Question about AI Quality

[Jolly Joker]

That's not the question.
Winterfate said H III AI was the strongest in the series.
I asked how it was strong and listed things to show that it was actually the the dumbest AT THAT POINT.
The strongest AI was in Heroes I (which is not surprising because it is the simplest of all Heroes games).
H IV doesn't count because the game setup basically made it impossible for the AI to exist at all. Actually, playing in Hard mode, the Heroes IV AI is, imo lightyears ahead of the AI in H 3.
Challenge-wise H 5 is on par with H 1.

[Asjo]

Jolly, by reading my post from which his reply originated you might be able to avoid that question.

[Blaze85]

"i kill its strong heroes and he hides in town"

This problem alone takes away the challange form H V...
In H 3 i had to keep fighting high level ai heroes with great artifacts until the end of the map. That kept up the challange.
In H V the Shackles of the Last Man is useless. It's a shame...

[Jolly Joker]

Jolly, by reading my post from which his reply originated you might be able to avoid that question.

But I disagree with that post.
When we talk about AI, we are surely talking about simple MP maps, not one-player maps with specially designed AI players. On those simple MP maps the H 3 AI is inapt, imo, because it simply IS NO challenge, even though it acts like it would play in a humanly fashion. The battles against the AI in H 3 are a cakewalk. It is VERY anti-climactic when you play a good map like, umm, Free For All, the H 3 version of Dominion, Europe and so on - large maps with the promise of a fair challenge -, but the map is basically over once your (first) main hero gets Town Portal and you start steamrollering the AI that offers no resistance at all.
In H 5 the situation is different. You might be steamrollered by the AI, and this can happen on smaller 1 on 1 maps early, but even on larger maps where one AI player starts conquering others and you may face something like an AI player who conquered already 4 positions and is no closing in on you.
However, on any map I would name close to stalemate the AI simply turtles in its town. Actually, since winning the map involves conquering that you cannot win easy - the armies are mostly under the best hero, the AI hires and the end battle(s) is (are) brutal at times. It is certainly more interesting than the floor-wiping happening in H3.
Moreover, for H IV, I'm no friend of that game, but the AI is not nearly as bad as it looks - it simply didn't get the chance to shine because it just can't function in Very Hard and Impossible environment. The paradoxical thing is that H IV plays most interestingly on Hard difficulty (only).

My opinion is simply, the worst that can happen is, that a game actually manages to create the impression in some areas, only to fail miserably in others.

[Alamar]

Can someone please tell me where the H 3 AI was strong?
1) Hero developing: Zilch. The AI favors all the crappy skills the players avoid like Scouting, Eagly Eye and so on. You could say the AI made it an art to create the worst heroes possible.

Are you implying that H5's hero builder is any better at picking skills. Why would I get L19 Necros whose only magic skill is basic light? Why would I get a high level Wizard with NO MAGIC SCHOOLS at all??

The only reason that H5's AI looks better is that there are fewer "boner" skills.

Considering that the HoMM series has had 10 years to resolve this sort of issue you would HOPE that we'd have gotten it right by now. BTW [IMHO] I believe that it is fair to say that H3 is better along these lines because expectations weren't as high 10 years ago .... and we have the EXACT same bad heroes that we've always had.

2) Battle AI. Bad. No grasp of the waiting command. Silly preferences. Easily tricked into attacking single level one creatures when it could have wreaked so much havoc. Extremely exploitable. Because of 1 Heroes have no mass spells. Spells actually cast are a joke.

I generally don't see the H5 AI using a wait command either. As for the H5 AI I can easily trick it into attacking a single IMP and let my shooters pelt whatever creature it is from long distance. [Assuming creature can't reach me in 1 turn that is]

I can [and sometimes do] exploit the H5 combat AI terribly. I sometimes find that it is the only way to win on heroic difficulty.

I can also say because of 1 above heroes don't cast mass spells. I went for over 2 months playing this game before I EVER saw a hero cast a mass spell .... then that was only because they were high level and started with Light Magic so they probably couldn't choose anything else.

Spells cast are often a joke with H5. I went a couple of months before I ever say a Meteor Shower and I don't think that I've still ever seen an Implosion [cast by AI]

I rank H5 as only SLIGHTLY better than H3 in this regard. If you add in 10 years of advancement in computing technology I'd say H5 ranks below H3 in taking advantage of what's available.

