I'm still steamrolling on ridiculous. I've tried everything to lose... big maps, small maps, lots of opponents, few opponents, different victory conditions, etc. I just can't lose. Any advice?
First you post a thread about Derek sleeping, then this?
Are you working on your post count or what?
I'm on the AI today actually.
But I need your help for those of you who are stream rolling. The AI's weaknesses are more to do with strategy than anything else.
To use a Starcraft analogy, it would be like the AI developer not realizing how important it is to block off their base making it easy for the Zerg to win.
Some of the steamrolling comes from AI doing stupid stuff, but much of it comes from the game being so easy for a player to exploit.
Armor is a huge problem right now, as is blink (as I mentioned in my previous thread).
Really I think there are three things that make the AI weak vs me.
Heavenfall aka "the broken record" says: The strength of the AI on higher difficulties would be multiplied if they focused on building groups and squads of units with high quality levels, instead of basic quality solo units. As soon as I get a group up, it can defeat a nearly infinite amount of solo enemies.
I would go as far as to say that as soon as the AI has access to quality and group upgrades, it should never build anything that is "less than max".
Pulling numbers out of my ass, a 4-man party of elite enemies would probably be 12 times as dangerous as these solo units running around.
I would also like to see the AI not make constant demands for gildar. It is way too easy to pay them off. Ask for gildar once, then invade!
My basic strategy on ridiculous:
1) Expand as much as I can. Get Tools of War / Equipment to build units with 9att 5def. After that, rush to libraries (or whatever the empire version is called).
2) Continue expanding until there is no more room to expand. Avoid war at all cost. Don't worry about AI declaring war, at the most they'll want some petty cash.
3) Research some weapons, armor, groups, quality upgrades. Train these with call to arms.
4) Invade. Game over, because AI has NO military strength.
Would it be helpful for you if we did some "how I make tactical decisions" flowcharts? (at least for the tactical battles)
I find that many of my decisions in the tactical portion of this game and MoM are relatively linear, and could probably be mapped out if I spent a little time with it.
I think the AI may be underestimating sovereigns and defense in its calculations.
Hi Brad,
1. Have them first research equipment/tools of war, then create a fully armored dagger dude, axeman, and/or spear guy. One to protect the city they're in and one to escort the pioneer that you build next. No more unprotected pioneer spam please. Maybe have them keep at least x armies in the city where x is the city level for defense. That's what I typically do.
2. Teach them to build teams of armies.
3. Teach them to use magic so they can tie up my best guy somehow.
4. In tactical battles, the AI tends to have a favorite target. I can run that favorite target around the map while the AI just chases him around. Meanwhile, I'm pounding the AI. He counterattacks, but does not attack because he wants to chase that favorite target around. Don't do that.
Thanks for your support!
1. The AI does not protect it's own town/cities, when I attack an AI town it rarely has more then 2 to 4 units in it, even after a 100 turns.
2. The AI builds way to many pioneer's, you can camp outside an AI town and just kill off pioneer's as they are produced.
3. Tactical combat, stop the suicide charge, the AI should wait and then try to overwhelm the nearest opponent, stop moving archer/casters, let them attack with ranged combat. Ambush mode, where a AI unit will hold their ground until an opponent get's within attack range, this will help the AI get the first attack.
4. Flanking, have the AI sent groups to attack towns to try to drawout town defenders.
The ai has been known to walk their sovereign into my land without anyone else. So I kill him or her and that's it for that faction.
So yeah, I think everyone here agrees that no amount of strategic brilliance on the AI's part is going to make up for the worst tactical AI I've ever heard of.
I get worried sometimes that people confuse Artificial Intelligence with game difficulty. A better AI doesn't mean it is extremely difficult to beat. It means it makes decisions more like a human being. It can see and carry out all of its choices, but it can still make mistakes.
This is something that bothered me a bit with GalCiv 2. The AI was really, really difficult. I don't really want Elemental to be like this.
I think the Pioneer thing is actually a bug because during the beta I actually started hard coding limits on # of pioneers so there's some weird issue going on there.
