Jolly Joker wrote:Just why is it so important, whether the AI cheats or not? I mean, you play against routines developed by someone, mechanisms of play which are blindly followed; you play against some soulless, brainless, clueless somthing, pre-programmed to do things in a certain way.
It's hard to explain why something is fun and something else isn't ... but let me try. For me, strategy games have (among others) two important aspects: one is the competitive aspect (i.e. you match your abilities against someone else's, be it man or machine), the other is the "quest for the best strategy".
The competitive aspect means that I want to challenge and beat my opponent. I want to win. The better the opponent's playing ability is, the more satisfying a victory will be. Now when I play against a poor opponent, which would be easy to beat, I can give him advantages to make the game more challenging for me. For example, I sometimes do this when I play games with inexperienced players. While this *does* make the game more challenging for me, a victory still isn't very satisfying, because in the end you just beat a poor player, who needed specifdic advantages and rule tweaks to become a challenge. It's not at all as satisfying as winning against a player who really *is* good at the game even without cheats. So, when I play against a cheating AI, a victory won't feel very satisfying because I just beat a poor opponent who cannot really play the game well.
The "quest for the best strategy" aspect that I always try to improve my playing, and I like to monitor how others play and what they achieve. I like these moments when I see somebody (or some AI) pull off something amazing, something I didn't think possible, and I'll try to find out how that can be done, and whether I can incorporate it into my own set of strategies. This aspect of honing one's strategy holds a lot of fun for me, and I consciously stay away from FAQs because they'd spoil the fun of finding those things out by myself. But with a cheating AI, there's nothing to find out. Most amazing things the AI can pull off are based on its cheats, so there's nothing to learn from it, nothing to gain, and not much fun for me.
Jolly Joker wrote:Playing (Gal)Civ it's rather easy to create "personalities".
Sorry for snipping a lot of your post here.
I'm not sure whether I understand what you're getting at. You're saying that in (Gal)Civ it's easier to establish different AI personalities, and that (Gal)Civ is more forgiving to early mistakes of the player (I'm not sure whether the latter is really true, but let's assume it is).
I agree that HoMM is a more straightforward kind of game, and (Gal)Civ provides more room for totally different strategies. I also agree that this makes it easier to create "AI personalities" in (Gal)Civ.
However, this doesn't make a good competitive AI. Take a look at Civ1 - the AI there definitely has different personalties. You'll always see the greek building masses of low-tech units (for example), there are preferred and shunned governments and playing styles for the different AI civilizations. But the AI does still, frankly, suck. And that's partly due to the game design. The same freedom of choice that makes it easy to implement believable different AI personalities, makes it also *harder* to produce a truly competitive or even adaptive AI.
Now if HoMM is a series which has more straightforward gameplay, and less complexity in tech trees, diplomacy strategies, economics etc., then certainly it's more difficult to implement believable different AI personalities in HoMM. But at the same time, it should be *easier* to write a competitive AI for it, *because* it's more straightforward. The AI doesn't have to worry about massive tech trees, intricate diplomacy, or complex economics.
In my experience with the HoMM series (which doesn't encompass H5 so far), the most important decision is which stack to attack and which stack to stay away from. It's relatively easy to get any town going and produce a respectable army - of course there's room for debate about the *best* strategy to do this for a given town type in a given situation, but even if you choose a suboptimal strategy, it won't immediately break your neck. You'll just grow a little slower. Attacking the wrong stack, though, can turn a promising game into a very difficult struggle. But how do you determine whether you'll attack a stack or not? You assess the strength of the creatures, and their numbers (a rough estimate of which is given to you). This assessment is something a good AI can do far better than a human, because the AI has more calculation capacity. If the AI sees "lots" of goblins, it can simulate combats against the smallest and largest possible numbers, see how they turn out, and see how many losses are to be expected. Then it can assess the possible gains - experience for its hero, perhaps access to a mine (which should be weighed accordingly, taking other mines in the AI's possession into account, as well as the towns of the AI and their needs). Finally, it can decide whether the expected gains are worth the expected losses. This decisions is easier for a good AI than it is for a human, without any cheating on the AI's part, simply because of its greater calculation power.
You're probably right that comparisons with (Gal)Civ games are a bit problematic because of the difference in game concepts, but if it helps, think about Warlords II and III. Gameplay was much more similar to the HoMM series than (Gal)Civ games. And interestingly, the strong part of the Warlords series (which was less complex than Civ, for example) was the strength of its AI. Since Warlords is a more straightforward game as Civ, it was easier to write a good, competitive AI for it.
(Side note: For a couple of decades, AI calculation power was severely limited by the fact that the AI had to do all its decisions in a very small timeframe between the player's turns. Modern games, like GalCiv2, let the AI assess situations and plan strategies while the player makes its turn. This is a very viable way to do things because while the player is thinking, there's usually a *lot* of unused CPU power that the AI can put to use.)
The really difficult part for the AI is to build a long-term plan, to evaluate long-term sacrifices, and to assess combat outcomes against the player (who may act totally different from what the AI expects). I don't think I've seen an AI yet that's good enough to rival a player's abilities in that regard. But I'm not talking about these aspects. I'm talking about the comparatively simple decisions like "Can I beat this stack and if yes, are the gains worth the losses?".
And seen in this light, I understand that people who play a HoMM game find it difficult to understand why the AI isn't competitive without cheating, when other, even more complex games manage to do this.
)