3) Sieges. Incapable, as attacker AND as defender. Give the AI a defending hero with Tactics and have a good laugh before the siege starts. No concept of moat damage. No concept of even using the walls.

I'll give the edge to H5 here although I'm not sure that the H5 AI properly calculates moat damage either ....

4) Adventure AI. Easily exploitable and tricked. No concept of the fact that visiting and garrisoning forces are being unified when the town is attacked.
So what was good?

I'll give a HUGE edge to H3 here. H5 STILL doesn't pick up artifacts, flag mines, etc. It's worse than a chicken with it's head cut off.


In summary point 1 is a dead heat. Points 2 and 3 are a slight to miniscule edge to H5. Point 4 goes to H3 in a LANDSLIDE. The close winner is H3 because of the huge landslide of option 4.

If you factor in expectations, advances in computing power, advances in # of people to work on the game, etc. the HUGE victor comes out to be H3!!!!!

[Jolly Joker]

The only reason that H5's AI looks better is that there are fewer "boner" skills.
Considering that the HoMM series has had 10 years to resolve this sort of issue you would HOPE that we'd have gotten it right by now. BTW [IMHO] I believe that it is fair to say that H3 is better along these lines because expectations weren't as high 10 years ago .... and we have the EXACT same bad heroes that we've always had.

Actually there are NONE (boner skills) in H 5. The only thing here is that the AI in H 5 could pick more abilities instead of skills. The bottom line is, however, that the the Heroes in H 5 are lightyears better than those in H3.

What you say about 10 years of resolving things is not applicable because the developer changed, and the game changed. So that point is completely invalid.

I generally don't see the H5 AI using a wait command either. As for the H5 AI I can easily trick it into attacking a single IMP and let my shooters pelt whatever creature it is from long distance. [Assuming creature can't reach me in 1 turn that is]

I can [and sometimes do] exploit the H5 combat AI terribly. I sometimes find that it is the only way to win on heroic difficulty.

I can also say because of 1 above heroes don't cast mass spells. I went for over 2 months playing this game before I EVER saw a hero cast a mass spell .... then that was only because they were high level and started with Light Magic so they probably couldn't choose anything else.

Spells cast are often a joke with H5. I went a couple of months before I ever say a Meteor Shower and I don't think that I've still ever seen an Implosion [cast by AI]

I rank H5 as only SLIGHTLY better than H3 in this regard. If you add in 10 years of advancement in computing technology I'd say H5 ranks below H3 in taking advantage of what's available.

We seem to play a different game. Just attack any stack of neutrals and switch to auto-combat. You'll see the AI wait a lot. It doesn't wait in full-out battles very often, but the reason for that is that it's not often paying to wait. You should maybe select a duel and let the AI fight it out, that's maybe enlightening.
Spellcasting is massive - if the AI has them. Again, try duel mode. I agree, though, that getting high level spells and using them in combination with abilities should be higher prioritized. Apart from that, the AI handles casting lightyears better than the heroes in H 3.

3) Sieges. Incapable, as attacker AND as defender. Give the AI a defending hero with Tactics and have a good laugh before the siege starts. No concept of moat damage. No concept of even using the walls.
[/quote]

Adventure AI
I'll give a HUGE edge to H3 here. H5 STILL doesn't pick up artifacts, flag mines, etc. It's worse than a chicken with it's head cut off.

If it doesn't pick up artifacts I really wonder where those are coming from I always get from the AI heroes I kill? Also I wonder who had those mines flagged when I enter AI land?
You are flogging a dead horse here.

In summary point 1 is a dead heat. Points 2 and 3 are a slight to miniscule edge to H5. Point 4 goes to H3 in a LANDSLIDE. The close winner is H3 because of the huge landslide of option 4.
If you factor in expectations, advances in computing power, advances in # of people to work on the game, etc. the HUGE victor comes out to be H3!!!!!

Yeah, H 3 is the big winner. That's why you lose duels if you are not playing sharp, that's why you can lose even on hard if you abstain from too much reloading in H 5 and that's why you wipe the floor with the AI on each and every difficulty level in H 3. Because the AI is so refined and subtle and well designed in H 3 while in H 5 it's a primitive, dumb cheater that swings its big money club without finesse.

