Question about AI Quality

[MistWeaver]

Ah,but without free gold when you capture gold mines and/or crucial cities you indeed hamper its ecenomy.With free money,theres no use for doing that.

Exactly.

[Mirez]

when I just got the expansion I was like WHOOT DWARFES
since I din't know anything bout them I set the AI on easy
it wtf rushed me, with a lotta effort I could defend my castle
in late game I > AI a lot and I was about to attack his main heroew, big battle big battle yeay yeay (joy)
he retreats the asap :'(
the day after I took his castle cause there wasn't anything in it

so I guess the AI don't really get smarter but just get more stuff and retreat more
and THE bug also returned in the expansion, the AI takes forever to make his turns
so In my opinion: the AI sucks

[Gaidal Cain]

Ah,but without free gold when you capture gold mines and/or crucial cities you indeed hamper its ecenomy.With free money,theres no use for doing that.

Not quite. If the AI has several towns, you'll actually hamper it's economy more effectively by capturing towns/mines if it has a fixed daily bonus than if it has cheaper buildings. I do agree that both bonuses were too much though.

[DaemianLucifer]

Not quite. If the AI has several towns, you'll actually hamper it's economy more effectively by capturing towns/mines if it has a fixed daily bonus than if it has cheaper buildings. I do agree that both bonuses were too much though.

Thats only because bonuses are too high.But generally speaking,a cut in the price of things is better than a fixed daily income.

[Alamar]

33, 50 and 66% cheaper stuff is definetly cheating.

Compared to free 10k its much better.

IMHO it's a wash either way.

A free 10K -> Total of about 14K / day with Capitol.

1/3 costs ~= 4K * 3 == 12K / day.

Personally I prefer the "free 10K / day" option because it's much easier for a map maker to decide to take away some of the flat 10K instead of having to figure how how much the AI makes then take away a fraction of that.

Actually 1/3 costs are MUCH worse when each AI faction has more than one castle / gold mine / etc. ... it's only a wash when each AI has only one castle and no extra funds ...

Either way we're looking at virtually unlimited funds ...

-----------To avoid double posts --------------------

Compared to free 10k its much better.

IMHO it's a wash either way.

A free 10K -> Total of about 14K / day with Capitol.

1/3 costs ~= 4K * 3 == 12K / day.

Personally I prefer the "free 10K / day" option because it's much easier for a map maker to decide to take away some of the flat 10K instead of having to figure how how much the AI makes then take away a fraction of that.

Ah,but without free gold when you capture gold mines and/or crucial cities you indeed hamper its ecenomy.With free money,theres no use for doing that.

The AI without extra mines already has basically UNLIMITED funds so your taking of its mines really isn't doing ANYTHING to hamper the AI economy because it has more $ than it could possibly spend anyway ....

[Campaigner]

Does it get resources for free to or does it trade gold for them? And the trade prices are the same for human and A.I huh?

Hey webmaster. The message about exceeding CPU quota is back.

[DaemianLucifer]

The AI without extra mines already has basically UNLIMITED funds so your taking of its mines really isn't doing ANYTHING to hamper the AI economy because it has more $ than it could possibly spend anyway ....

Depends on the map size,the day you rush in,and the development path it took.And I already did say that the numbers are too high,so its not about the system but its implementation.Decrease of 50% of the price should be the highest decrease it gets.

[Alamar]

Does it get resources for free to or does it trade gold for them? And the trade prices are the same for human and A.I huh?

Hey webmaster. The message about exceeding CPU quota is back.

I have no evidence to support it but judging from castle build rates I would THINK that both $ and resources are at 1/3 cost. Combine this with the day 1 bonuses that the AI gets on impossible and it will have virtually limitless funds.

[Jolly Joker]

Frankly, the more I read about this the less I understand the fuss about the AI having a monetary advantage on other than normal diff levels.
No one complains if you play a single-player map set up so that the player has to fight against a much more powerful opponent.
And that stuff about taking mines and "besiege" the AI, is someone really taking this seriously? I mean, if you are in enemy territory with your main hero you are running the danger to get victim of a surprise attack, if you are playing against more than one opponent, so IF you are on the march you better hurry and get things done one way or another. This is not H 3 where you could beat stronger forces with no losses and Town Portal anywhere you like.
Now, if you've followed discussions about playing Academy or other towns effectively and depending on the situation and the resources needed, do you really think you can make the game heroically difficult by not making things foolproof for the AI? Remember H IV and its inability to get things accomplished there.
On the other hand, if the AI would really play aggressively all the time, how are you supposed to win for example a six-player mp map? You struggle along and manage to beat the first opponent (who has a bigger army then you) with heavy losses. Now what? You have still four opponents (if they didn't fight it out, yet), but your forces have been massively reduced AND you have two towns to guard now, while still having only one good hero. With the won artifacts you may have a medocre second one, but the situation may still be hopeless.

