates me. Not being able to predict the hit/miss is the most annoying part of the DS in my opinion. Personally, I'd rather have a shield that simply cuts damage in half so I can at least predict whether or not something I am about to do, or am doing, will be successful or not.