[DaemianLucifer]

What you say about 10 years of resolving things is not applicable because the developer changed, and the game changed. So that point is completely invalid.

So what if its the new developer?What,they are using the same computers 3DO had 10 years ago?They are using the same AI algorhitms 3DO used 10 years ago?

We seem to play a different game. Just attack any stack of neutrals and switch to auto-combat. You'll see the AI wait a lot. It doesn't wait in full-out battles very often, but the reason for that is that it's not often paying to wait. You should maybe select a duel and let the AI fight it out, that's maybe enlightening.

Yup,auto combat became a bit better,but overal AI extremelly seldom uses wait.And no,first aid tent waiting doesnt count as waiting.

Spellcasting is massive - if the AI has them. Again, try duel mode. I agree, though, that getting high level spells and using them in combination with abilities should be higher prioritized. Apart from that, the AI handles casting lightyears better than the heroes in H 3.

Yup,spellcasting is massive,but its targets are...pittyful.Take frenzy for example:I only saw AI using it either on its own troops(after which I move away and it attacks its own),and on my treants that attacked unicorns after that,while my hunters diminished its ranks.And those arent the only idiotic things AI does.

[Elvin]

Instead of saying how adequate the H5 AI is and how lame the H3 was you might try to see them in action without the cheating.Then H5 is really exposed.Yes the combat AI is a lot better in almost every sense but in the rest areas...

[Gaidal Cain]

Actually there are NONE (boner skills) in H 5. The only thing here is that the AI in H 5 could pick more abilities instead of skills. The bottom line is, however, that the the Heroes in H 5 are lightyears better than those in H3.

There are at least two skills that an AI probably should avoid- the magic schools not associated with it's starting town. War Machines and Sorcery might not be very good choices either, depending on hero type. But it's right that the situation is better now than it used to be, even it's far from satisfactory.

[Khelavaster]

*****
I'm surprised there are people who still take what Jolly Joker says seriously.

The guy has always made a fool of himself, ever since I first saw him back in the day when he was praising the braindead HOMM 4 AI to high heavens on the 3DO boards.

His job -literally- is to circle-strafe arguments and spin facts to always cast a favorable light on the game, regardless of how obvious the facts are, while occassionally griping over minor issues to throw off the scent. And always in that haughty tone of his.

Seriously guys, you may want to look through the ruse already.

*****
MEANWHILE, IN HOMM V's WONDERFUL AI WORLD...

An enemy hero attacks my scout leading 1 bone dragon and 4 War Dancers. My scout has only a Master Gremlin to his name.

The enemy hero spends ALL of his turns casting positive spells on the bone dragons, who strangely remain fixed to their initial position. My Gremlin picks off the War dancers and finishes the bone dragon helped with my spells.

After several turns, my Master Gremlin comes off victorious. The enemy hero never even used his normal attack.

All hail the mighty HOMM V AI ! And yet the battle AI is miles better than the strategic AI. Do the math.

Khel.

[Elvin]

Occasionally it does stupid things like that but overall it's decent enough.It's the adventure AI which is really troubling...

[DaemianLucifer]

I'm surprised there are people who still take what Jolly Joker says seriously.

In his defense,he does have some moments of clear thoughts.Maybe he is an AI himself

[Jolly Joker]

*****
I'm surprised there are people who still take what Jolly Joker says seriously.

The guy has always made a fool of himself, ever since I first saw him back in the day when he was praising the braindead HOMM 4 AI to high heavens on the 3DO boards.

That looks like you were not able to grasp the problem of the H IV AI then, and it looks like you didn't learn much in the meantime either.

I remember very well the days when people started bashing onto the H IV AI, when It wasn't the AI that was the problem, but the difficulty level system and the inability of the programmers to take the wandering monster stacks into consideration. I don't want to go into detail here, but the main thing is, that the H IV AI LOOKED braindead, but actually never had a chance in game due to the fact that it had to cope with the massive creature stacks the players got under Very Hard and Impossible.

