With brush tests you might easily end up hiring someone who excels at your tests and nothing else. Overfitting can be a problem even with subjects that don't actively try to adapt, and for Google they most certainly do (I read the "please take some months to prepare" suggestions as a clear indication that they want a level playing field wrt adaption, kind of cheap to leave that problem to be solved applicants)

In the long run you might end up with better results if you don't even try to identify the one optional choice and only focus on weeding out false positives, leaving the rest to a diceroll.

(I'd be surprised if what happens at Google wasn't a weird compromise Frankenstein of both)