[Solo-Mode TFTCG] Building an AI System

In the previous article about this quest, I laid out some of my initial thoughts and inspirations behind trying to create a solo mode for TFTCG. I also blatantly lied to you and told you that I’d follow up within a week or two.

Now that it’s two months later, I finally have a moment to continue my thoughts from before. And when I say that, I really mean that I have an hour long break between meetings, and this is how I’m going to decompress a little bit.

As a reminder, my two main points of inspiration come from the Lord of the Rings LCG and the Marvel Champions LCG, both by Fantasy Flight Games. In each, there’s an Encounter card deck that serves as the driving force that antagonizes the player(s).

In LOTR, the encounter deck spits out enemies that threaten the health of player characters and allies, locations that generally make it more difficult to put progress on the current stage of the quest (progressing through all quest stages is the win condition), and treachery cards that serve as one time events that can hamper players in any number of ways. On top of that, many enemies have Surge abilities, which are activated by flipping over the top card of the encounter deck to check if that top card has the Surge keyword. This mechanic adds an additional degree of randomness that can cause the board state for players to cascade further into greater degrees of difficulty. On top of that, it helps to cycle through the Encounter deck, which can cause discarded cards to be reshuffled back into the deck if it ever runs out, forcing you to once again overcome the difficulties that you’ve already beaten (or avoided).

In Marvel Champions, the win condition is to defeat all stages of the main villain. Along the way, you need to contest with schemes that begin in play or which enter play later, always designed to make the player’s life more difficult. The Encounter deck can spit out additional minions, side schemes, or attachments to power up the villain, along with Treachery cards which function in a similar way to LOTR. On top of that, if the Encounter deck ever runs out and needs to be reshuffled, the main scheme gets an acceleration token which causes additional threat to be placed upon it, hastening a possible game loss if the main scheme is ever completed.

For a solo mode for TFTCG, I wanted to borrow some of these concepts, without making any major changes to the framework of the game as it currently exists. At the same time, I also needed a way to determine the order in which AI characters would attack and how their targeting would work. Above all else, I wanted the initial version of the AI control concept to be fairly basic, while also having the ability to build upon it in the future. Here’s what I came up with:

In addition to a battle card deck, the AI player will also have an AI Behavior Deck, which will determine the order in which its characters will activate, and who they’ll target with their attack. The very first thing that will happen on the AI turn is that it will reveal the top card of the AI Behavior deck. The AI will then proceed from top to bottom through the steps on the revealed AI card (the final versions of these cards will likely be numbered 1, 2, 3 as opposed to using bullet points in order to make this crystal clear).

I’m also attempting to streamline any additional rules text that might be necessary, hence why the AI cards also instruct the player to flip the indicated character if no other character has been flipped by the AI this turn, and to reveal and play a battle card if the AI has not yet played one this turn. In this way, it will mimic a typical game, where if the AI’s opponent is totally tapped out, they will only get 1 flip and 1 battle card. There will be another mechanism that will allow the AI to play additional battle cards as the game progresses, but I’ll talk more about that in the next article (which hopefully won’t take another two months to write). In the initial version of this scenario, I’m keeping the AI to one battle card played during its turn in order to simplify things. As the scenario ramps up, the AI will have the opportunity to play more.

One of the other elements that I wanted to pull in from the LCG games was the fact that the Encounter deck is simultaneously predictable and unpredictable. Especially after you’ve played a scenario once, you know what’s possible and what might be coming your way. You don’t know, however, exactly when certain effects will be revealed and how they might be combined together. In this way, if your first attempt to win an encounter is unfruitful, you can change your strategy and possibly your deck to make another attempt.

I wanted the AI deck and the AI targeting to mimic this. Much like the battle deck, the AI Behavior Deck will be shuffled before the game. In this case, it will randomize the order in which the Predacons will activate and attack. Once all characters on the board become tapped and subsequently untap, the AI Behavior Deck will be reshuffled in order to prepare for the next “round.”

I wanted the targeting to also be both predictable and unpredictable. After you’ve played against it once, you know how each different Predacon prioritizes its target. In the above case, Razorclaw will attempt to attack the available target with the highest star count. In the event of a tie, it will then go for the available target with the highest ATK value. In the event that there is still a tie, it will go for the available character with the highest HEALTH value. In the unlikely event there is STILL a tie, the target is randomly determined (roll a die, flip a coin, etc.).

I also wanted this to highlight each Predacon’s personality. In Razorclaw’s case, he’s a consummate hunter. He wants to go after the target with the highest prestige. Star count becomes the representation of that, followed by how dangerous the opponent is (ATK value), and its overall vitality (HEALTH).

Rampage is described as being totally unpredictable, launching into a frenzy of uncontrollable action at any moment. In this case, none of your vulnerable characters will be safe, as his target will be randomly determined.

A scavenger by nature, Divebomb prefers to attack the weakest targets available.

Stubborn and seemingly undaunted by the most challenging opponents, Headstrong goes after the toughest available target on the field.

Tantrum (suck on that, expired copyright!) is vindictive, preferring to attack the available target who attacked last.

As I’ve mentioned above, I consider this to be a fairly simplistic initial version of an AI Behavior deck. In future versions and in other encounters, I think it could do more than just tell the player what order to activate the AI characters and how they’ll prioritize targeting. For example, it could instruct the player to pull in additional enemies from outside the game. It could also instruct them to untap a character that has already activated and attack with them again. Against a single “boss” AI character, it could automate them through certain movesets, similar to bosses in games like Dark Souls. Upon reaching a certain threshold of damage, the AI deck could be swapped out for another one, making the next stage more unpredictable.

It could also be used to introduce secondary win or loss conditions (I previously gave an example of a Combaticon AI encounter having a secondary condition involving the Combaticons trying to stockpile weapons from their battle deck, or perhaps even from the player’s battle deck).

By this point, you’ve likely noticed that the art on the AI cards is using different Transformers Legends art than the Predacons that we received in Rise of the Combiners. In order to both make them a challenging opponent and to automate their character card abilities, I’ve tweaked or in some cases completely changed them in terms of both stats and character card abilities. In the next article, I’ll go into more detail about that process, while also explaining the battle deck that this specific AI encounter will use. In that article or the next, I’ll also explain how the endgame will be attempting to escalate in order to make things more dangerous for the player. Yes, part of that will include combining into Predaking, because duh, of course.

Until then, stay healthy and sane. Ish.

As always, keep your eyes on the Wreck ‘n Rule YouTube channel for our usual shenanigans. Also consider supporting us on Patreon, since the proceeds go toward helping us upgrade our recording set-up! Plus, we now have merch available too!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s