MArcomage

Just had an idea about a game mode/option where you get a warning if the opponent could win next turn, similar to a "check" in chess.

For non-hidden games this would be mainly a convenience, but for hidden games it could alter gameplay a lot. I definitely don't want to see the current system done away with, but maybe something like this as an option? It would lower the occurrence of those "cheap shots" like a sudden Orc Regiment/Pyramids/Duplication.

This might be difficult to code seeing as each card might have different options and some can work randomly, and even discarding would need to be checked for a win condition.

Consider Chaos dwarf. This one could multiply production and give a resource victory, or it might not. Would the system call this a check?

Additionally, Swift and Quick would render this useless. It's possible that your opponent could draw a card and win on the next turn without you ever being warned.

Chess is a deterministic game. MArcomage isn't. Even the play preview functionality that we currently have is only hypothetical and it is only guessing the impact of card which you are going to play. The Check functionality would have to do much more guessing than this. It is probably possible to implement a hypothetical "Check", however it would be highly unreliable. I don't think this is worth the effort.

Fair enough. I didn't really intend to equate MArcomage to chess, but the point about random chances is a very valid one. Maybe it's not the best idea for this game. Still, good to be thinking of these things, eh?

It's not a bad idea at all, no. I think I recall someone a long while ago suggesting a similar feature where a card capable of causing a destruction victory for your opponent would be displayed with a red border or something as a warning to the defending player. Not too far off from your proposal, though simpler and something you should be noticing anyway if you're paying attention.

Anyone who's played me a lot in the past knows that I generally thank my opponent for the game in chat when I spot that either of us has sealed the win. Occasionally my opponent doesn't play the right card, though, and the game goes on. I wonder how many losses I could have avoided by not doing that, but I don't mind pointing out a win here and there if the player is inexperienced.

Mojko wrote:
Chess is a deterministic game. MArcomage isn't. Even the play preview functionality that we currently have is only hypothetical and it is only guessing the impact of card which you are going to play. The Check functionality would have to do much more guessing than this. It is probably possible to implement a hypothetical "Check", however it would be highly unreliable. I don't think this is worth the effort.

Hm, maybe Marcomage should be ported to a nondeterministic Turing machine?

dindon wrote:
Hm, maybe Marcomage should be ported to a nondeterministic Turing machine?

That would be interesting, however the hardware limitation (like finite memory) constraints the model to finite automaton :(

DPsycho wrote:
I think I recall someone a long while ago suggesting a similar feature where a card capable of causing a destruction victory for your opponent would be displayed with a red border or something as a warning to the defending player.

You could make it so one that could guarantee your defeat has a red border, those where it's not certain, either due to random chance or lack of opponent's knowledge (how rare is the card opposite his Mirror Golem?) have a yellow border.

But yeah, I understand it's hardly on the top of anyone's list of priorities. regardless :)

Spoon wrote:
You could make it so one that could guarantee your defeat has a red border, those where it's not certain, either due to random chance or lack of opponent's knowledge (how rare is the card opposite his Mirror Golem?) have a yellow border.

But yeah, I understand it's hardly on the top of anyone's list of priorities. regardless :)

While this seems reasonable, I think it would actually require a huge amount of work. It would require the system to in some way "understand" how different cards work, which it doesn't. The preview function basically works by doing a dry run of the next turn playing each card. It doesn't know why Mirror Golem is doing 7 damage, it just pretends to play the card and sees what happens. The system doesn't have each card's damage range explicitly encoded anywhere.

Yes, this is the source of the problem. The system doesn't understand cards at all. To decide if the danger assessment is reliable or not is equally difficult to do as to make a reliable assessment. It's kind of small miracle that the AI has about 43% win rate (AI challenges excluded), because the AI reasoning is built on preview functionality. When I added the AI my expectations were about 20% win rate.

you have to consider that the AI is a perfect opponent to test new decks, so no wonder if it have many wins.
(anyway, I am aware that this AI does not really exist: it's just Mojko and the Ultramage pretending to be a bot)

Fithz wrote:
(anyway, I am aware that this AI does not really exist: it's just Mojko and the Ultramage pretending to be a bot)

You got me there :D