Every idiot can see that the way Heroes IV difficulty levels worked was massively flawed. Neutral creatures worked that way that a stack was always comprised of a number of XP, but that creature value would be less the higher the diff level.
So if you would have a stack of 12 creatures with 240 XP on normal, you'd get 18 on Hard , 24 on very Hard and 36 on Impossible (8 on easy) for the same 240 points.
While this made the game more difficult against neutrals the same was true for the AI as well which had to fight against those massive stacks as well - and obviously couldn't do better against them as a human, because of the auto-combat system that didn't take into account finesse, but only pure strength.
Now, it COULD have worked, but only if those stacks would have been different for the AI. Logically, for a Human playing on Impossible (and said stack would have consisted out of 36 creatures for 240 XP), if the AI would fight the same stack it should have had to have them on EASY diff which means, for the AI the stack wouldn't have been 36 creatures, but only 8 - for the same 240 XP points.
Clearly, this would have made the game a challenge. In fact I think it could have been a brilliant way to do things. The way it was, however, the game became ever more difficult for the AI with each difficulty level.

You can just as easily check that on HARD diff level (the only playable diff level in terms of the AI) the H IV AI is not bad at all - on the contrary.

[Elvin]

An AI that can't deal with neutral stacks or attacks them earlier than it should,taking heavy losses in the process doen't strike me as a good one.And it never occured to me it was good on hard either.

[Khelavaster]

Jolly Joker, please drop the act. It's cute but tiresome. There's no way in hell that a sentient being could call HOMM IV's AI anything but awful.

It was terrible. Disgraceful. It stank worse than a dead woodchuck under your porch after a month. Even more so after the ton of undeniable fun and challenge HOMM III brought to the table.

HOMM IV was the game that single handledy nearly obliterated the series for me and my friends (who by the way never came back after the fiasco).

And it wasn't because of the new system, the fighting heroes, the potions, the lousy combat grid, not even the terrible bugs.

It was the broken AI. You see, it had nothing to do with neutral stacks or XP. It had nothing to do with difficulty levels. We used to play cooperative against the computer and it wasn't long until we found out that the AI couldn't find the way to your castle. Hence the multitude of ONE-WAY PORTALS LEADING TO YOUR BACKYARD.

The AI had no aggressiveness. Its priorities were all wrong. I made a map with the editor once. I stood with my hero and just one skeleton in front of the computer's castle, full of vampires and assorted ghoulies. The computer NEVER CAME OUT even when it would've meant instant victory.

In all our games, we never even saw the computer until we reached its castle, and even then it was locked inside with a handful of units. Games quickly became all about fighting neutral stacks, which some people may find an amiable passtime, but we had cut our teeth on three previous incarnations of challenging, slightly flawed maybe, but completely satisfactory Heroes of Might and Magic games.

No matter the causes, the reasons, the budget, the new system, HOMM IV's AI was embarrassing and a nearly fatal blow to the series for us single player gamers. If you can still defend it, I'm not surprised that you see no fault with V's.

Khel.

[Jolly Joker]

To get this straight. I never said the AI was good in H IV. My point was that the AI basically had no chance to do anything else than look stupid. The problem wasn't actually the AI. The problem was, that the AI had a more difficult game to play, when it was actually the human who should have the more difficult play while the AI should have an easier one. The problem is, even if the AI would have been more aggressive and so on, there is no way in hell it would have been able to beat a stack of 50 Nightmares the way a human could because of quick combat and the consequences of it on the higher levels. Additionally there was the wandering monster stack problem.

Anyway, even you with your doubtless praise for something called adventure AI in H 3 should see, that the "adventure AI" in H IV could have been the best ever in the whole wide world - no one would ever have known it due to the fact that there could be no adventuring.

All of this is rather ironic, since I'm known for not liking Heroes IV. It was one of the points why I stopped playing H IV rasther soon: on the lower levels the AI didn't have the problems with the neutrals, but as a player you would flatten it. There was no AI on higher levels. What I did in the end was playing on HARD and exchanging my money and resources until I had the Impossible situation of no resources at all.

[Gaidal Cain]

Actually, even without stack bonuses, the H4 adventure map AI is terrible. If it's threatened by a human, it'll go and loose against neutral stacks far more powerful than the humans army. It can't manage daily produced creatures either. The stacks (and horrible quick combat) could easily have broken more powerful AI:s, but it wasn't needed for H4's. I've mostly played at Hard, and the h4 Ai hasn't challenged me the way the H3 one managed to do.

[Elvin]

Gothrak had some things to say on the matter.
You still believe it could have been the best in the whole wide world?

