dannythefool
King
- Joined
- Jul 15, 2009
- Messages
- 657
Many of the AI problems seem to be dependent on difficulty levels:
Just my thoughts, I'm probably missing one or two things.
But actually some of this could explain the alleged discrepancy between AI performance in the press build and the final release. E.g. if the pathfinding bug appears only in the final version, that would have made the AI in the press build stronger, etc.
- Warmongering: This seems to be much less of an issue on lower difficulty levels. My impression is that when deciding on a war, the weight of military superiority is much too high and past peaceful relations are valued much too low. On higher difficulty levels you almost automatically have weaker military because to a human player, large numbers aren't as useful when a small army will do the same job as effectively. This appears to cause every civ to act like Civ IV Monty.
- Combat: The AI is bad at tactical combat. You'd think that this is independent of the difficulty level, but I've been noticing that AI attacks on higher difficulties often devolve into a chaos because the AI has so many units that placing them tactically is no longer possible within the constraints of its programming. Whereas when it only has two archers and three melee units, I've seen it pull off a fairly good attack when the terrain was in favour of it. (It seems that the slightest terrain movement malus will break the AI's formations, it appears to assign a higher value to moving fast than to keeping a working formation.)
- Pathfinding: Sometimes pathfinding seems just plain broken, e.g. when the route preview shows a different cost from what you actually get. This happens a lot to me, especially for travel that involves water tiles, and I think it should affect the AI even more than a human who is capable of noticing the errors and finding his own routes. I also think that the AI is too scared of using water tiles in addition to that. You can often block a land bridge with a handful of units and the AI will only embark its units and move them through the water if it has so many units that it doesn't know where to put them anyway. While this isn't strictly related to difficulty level, it does mean that the headcount advantage the AI has on higher levels is practically moot because it can't use it. What the AI needs to do is figure out choke points and then secure alternative routes.
- Expansion: On lower difficulty levels, the AI will not expand enough. To some extent that's reasonable, as the cost of maintaining a large empire can be quite high. On the other hand, this dramatically changes the game when you increase difficulty. The game doesn't just become harder, the bonuses actually change the way the AI plays.
- Naval Exploration: This may actually be related to the pathfinding problem. Planning a route that involves water will show you that the cost is calculated incorrectly. If this affects the AI as well, it would explain in part why colonizing another island is such a low priority: Crossing a few water tiles to get to that island with no insane resource bonuses (since they've all been reduced a bit) would appear as much more costly than riding across a plains. I've only ever seen an AI embark settlers when it had NO uncultured land tiles available on its continent at all. It also seems to have too few ship units, and it doesn't seem to be interested in exploring islands beyond finding them. Even a trireme has enough visibility to uncover a normal sized island, but the AI isn't, for example, trying to take the ruins from the human players, which is a real problem in single player games. Ruins don't seem to be very desirable for it in general, I've seen them sit right next to an AI's border, but ruins on other land masses are just ignored entirely.
- City Trade: A quick money strategy is spamming settlers, founding remote cities in inhospitable land, and selling them off to the AI. Especially useful on archipelago maps. The AI greatly overestimates the value of these random remote settlements. This is more of a problem on higher difficulties because that way the player benefits from the AI's resource advantage that was supposed to counterbalance the player's ability. This also affects AIs that sue for peace. On low difficulties, you don't get ridiculous amounts of cities, resources and gold for peace because the AI doesn't have them in the first place, but on higher difficulties you get the benefit of the AI advantages that were supposed to work against you. Selling a useless city for 1k gold plus resources is a much too efficient hammer/gold conversion ratio (with hammers being production you can't use while city is building the settler, of course, since it doesn't use hammers).
- Resource Trade: Another early money scheme is selling your first luxury resource to the AI. The AI will pay way too much at a time where every civ should still be happy anyway. This is another case where, on higher difficulties, the AI's advantage that lets it have more gold is transferred to the player.
Just my thoughts, I'm probably missing one or two things.
But actually some of this could explain the alleged discrepancy between AI performance in the press build and the final release. E.g. if the pathfinding bug appears only in the final version, that would have made the AI in the press build stronger, etc.
