Yep, statistics and probabilities are great when looking at a basket of events. But each event is a binary result, it either works, or it doesn't.

drugs are unique. The failure of one proves NOTHING about whether another is likely to fail. -biomanbabba

So the failure rate could be 80% (just to pick an example)

So you have a new drug and think, 'well this has only 20% chance of win' wrong. It either wins or loses. If it is a winner, it is a 100% probability.

We just don't know in advance so we look at these probabilities as comps.

But like you say, it is apples and oranges with the actual event which is either 100% win or 100% fail.

I once was playing poker and the lady to my left was upset and incredulous that her aces had been beaten. But I had a 80% chance, how could I lose? (I forget the actual number)

I said nope, you had a 100% chance of losing, you just couldn't know in advance. She didn't understand and insisted I was wrong.

With that deck's shuffle and those players in, she was always going to be beaten by the flush. 100% sure. So the probabilities only work out over thousands, sometimes tens of thousands of hands. But each event is a binary result.