I think what's really interesting about this formulation is that after one player sets it up, every step of the resolution of the rules for the simulated Turing Machine is forced for both players.
The average number of turns in an online game is less than 7 turns. I would further posit that if you constrained possible actions to decisions that good players are likely to make, the game is no longer Turing complete.
You could make a similar claim about a game where any player can win at any time if they choose, but can alternatively do anything else. It’s not computable, but it is irrelevant to the complexity.
You could probably also argue that a Turing intending player is still computable if playing against someone with a defined strategy. Because they will be halted by the other player winning.