I guess that the AI will have problems with planning and strategic "awareness" even with this cheat in place. I mean, how many games on the EUIII's level of complexity have a non-cheating AI that can provide a good challenge to the player?
-
We have updated our Community Code of Conduct. Please read through the new rules for the forum that are an integral part of Paradox Interactive’s User Agreement.
I think the problem is, the ai can not guess the same way as a human can. So it must be substituted with knowledge.
This Artificial Intelligence is not sentient, so it plays by different rules, rules which it never ever breaks.
The AI does not cheat, it merely has some advantages to overcome it's inability to think.
Be very thankful that the AI has to have advantages in order to beat you...
What would it signify if it didn't need any?
I'll be back...
The AI does not cheat, it merely has some advantages to overcome it's inability to think.
Be very thankful that the AI has to have advantages in order to beat you...
What would it signify if it didn't need any?
I'll be back...
It can do better. It can read your unit amount off the ledger the same way a human can, then accurately calculate travel times for any unit it knows of as well as minimum travel times for any unit it doesn't know the position of given your technology, type of units available, weather, unit composition and a hundred other factors. It then can make a very accurate "danger map" of the whole world, showing which areas you can reach at most and with which forces for any time in the future. It can do so a thousand times per second or more often, as needed. It can calculate its own worst, average and best case scenarios in terms of amortised costs in money and manpower lost and gained for any plan it comes up with.
That's not something a human can usually calculate - we have to rely on rough guesses.
Yeah, right, that's why the player can beat it so easily 
. AI development is a tricky thing and very time-consuming, while good AI is not so flashy as other features, so it's harder to "sell" it. I guess that in theory the devs could just copy-paste features from EUIII, redesign the UI and devote the rest of their time to AI development, but then people would be complaining that it's just EUIII in a different package and there is no new stuff in EUIV. There are CPU limitiations as well. We won't get Deep Blue, folks. It's best to stop dreaming early on...
. AI development is a tricky thing and very time-consuming, while good AI is not so flashy as other features, so it's harder to "sell" it. I guess that in theory the devs could just copy-paste features from EUIII, redesign the UI and devote the rest of their time to AI development, but then people would be complaining that it's just EUIII in a different package and there is no new stuff in EUIV. There are CPU limitiations as well. We won't get Deep Blue, folks. It's best to stop dreaming early on...
Hmm, this might be true for a game of chess. But there are more possibilities in a game like eu. And we don't have one ai, we have many ai around the globe. The performance could be really slow through this. Don't you think?
Well, when your navy moves out of its visibility range, what should the AI do? You would have to program some procedure by which is estimates the position of your fleet, based on where it saw it the last time. It would then plan its movements accordingly.
I would assume that 99% of naval AI work is about two things: Not getting the AI's small fleets jumped on by a player's superior fleet, and maneuvering the AI's larger fleets to intercept smaller player fleets. When you always know where the player's fleet is, this is a possible task. If you do not know, you must program exactly what it should assume.
Here's where the challenge would come in. Should it assume that your fleet kept moving into the same direction it was last seen heading? That would make the AI awfully easy to trick into traps. He would still want to chase your small fleets, you could easily keep a large fleet in a seazone where he has no visibility yet, and dangle a small fleet in front of his eyes like a carrot. Because he does not know about the large fleet, you could easily lure him into traps. So the AI also needs to make some assumptions about what the rest of your fleet (aside from the ships he last saw) is doing. Should he make "defensive" assumptions: that the unseen ships are always stacked into one huge battlefleet, and luring just beyond visibility? This would mean the AI would never leave port unless all your ships are visible, and you could just do whatever you wanted with your fleet. You'd never get a naval battle unless the AI had more ships than you.
Programming that sort of AI would take a lot of effort. And you would still have to program all the stuff it has now... how to intercept, how to avoid getting intercepted, how to form concentrated fleets in war time, how to split off fleets if land units need transportation, and so on. It's just so much easier to just let it know where everything is. It still makes for a fairly predictable AI, but at least it does not fall for the most obvious traps.
Creating such danger/influence maps for the first time would take a few seconds. Progressively updating them is comparatively cheap on processor time though, but relatively heavy on memory (on the order of up to 10 bytes per province and active county squared, so something like 10 * 3000 * 100² = 286MB for the unoptimised case). Lookups are cheap too.
The general technique is called Influence Map (check the link for pretty pictures). It allows the AI to quickly determine if an area is safe (that is, the opposing player or AI is guaranteed to not be able to get any forces in time to that place) or if not, and in the second case to calculate a good estimate on what forces could be at worst in this area. This usually leads to the AI over-estimating the opponents if it looks at a single "province" of the map, but also means it will identify future front lines correctly and know which forces to allocate to be able to fulfil its goals there.
Looks pretty clever. Maybe its to costly to implement for Paradox? Or didn't they recognize such possibilities?
I would guess that it's too resource expensive for just one small (albeit important) aspect of the game.
It would multiply the amount of processing the AI has to do massively for every, single, country.
And be a huge coding pain in the ass.
Yeah, I guess you are right. I never thought of that.WhitemageofDOOM said:
I'm sorry to butt in, but it always looked to me that way, I've lured countless fleets of AI that way (and armies on land as well). I've always assumed that AI was just confused by FOW, but if it can see through it, how does it still fall for it?
I don't see why the AI needs to be given cheats to be made more challenging. If you want a challenge can't you just play a weaker nation?
I think difficulty should solely be factoring the AI's awareness of your presence. Not the tech, manpower (limit), gold, inflation, w/e, bonuses, and ESPECIALLY not the morale bonuses. Nothing is more annoying than when an equal army defeats you regardless of how fortified you are, what terrain you sit on. If I play on high difficulty, the only time I want to actually see it is when I stick my nose out. The AI would go: Player expanding - threat level compute - need for destruction compute - ERROR ERROR - Computes - Coalition successful, commencing total face**** mode - Computes - player ragequit, victory, feels remorse, must buy flower and candy to make amends.
The AI should always be made smarter, regardless of difficulty. He should always try to flank, set up defences, guard his coasts, "avoid" tensions and flex his muscles. In other words, become more "human".
I also hope AI agressiveness will be in still. But that should also determine the AI's attitude towards other AI's.
The AI should always be made smarter, regardless of difficulty. He should always try to flank, set up defences, guard his coasts, "avoid" tensions and flex his muscles. In other words, become more "human".
I also hope AI agressiveness will be in still. But that should also determine the AI's attitude towards other AI's.
The only thing I really want from the AI is to want to reduce casualties as much as possible (meaning trying to avoid sending 100 000 men into a scorched province and just sit there, not storming, the fort) and to make them a tad bit smarter when it comes to peace deals.
That's all I want
That's all I want
This is an example of a bad cheat - the one that gives idiotic results and annoys the player. Well-designed cheats only HELP the AI in areas when it needs help, but they don't destroy the whole game balance or make realistic strategies totally useless.
AI cheating and AI competency are not contradictory - they can supplement each other. Nobody is saying that the devs should stop AI development - quite the opposite. However, realistically the AI will always need to get some help in X or Y.
I know he has to be joking, but reading this post made me realize to my dismay that I've seen stupider ideas on the wed before.