Now that I'm free to...indulge myself...we'll see what can be done.
I've never looked at the tactical AI, so I don't know how much I can do (and I don't want to step on someone else's area even if it's during the holidays).
There's probably a handful of stupid things the AI does in tactical battles i could address though while I'm messing around.
The AI way way way underestimates armor. It might have 100 to 1 the number of troops, but it dosen't mean squat if they all have clubs and my one guy has platemail armor. I just keep one armor level ahead of the AI's weapons and they can't ever hurt me. That and the AI dosen't use things like helmets and such, it just throws out the default soldier designs, which are far from the best.
I think the groups are a main problem of Elemental, because they make the scaling very difficult. A group has a higher damage output and higher HP. This makes the balancing of single units or spells very difficult. Other games like Age of Wonders: Shadow Magic do not have groups and they should be removed from Elemental, too.
A lot of the steamrolling for me is tactical AI which I think you mentioned you aren't involved in. Not counting lack of equipment variety and stuff like that.
I think the strategic level AI has come a long way though, a few points I can think of for improvement:
Pioneers are still a bit of an issue, and if at all possible I think you might want to program the AI to settle near their own borders and not expand past an enemies border to setup cities on the other side of them just because there is empty space. In this situation the AI should swap focus from expansion to fortifying its position, increasing aggression, or emphasizing tech, etc.
Resources being scarce, the AI needs to use them more wisely, Iron is a huge glaring issue atm. Better city placement to take advantage of resources would be good too.
Technology, AI should be placing less emphasis on adventure techs on more on techs that actually matter like armor and weapons. They should then USE that armor and weapon tech. Early game seems ok, with AI massing spearmen in padded armor. It goes downhill from there though which I can only assume is due to improper management of iron. Late game I will see AI with lord hammers and....leather armor.
Each AI should be trying to work on some win condition too. More aggressiveness overall for AI would be great, though the AI can't judge its opponents very well yet.
How about actually never seeing the AI attack another player?
I had two ridiculous AIs declare war on me, but all their armies just sat around doing nothing.
Adding to what TheProgress said, on anything less than ridiculous the AI wont even declare war and when the player declares war, as said before the ai wont attack.
Oh this brings up another common player tactic that AI should use. AI should stack its war armies on the border of the player before declaring war. It can then march in and stomp the area a bit. Though Teleport really breaks the game a lot, this AI change can really make a player feel that this "war" thing is something he should probably take notice of.
So yeah, the main issue isn't an AI one. It's a balance one. Seriously decrease obsolescence, and you'll see the AI doing much better. We've discussed this countless time: quality versus quantity doesn't work, and the best research strategy is to beeline for heavy plate (and a couple adventuring techs), while the AI will be busy doing quests and researching diplomacy, you'll very quickly have an armor he won't be able to fight until maybe 100 turns later. As long as this won't be fixed, I'll steamroll the AI.
Then, it's what Heavenfall said: the AI needs to start building better units, and in more substantial quantities. Actually, it needs to adapt to what I'm doing a bit more and respond with the appropriate units and improvement building strategies. It needs to optimize its city production more. It needs to regroup inside the cities in case of an invasion. It needs better threat assessment: if my units are within range of his city to attack it the following turn, don't lose time patrolling: reinforce ! Only sortie if you can afford it ! Put your back lines cities into full unit production mode so you can actually kick the invader out of there ! And if you can afford it, send one or two fast, cheap units destroy iron/materials/gold producing improvement and caravans in enemy territory ASAP, while avoiding enemy troops (so you can prevent the opponent from swarming you). Use two, so he can't easily teleport-hunt you before you destroyed at least a couple improvements and caravans.
Oh, and it needs a few lessons in city-optimizations. It needs to start saving gold and resources. Libraries in a research dedicated city only (almost), for instance.
I've steamrolled quite a few myself, and I'll share my thoughts on the matter.
