Single player mode (AI/Solo Games)

This has been a long standing request by the community. Here are some suggestions submitted to tcgone.uservoice.com:

image

image

image

image

image

image

image

2018-05-29:

So now we have an official forum, this topic will serve as basis for single player mode. I do not have any timeline yet. I am thinking the best and quicker way to implement this in a basic setup would be to implement a deck based AI that needs to be programmed for every type of deck. I’d appreciate your inputs on this matter.

Consider this a side project for the future when your not focusing on fixing all the bugs and things. Plus Forbidden Light.
If you have some time on your hand, then maybe have a go at this.

I would love to see how an AI bot would play against a human. and have several questions for it. such as:

how will it know the best way to use trainer cards. like if it has professor oak, to use all the other cards in hand first like energy removals, gusts, switches (if it wants to retreat), potions etc…
how will it know that attaching the wrong kind of energy won’t allow it to attack.
how will it know when to retreat, and which attack to use to deal more damage.

for the first basic AI program, i think it should run single colour decks, so it doesn’t get confused with the energy types. and limit the trainer cards to say around 20. about the average for a theme deck.
Maybe we can incorporate some of the in game decks from the gameboy colour version of the pokemon tcg?

2 Likes

Necessarily the AI would need to run on specific deck. The effect of some kind of cards are too complex for being interpreted properly by an AI.
Each deck should come with a well defined simple strategy, mostly setting up the fastest possible and spamming damages. Some mono attacker deck would be ideal.
I think the cards and moves/ability should have some condition and conditional priority. As said previously, professor oak like cards should be used only when there is no other choices.
Multicolored deck shouldn’t be that much an issue as soon as the attachment is prioritized (favorite the attacker, the healthier, the one with energy already attached…).
An additional possibility would be to let the AI cheat, a bit like most of the existing AIs. Like drawing the right card at the right moment - some drawing items if the hand is empty - or biased coin toss for example.

I thought of a cheating AI that plays according to a preset plan and always has the necessary cards in hand (it will silently auto add cards to its hand). I believe this’ll be the easiest way to implement. The strength level of each AI will be determined by the pure strength of the deck they play. So for a base set blastoise deck, the AI will have some squirtle, breeder, blastoise and water energies in hand to get into the full assault mode as soon as possible. The power level can be somewhat eased by changing the turn number they’ll be ready and limiting number of actions in a turn.

In Pokemon trading card game 2, (The japanese only game on the gameboy colour) the ai, was programmed to always say “No” if you ever played the card “Challenge!”

Thereby every deck essentially had 8 Bill cards XD

1 Like

Indeed it is very hard to program how to respond to each effect that requires an input from the ai. It is also not feasible to simulate all possible moves like in chess because there are millions of possible moves.

Although I agree that an AI to play against would be much appreciated, I believe that could be a side project for the future, as DKQuagmire said. Right now, the most urgent thing to be implemented I believe is the inclusion of new expansions and the elimination of the few bugs throughout this amazing site. <3

tl;dr > same opinion as DKQuagmire hahaha