While in any CCG there is some amount of luck with regard to draws, I do not believe that cards like this should be in the game, where it's entirely based on RNG percentages. I do believe that it actually does block 50% of the time (I'll go do a study on this later), but this is on average. This means that, by the definition of average, there will be games where it blocks less than 50% and there will be games where it blocks more.
I suggest that this shield blocks 50% of every attack, period, instead of having a 50-50 chance at blocking it all. In this way, the expected value does not change, but the damage can at least be predicted.