You are allowed to build up too much without harassment. I can literally turtle myself during what I like to call.. "The Closed Building Stage" and then commence with the "War Machine Stage" with my super-powered economy. At that point, I build up large armies of armor-clad juggernauts, and it's basically just a slaughterfest from then on.
Coding the AI to battle the human brain will be very tough, especially with a game that lets you do so much with everything. If the AI can build up groups of soldiers that would give me a problem, I can just wait a ways off with my army until that group leaves their city, and then I swoop in for the kill.
Game Alteration Ideas to make conquering more difficult
This is actually one of the things that is the oddest about AI in current format.
In my experience, the AI does a decent job of putting together appropriate sized groups, but does not use them well. It will send a couple stacks out to engage you on the way to the city, but by the time you get there, they have sent all but 1-2 models out to get beat up already.
I totally get the first "get off my land" wave to either soften up the invading force or repel it, but if that fails, it ought to curl up behind walls and start cranking out troops. Staying in cities gives its models a bonus to health and armor, and the AI does not seem to appreciate that.
Additionally, it really needs to start differentiating between appropriate peace time and war time standard garrisons. Someone tells me we are going to fight, and all but the most remote of my cities get at least 4 models (typically 2 melee and 2 ranged) or more garrison as quick as possible. AFTER that, I develop the invading army.
On the same note; it would be nice if the AI started its buildup a couple turns before declaring war (yes, I've seen it a couple times), so that the turn or two after it tells you that its time to fight, you are fighting. (not 10 turns)
Part of the AI problem is that it needs to predictive as there is no way to be reactive. If city zones had a slowing effect on enemy units, it would allow the AI to maybe mount a defense. It might also allow a sovereign to gather units via teleport or "return" and otherwise react to an assault.
In the modeling of AI unit approaches, I would like to see an elitist style (expensive, effective maximized units), a soviet russian style (masses of inexpensive, moderately effective units), and a mythical style (armies primarily composed of fantasy races or mythic beings). You could toss in stuff like lots of mounted units, massive recruitment of champions with magical gear being a part of that plan, leaders that utilize terraforming to devasting effect, etc.
Regardless if these are well liked or not, there should be an overriding approach or technique that gives our AI opponents some discernable differences in how each one conducts the business of war.
Top
The AI needs to build more tech libraries and use more spells. It needs to 'zero out' it's resource piles through trading or building.
I was fighting Resoln the other day, and I noticed that the faction had stockpiled like 2000 Gildar. They had 50+ Power level above me. Eventually, they decided to declare war. I crushed Resoln utterly. The only small stumbling block I had was that they HAD purchased a demon. Individually it would have caused trouble, but I bombed it with Arcane Arrows and killed it utterly.
Really?
RESOURCE MANAGEMENT
If the AI has Gildar and Materials, it should be building basic troops. War Staff + Leather
If the AI has Gildar and no Materials, it needs to invest in more workshops.
I think the AI does this already, but it should research Magical Weapons/Equipment to drain it's excess Gildar.
If the AI has too many resources, and nothing to do with it, they should start building outposts to take advantage of empty queues. That, or forging alliances and paying other people to do it's bidding.
MILITARY STRATEGY
-Maintain large army groups! The AI needs to have at least one 'main stack' that is composed of the strongest units it can muster when at war.
-if the AI is at war, and there are units at it's borders, the AI should use casters and teleport to 'gather up' the strongest stack it can, and try to crush enemy formations.
-Maintain parity with neighbors! If the AI is running caravans, it can see the cities of other players. The AI should try to maintain an army of equal (preferably better!) size and combat rating. It can be a champion type army, a mundane army, a summon army, whatever. The AI should calculate every player's total combat rating and try to maintain a force depending on it's disposition. (Of course, combat rating calculations need to change as well)
Military/Conquerer AIs want a force 50% larger than the largest neighbor by default. If a 'conqueror/Hitler' type of player has appeared and is waging aggressive war, then all AIs will want a 50% larger force than the conqueror
Peaceful AIs want a force 25% larger
AIs that have a good relationship with the largest neighbor can maintain an equal force.
