AI

The Warzone AI's behavior is determined by a set of algorithms. The source code behind these algorithms can be viewed on GitHub at https://github.com/FizzerWL/WarLight.AI.

Weighted Random

The AI constructs a list of actions it could take on each turn, along with a value of how strongly it feels it should do that action. It then does a weighted random against that list. This means that the AI is never perfectly predictable, but it does have tendencies.

General behaviour of the AI

Confirmed:

• The only cards it can play are the reinforcement card, sanctions card, bomb card, blockade card, and diplomacy card.
• It prefers to expand as long as it still sees neutral bonuses to expand to, as long as the bonus meets its criteria for being a good bonus. ^[1]

Behaviour of the AI in team games

• It will only attack teammates in two circumstances:

1. If a bonus is entirely controlled by teammates and it has >= 50% of the territories. It will only do attacks of 2 then.
2. If it's expanding into a new bonus that no teammates control. ^[2]

• The AI will not play cards when teamed with a human teammate unless it must play cards by rule. The only circumstance where an AI must play cards when teamed with a human teammate is if the last player on a team turns into an AI via boot or surrender and cards must be used by that team that turn.

Playing Itself

Under the single-player tab, you can set up a game full of AIs, no fog and watch the entire game play out on its own.

However, under the multi-player tab, the AI will not play a game without any humans involved. If only AIs remain in a game, the AIs will simply vote to end. If voting to end is not possible, such as in a tournament game, the AIs will just pick a winner randomly. This restriction is necessary due to the way that multi-player games work. Due to Warzone's flexible engine, someone could construct a game that took thousands or millions of turns to finish. For example, a game with no bonuses or large wastelands separating players. The Warzone server would strain to complete these games, likely causing the entire site to grind to a halt.