Edit:
Referring on the artificial intelligence playground at the end of the page.

Mod note: fixed the link. GC

[Khelavaster]

I still maintain that neutral stacks were at most just a part of the problem.

Here's a test that could be enlightening, if anyone feels up to the task. Create a map with the editor with no neutral stacks. You'll see that the computer acts braindead nevertheless. And again I refer you to the example in my previous post.

EDIT: NEVER MIND!! Please all read the HOMM IV playground Elvin is quoting in the post above. It's absolutely BRILLIANT. I never had the patience to go that far. That post should be made a sticky for generations to come.

What I did in the end was playing on HARD and exchanging my money and resources until I had the Impossible situation of no resources at all.

See, that's unacceptable in my book. If a game requires that for me then it's unplayable (which HOMM IV was). Again this makes me understand more why you're so happy with HOMM V (whose AI is, for all its faults, a lot better than IV).

Khel.

[Alamar]

Actually there are NONE (boner skills) in H 5. The only thing here is that the AI in H 5 could pick more abilities instead of skills. The bottom line is, however, that the the Heroes in H 5 are lightyears better than those in H3.

You made my point for me. The AI is better because [credit to Nival design] having fewer skills but making sure that most of those skills are "good" helps the AI.

I still say that a L19 necro with only basic light magic is broken. Wizards without ANY magic schools also rank up there as silly AI mistakes. IIRC one poster mentioned a fully developed Deleb without the ballista skill.

IMHO I rank the H3 ability to pick skills the same as H5 ... both pretty much stink.

What you say about 10 years of resolving things is not applicable because the developer changed, and the game changed. So that point is completely invalid.

I disagree.

Changing developers does not change the fact that PCs have more memory and faster CPUs. The OSs can handle more threads / processes/ etc. at one time with fewer problems. The possibilities for a better AI are there. This doesn't count possible enhancements in general "scripting" knowledge.

In addition picking good skills for a hero is not a "hard" problem. Heck as far as I'm concerned I'd rather the AI have fewer bonus resources and let it "cheat" to pick skills that would be more helpful [even if that means some form of scripting]

We seem to play a different game. Just attack any stack of neutrals and switch to auto-combat. You'll see the AI wait a lot. It doesn't wait in full-out battles very often, but the reason for that is that it's not often paying to wait. You should maybe select a duel and let the AI fight it out, that's maybe enlightening.

I must admit that I don't often fight in that manner. Usually it's me against AI controlled neutrals OR me against an AI controlled army.

I still contend that I don't see waits as often as there should be.

Spellcasting is massive - if the AI has them. Again, try duel mode. I agree, though, that getting high level spells and using them in combination with abilities should be higher prioritized. Apart from that, the AI handles casting lightyears better than the heroes in H 3.

I think that H5 spell casting is about the same as H3 IF you factor in the differences in skill systems.

As I mentioned before it took me a month or two before I saw the AI cast its first Meteor Shower. At least in H3 I could see those spells go off even if they weren't backed by "Expert Earth" or similar.

H5's reluctance to take perks [as we agree should be tweaked] leaves us with H3 & H5 in a similar state in regards to mass spells.

Adventure AI
I'll give a HUGE edge to H3 here. H5 STILL doesn't pick up artifacts, flag mines, etc. It's worse than a chicken with it's head cut off.

If it doesn't pick up artifacts I really wonder where those are coming from I always get from the AI heroes I kill? Also I wonder who had those mines flagged when I enter AI land?
You are flogging a dead horse here.

I have to wonder why [when I stroll through AI controlled territory] that I see resource piles & artifacts laying around that H3 would never have left around.

Yeah, H 3 is the big winner. That's why you lose duels if you are not playing sharp, that's why you can lose even on hard if you abstain from too much reloading in H 5 and that's why you wipe the floor with the AI on each and every difficulty level in H 3. Because the AI is so refined and subtle and well designed in H 3 while in H 5 it's a primitive, dumb cheater that swings its big money club without finesse.

H3 AI is not perfect. Your last sentence does sum up things nicely though.

In my own words the challenge from H5 only comes from its ability to:

1. Dominate a player economically the first month or two.
2. Fight nuetral stacks with taking very low losses that a human [often] couldn't achieve.