Chomp is a game where two players attempt to eat peices of a chocolate bar without being the one to eat the "poison square". The rules are as follows: We start with a rectangular grid, and players alternate turns. On each turn, a player chooses on tile on the grid and removes all tiles that are at least as high and at least as far right as it. You can only choose tiles that haven't been removed yet. The player that has to pick the last tile (the bottom-left tile) loses.

An example game is here:

Player 1 removes chooses (5,3) on the first turn. Then Player 2 chooses (2,4), Player 1 chooses (2,1), and Player 2 chooses (1,2), winning the game.

I found a place where I can play a game of 4x7 Chomp against a computer and input a starting sequence of moves so that I can effectively save a game. I will have Manifold play as Player 1 and the computer as Player 2. This means that Manifold can win if they play right, since Player 1 always has the winning strategy in Chomp (except on a 1x1 board).

The moves will be chosen as follows: I'll create free-response markets asking which tile Manifold should choose for the next move, and the top three options at close will be the contenders. I'll then create three conditional markets asking whether Manifold will win, given that each move is chosen, and choose the move with the highest average probability (i.e., the one that Manifold thinks is best). I'll then see how the computer responds to that move and continue until the game has ended.

Thanks for making these markets. This was a fun game to bet on.

Chomp is perhaps not the best game for the Manifold Plays series, since there are quite a few moves where one player is already known to win quite easily.

Now that the market is over, let me share my thoughts on how easy it is to game these markets. I don't know to what extent you tried to avoid this, but it is incredibly easy to make the players lose and make some mana in the process. All you need is to make some last-minute bets (and quite some mana) So as someone betting for Manifold to win, I was constantly at risk of losing mana.

For all the "what move should Manifold make next" markets, you can just buy up 3 losing moves to 26%. Once you do that, you are guaranteed to make mana on that market, since only those three moves will be eligible to be picked as the actual next move. Anyone betting on the winning move(s) will lose their mana. It was incredibly risky that I had some large limit orders on some of these markets.

The "If Manifold plays X, will we win" markets can also be manipulated, but less profitably so. If you buy a lot of YES on a losing move, you will lose that mana. However, you can buy a lot of NO on all winning moves, and a tiny bit of YES on a losing move so that it gets the highest probability. This way you barely lose any mana (since the winning move will be resolved N/A) and can still make Manifold lose.

If your goal was to make it impossible to game these markets, here are some suggestions:

Add all moves with a significant amount of YES mana bought as options to bet on in the secondary market (e.g. you can vote on the top 3 moves, and all moves with at least 1000 YES shares bought)

Possibly: extend the close date if there was a last-minute big swing in market probabilities (e.g. if an option is ahead at close that was >20% behind in the last 12 hours, extend the close date by 1 day).

That said, these markets were fun to participate in. Thanks again for making them!

@FlorisvanDoorn Yeah, I tried improving it by making a bounty, but you're the only one who has actually claimed it, so idk how to get more players.