My nephew thought of the following problem over the holidays:
I have 100 coins, half these coins are currently face up heads, the other half are currently face up tails. Currently, my score is 0. Each coin has a 0.5 probability of landing on heads and a 0.5 probability of landing on tails
Round 1: I randomly generate an integer "n" between 0 and 100, select "n" coins and flip them - naturally, some change sides, some don't. the coins that weren't selected stay as they are.
After Round 1 - each coins that is now facing the opposite side gives me 1 point. And each coin that is facing the same way takes away 1 point. Record the total number of points.
Round 2: I now again generate another integer "n" and flip these "n" coins. For coins that were not previously selected but were now selected if they land on the other side, I get 1 point but if they land on the same side, I lose 1 point. And all coins that were not selected in Round 2 automatically deduct 1 point each. Record total points from Round 2 and add to total points from Round 1.
Repeat for many rounds
Now, my nephew's game lead me to the following questions:
- After Round 2, what is my expected score?
- How many Rounds do I need to play to get some expected score of "x"?
- On average, what is the optimal number of rounds I should play to give me the highest possible score (before my score starts to decrease and decrease)?
My feeling is that initially it it makes sense to play this game for a few rounds regardless - but then, based on your score, there is probably an probabilistic optimal number of rounds to stop playing this game.
I think that maybe you can use a Markov Chain in where states are either the "score of the game" (e.g. probability of transitioning to -5 points given you are at -1 point) and then estimate the "time to absorption" for different questions based on simulations or solving this analytically - but I am not sure.
Any ideas?