After a recent discussion I had on the /r/4XGaming subreddit concerning the Stellaris AI, I’d like to elaborate on the role of the four X’s in the development of a good game AI.
As a nascent game developer, one thing I’ve learned is that there’s quite a bit of difference in difficulty between the implementation of an AI to handle each of the four X’s: Explore, Expand, Exploit and Exterminate. Most notably, the first three X’s have been fairly straightforward to model while the fourth X is more challenging. And by “model”, I mean making the AI competent enough so that it can keep up with most players, even the good ones.
There’s a good reason for this, I think. The first three X’s are essentially the “AI vs. Randomized Map” or “AI vs. Game Mechanics”, which gives the AI a relatively stationary target at which to shoot. Once the developer understands how to explore a randomized map, how to prioritize the best colonization locations, and how to efficiently spend resources to develop your colony, then it becomes a fairly straightforward project to write an AI to do those things.
However, it’s also important to remember that the AI perform the first three X’s in a fashion repeatable by the player. The AI should never exploit game mechanics through some form of tedious “micro-management” that would be onerous for the player to reproduce. In other words, if the player wants to micro-manage the game in order to get an edge on the AI, he should be allowed to do so. But he should never feel forced to micro-manage in order to keep up, because that will suck the enjoyment out of the game.
A quick diversion on micro-management
A quick definition: micro-management is the constant and repeated use of minor actions that, on aggregate, provides a player with some tangible advantage that is unavailable otherwise. They often exploit edge conditions in the game mechanics that less competitive players disregard as inconsequential and/or not worth the effort.
Although there has been a concerted effort within this game to remove known micro-management issues from the original MOO1 mechanics, I am under no illusion that the ability to micro-management will ever be eliminated. One thing I’ve realized is that micro-management is not a game problem, but a player problem. Let me elaborate…
The real question is not, “Does this game have micro in it?” but rather, “Do players choose to micro in this game?”. There are three ingredients required for micro to exist and, if these ingredients are present, then there will be players who micro-manage no matter how the game is designed. These ingredients are:
2. A Challenging Opponent
3. Player Desire to Win
Let’s start with Time. Turn-based games have this in spades. After all, you are allowed each turn to spend as much time as you need before pressing “Next Turn”. This enables micro-management and is also why you don’t see real-time games with micro issues.
A challenging opponent creates the necessity to micro. If an AI is easy, then there’s no point in spending all of the effort. This is why you see players resorting to micro-management almost exclusively when playing in Hard and Impossible modes, but not in the standard Easy modes.
The player’s desire to win is also important. Of course, all players want to win, but the reality is that most players are looking for a fun game. If the difficulty level makes it too hard to win, then they turn it down. It’s why the default difficulty level for most games is set on Easy. Hardcore players who want a challenge are the distinct minority and they will ratchet up the difficulty as they get better. By the way, that the majority of players are not hardcore comes from an old Explorminate forum thread in which Brad Wardell of Stardock noted that vast majority (90%) of Gal Civ players never play on anything but the default (Easy) difficulty level. They’re just playing for fun!
So therefore, my goal in Remnants is not too eliminate micro-management (an impossibility), but to create a game where it is never necessary unless players choose to play on high difficulty levels. In those cases, they are intentionally allowing the AI to cheat with bonuses, so they can micro their hearts out!
Back to the Fourth X
The fourth X, Extermination, at some point pits the AI against a human player, who is not a stationary target and is therefore much more challenging to compete against. As a result, it is absolutely crucial that the AI perform the first three X’s very well so that it is well-positioned to take on the player with the fourth X. Even if it can’t keep up and ultimately loses, it will at least have given its human opponent the satisfaction of defeating a well-armed opponent. There should never be a situation where the player captures a few enemy colonies only to discover that the AI is fraudulent and has been cheating to keep up. That would be very unsatisfying.
To me, the ideal situation is where the AI is equal or better than the player at the first three Xs — because it has patience to check and re-evaluate every system on every turn whereas most players do not. This is very possible and I am gaining confidence that the Remnants AI may already be at this point. But afterwards, the player needs to heroically come from behind and achieve his glorious and well-earned victory through better strategic and tactical planning. Of course, the reality is that this strategic and tactical expertise does not magically appear in Game 1… it’s developed over time as the player learns what tactics work better than others.
Ultimately, the fourth X belongs to the player.
Thanks for reading!
Discuss this article on the Remnants subreddit at:
Other game design discussions
- The Case for Limiting Complexity and Depth
- 9 Ways that 4X AIs disappoint us
- The Case for Limiting Ship Designs