Hacker News Reader

Comment by Cthulhu_ | Hacker News Reader

This is always a challenge in video games; the goal of enemy AI is not to beat you, but to make for an interesting and fun challenge.

I'm sure there's a community where they build alternative enemy computer players though. And else there's always online multiplayer, but that's not for everyone.

View on HN

The problem with AI 'cheating' is that it often shifts the challenge away from strategy and more toward figuring out how to outpace artificial handicaps. A well-designed AI should create challenge through better decision-making, not arbitrary advantages.

Exactly this. I'm not at all complaining about the game being too easy or too hard. I'm complaining that when you create a challenge by taking a horrific AI and then just letting it cheat to the point that it poses a challenge, you create an entirely new game. And that new game is often nowhere near as fun as the game itself "really" is because it all comes down to meta-gaming the horrific AI to overcome what should be an insurmountable handicap if the AI played even remotely decently.

> And else there's always online multiplayer, but that's not for everyone.

Civ's multiplayer is traditionally very poor (frequent crashes, desyncs etc), which is a big reason why there's such demand for good AI, moreso than other games.

I wish they would invest on the reliability of multiplayer... every time I hope it will be better and every time I am disappointed.

You are right, but the issue here is slightly more nuanced. The CIV AI is just plain atrocious at playing the game. It's not about having a personality, it's that the developers add game mechanics that the AI is fundamentally incapable of dealing with. To give you an example, CIV 5 introduced the concept of 1 unit per tile - it used to be that you could stack as many units as you wanted into one tile before. Neat idea, right? Terrain and unit placement suddenly mattered!

But there was a significant issue with this: the AI was terrible at moving units and armies. It'd get confused by the smallest congestion. The AI would build a large army, and then be incapable of moving it anywhere.

The AI can still have personality and roleyplay instead of attempting to win, and yet be a competent strategist and tactician.

No unit stacking even breaks the automated worker orders which are meant to help the player by reducing the amount of micro. Automated workers are borderline useless in Civ V compared to Civ II.

This isn't a case where the AI is meant to have "authored personality" or whatever.

That's a very good point as well. Again, I think the point is not about making the AI human-like (ie play to win), but making it competent. Firaxis seems unwilling or incapable of making the AI competent.

We could argue that making a competent AI is difficult - it is! But then a competent game designer comes up with mechanics that the AI can handle. A well designed game is put together knowing the capabilities of each game component so that they work together, not against each other.

This is I feel the source of brilliance of CIV 4. It was a game running on vastly less powerful machines than what we have today, and yet the AI felt vastly more competent. It is likely actually less "smart" than the AI in CIV 7 in absolute terms, but it was operating in a world which it understood and could reason about much better.

It's an issue with a lot of modern games. People have starry ideas about fancy features, but then they can't integrate the features together into a cohesive whole. I think this is something we lost in the modern days of game development. It has never been easier and more accessible, but this accessibility (which I support wholeheartedly), has made it easy for people to design themselves into a corner.

The problem with multiplayer in Civ games for me, aside from the really long play times, is that their networking code is extremely buggy so 80% of games I play end with the game state freezing or someone randomly dropping.

Reply on news.ycombinator.com