Just imagine drawing a stick man battle and be able to hook controls up to it to play it... thats what it could do.
But that's not quite it, is it? If I understand correctly you can't just draw a stick man battle from start to finish and then have it become playable, you would have to draw every possible permutation of different decision/outcome allowed in the game, unless I'm missing something?
You would therefore have an absolutely HUGE overhead in both time and effort needed to create your content as another major disadvantage to go along with the huge file-size to store the data. To create a completely non-interactive stick man battle lasting one minute you would have to create one minute worth of animation -- to add just one decision in the middle of the battle you need to create one minute and thirty seconds worth of animation -- that's an absolutely MASSIVE increase.
...but I'm still not seeing any actual advantages? What advantages do you think this technique could offer? A working prototype might help to show what you're thinking of, but even theoretically I'm just not seeing any upsides to the idea based on your present explanation.