-The AI needs to identify where and when a 'climactic battle' will occur, IE when each player's largest armies will clash. Preferably the AI will try to force this battle if it is stronger. During this 'climactic battle' the AI should cast it's strongest spells with no regards to mana. Before the 'climactic battle' the AI should try to use all of it's enchantments, namely cheap stuff like Arcane weapon.
-The AI's military units should ALWAYS STAY CLOSE TO EACH OTHER! It should have one 'main force' or 'flagship stack'. If it needs to garrison remote territories, it should create separate units for that purpose.
-A clever, devious AI might have multiple strong stacks for the purpose of 'forcing choices'. It might attack two cities at once, or run through the countryside destroying improvements.
-If the AI has access to strategic combat spells, like volcano or column of fire, it should use them on the enemy's largest stack.
-If the AI is weaker, it should send fast units deep into enemy territory to attack resource nodes. At least 25% of the AI's military should have a movespeed >2, preferably as high as it can go.
-The AI should use Terrain spells to create chokepoints between it, and any neighbors. The AI won't do this if it is allied to a neighbor. It should pre-emptively close off paths to vulnerable cities.
-An AI on the attack should use Terrain spells to create a direct path to target enemy cities.
DIPLOMATIC STRATEGY
-Unless it is the strongest faction, or a homicidal maniac AIs should try to band together with 'like' factions.
-The AI should feel free to 'gang-bang' smaller factions when it has the resources to pay off other factions to attack.
HERO/SOVEREIGN STRATEGY
-The AI should specialize it's Heroes. There are three types, CASTER, COMBAT and CHARISMA. How many it has of each type depends on the sovereign and faction.
-AIs with many caster heroes should focus on getting tactical spells, and stuff like the fire books. Caster heroes should ONLY have Int raised, and nothing else.
-The primary trait of combat heroes is Strength. An AI with a lot of combat heroes wants magical equipment and lots of gildar to buy it.
-The AI should level up CHARISMA heroes in combat, and then leave them in a city when they are level 5 or so.
CITY STRATEGY
- The AI should have separate, specialized cities focusing on TECH, MAGIC TECH, and GILDAR advancement. What city receives what upgrades is based on nearby resources.
-The AI needs to build MORE TECH buildings. Tech is the most vital thing in this game. Without tech, magical spells are very limited, and there is nothing to spend resources on. If the AI has a glut of resources or mana, it should focus on building technology fullstop.
-Build queues should never be idle! That is, unless the AI is running out of money/resources.
VICTORY STRATEGY
-Conquerer/Evil AIs should aggressively attack smaller factions. They should declare war once they've moved their forces in position, NOT before. They should make alliances of convenience and not be shy about backstabbing and absorbing weaker allies.
-Peaceful AIs should either focus on creating and maintaining alliances, and trying to win through an Alliance victory. They should strengthen 'friends' and weaken 'enemies'
-The strongest faction in an Alliance should be a 'peacekeeper'. If there are no more potential allies, they should move the alliance in the direction of isolating and destroying other factions.
-Weaker members of an Alliance should not be shy about gifting resources and units to a stronger ally if they are not immediately threatened by other neighbors.
-Neutral AIs maintain alliances to keep security and maintain a 'balance of power'. They try to win the game through magic or quest victory.
That's all I can think of for now.
What I hope to see in the future... is an AI that can determine that the best approach is to bypass small towns and even use terraforming to level mountains that are in the way to get to our capital and cut off the head, then mop up the rest from there.
I'd like to see an AI that's spent a lot of time researching spells and gaining mana (from essence buildings, etc) to throw up a mountain pass between you and him to protect himself.
I'd like to see a volcano suddenly appear on my city, or a wall of flames start burning my caravans at a major intersection.
I'd like to be able to call to the aid of an ally and actually see them attack our mutual enemy, instead of having to do it all by myself.
The game has some great spells in it that could really change the playing field (some quite literally), and it might make the game feel more magical if the AI used it to the extent that I'm talking about.
There are many great features available to you once you register, including:
Sign in or Create Account
