Tic-Tac-Chess-Checkers 4 puzzle

Ok stupid question - in the rules it has three pieces for each player - a knight a pawn and a checker King - where do the plain checkers pieces come from, please?

Awesome game! This is my first time hearing of TTCC4.

alandau, thank you for the explanation, I was thinking that the Connect 4 pieces could be "dropped" meaning "placed" on any square in the middle three columns.

Without that Connect 4 dropping restriction, that just helps white/red win even more.

One possible victory: king moves left-down to b3(?). Then, black doesn't have a winning move. So, white/red king can capture c4 to win.

I hope I understood the rules correctly. I would definitely like GGP.

Actually I dislike this game. Between strong players it has a tendency to get into situations where both players repeat moves to an eventual draw too often. Since it doesn't terminate until move 100 in such cases it takes a LOT of time to play for relatively uninteresting draws quite often. It really needs a max-repetitions count, but encoding such a rule in GDL is hard.

Indeed in general it is impossible (or at least infeasible??) to encode 'a move that would result in a repeated state is illegal' (which is something you would ideally want to do in Go incidentally, at least with Chinese rules). Doing so requires reasoning about the state, which necessarily involves the addition of meta-state, which logically sits outside of the game rules. In principal that can be encoded into a metagame (i.e. - expand the GDL model by adding a meta-layer whose job it is to track the state of the inner-game to monitor repetitions) with expanded GDL. However, in practice GDL does not lend itself to such encodings (at least for propnets you'd need a ground proposition for every possible inner game state!).

As a possible addition to GDL an explicit maxStateRepeatCount specifier would be useful to address this (a player can easily monitor that with a list of antecedent states as it expands the tree, and using something like Zobrist hashes this can be made fairly efficient [subject to accepting errors at an arbitrarily low rat governed by the hash size anyway]).

In a slightly more limited possibility, you could consider a slightly more restricted primitive illegalRepetitionScope(maxTimes, numTurns) which makes moves illegal that would cause more than a certain number of state repeats within a certain number of turns (the turn scope limit means what needs to be stored can be reduced to a small window)