this reply is explaining some of the stuff I have edited or feel needs to be expanded independent of the original post
expansion: my method involves the pred in an empty state, so if you get swallowed with paper and survive the first attempt my data requires her to digest the paper then you get eaten with another piece of paper to test second escape, ect for paper. Under the Apples I made sure that the pred was not under the effect of another apple, so I DID NOT rely on escape 1 being an apple, then escape 2 being another apple ect, I SPECIFICALLY used non-apple escapes to do this and this part is by far the largest part of why I took a bit as if for every 1st escape with paper I have a 50% chance of escaping I now need to do 100 apples with that coin toss, which makes 3rd escape closer to 15% chance after normalizing data but if you want to do it I am pretty sure apples lose their effect when the pred burps them up so if you are willing to wait and stalk them for that make a note of it in your results. I also forgot to mention the margin of error in everything is around 3% so I considered Bella and Satie's results to be the same despite them having a small percentage change due to RNG noise. I also did not include my original/new hard numbers numbers as it was already getting pretty long and this was directed more at the developer who has their actual RNG number chance known in their coding so the actual hard numbers were not a factor to my intent, but I would like to see everyone's hard numbers if they wish to perform this experiment to see a larger data point on how RNG percent affects gameplay chance whereas my data points are about version trends since I did this with the old version as well. This means that in total, you need to do at least 5040 games which is...yeah, it's not fun after a while, especially the hunt for those small red and crazy apples
edits: fixed some numerical errors on second readthrough and added some additional context to crazy apple results