So for me on any multiplayer map it's neither fish nor flesh. If the AI is not getting advantages it will be a pushover, that much is for sure, because it is impossible to make an AI that will adapt to everything and do it as well as a good human player.
On the other hand, if the AI is no pushover, mostly due to having the resources things may get very ugly if all the AIs make good use of their material superiority. With the limited movement spells, sttacking is a rather risky business.

[DaemianLucifer]

Frankly, the more I read about this the less I understand the fuss about the AI having a monetary advantage on other than normal diff levels.

Who mentioned that?The AI gets the advantage on normal as well.Or at least it did.Was it removed or not?On normal,you should play only against a good AI,not a cheating one.If its possible to make an even better AI,put it on a harder difficulty over the cheating one.If not,well a slightly cheating AI is ok on higher difficulties,but not on normal.

[Qurqirish Dragon]

Frankly, the more I read about this the less I understand the fuss about the AI having a monetary advantage on other than normal diff levels.

Who mentioned that?The AI gets the advantage on normal as well.Or at least it did.Was it removed or not?On normal,you should play only against a good AI,not a cheating one.If its possible to make an even better AI,put it on a harder difficulty over the cheating one.If not,well a slightly cheating AI is ok on higher difficulties,but not on normal.

I disagree. If there are additional skill levels, all cheating skill levels should be higher than all non-cheating ones, and cheating difficulties should always play with a full-strength AI (full-strength meaning it uses all legitimate tactics available). If such is the setup, I would always play on the highest non-cheating level.

This is a problem I had in H3 - in order to play with an unfettered AI, you had to play on at least hard (and even there, there were limitations imposed- hero count, for example), and then the resources became unbalanced, and the AI got bonus mine production. There was no skill level that allowed you to play against a full-power AI that didn't cheat.

The same is true in H5 - if you want the AI to be able to build all dwellings, and build every day, you have to play at a level that includes cheats.

I think there should be 2 difficulty settings- one for economy, and one for AI. The AI setting lets you choose between different handicap levels of AI, with the 2nd from the top being unfettered, and only the top level adding in cheats for it. The economy setting lets you choose the relative amount of starting resources between human and AI, again with only the top level giving the AI bonues resources or discounts. Thus, if I want an easy AI, but with me having no starting resources (say, you want practice on development in resource-poor maps), I can do it. I can also have a strong economic advantage against a killer AI (say, for battle practice- you want to be able to get any force you want to field)

For those who want additional challenge, they should also add in a "king-of-the-hill" setting (AFAIK, only Heroes 1 had this), which basically makes all AIs go after the Humans.

[DaemianLucifer]

Sorry I misswrote there.I did mean what you said:A noncheating AI first,and then the cheating one.Or,as JJ said somewhere else,separating the AIs skill with its starting conditions and possible handicaps would be the best imo.

[Jolly Joker]

Frankly, the more I read about this the less I understand the fuss about the AI having a monetary advantage on other than normal diff levels.

Who mentioned that?The AI gets the advantage on normal as well.Or at least it did.Was it removed or not?On normal,you should play only against a good AI,not a cheating one.If its possible to make an even better AI,put it on a harder difficulty over the cheating one.If not,well a slightly cheating AI is ok on higher difficulties,but not on normal.

What are you talking about?
For starters on normal the AI starts only with what a human gets on heroic (10000 Gold plus res) while human starts with 30000 gold and res. Furthermore the AI builds only every second day. So who cares whether the AI pays the full price or not? It's massively handicapped anyway, and even if it would pay only a tenth it couldn't make use of it.

[zuraffo]

The AI is horribly weak, even in Heroic. They'll have more troops, but that's about it. They neither use them well nor the spells well. So far I have only had problems with A.I when I am using dungeon (although lately I have improved on that by taking money from chest... dungeon troops and buildings are just too darn expensive). If I am using Academy against AI it would be a pushover. Undead was easy as well. I am not much of a Sylvan/Haven/Demon Player so I won't comment, but so far I think Heroic AI is manageable.

[DaemianLucifer]

For starters on normal the AI starts only with what a human gets on heroic (10000 Gold plus res) while human starts with 30000 gold and res. Furthermore the AI builds only every second day. So who cares whether the AI pays the full price or not? It's massively handicapped anyway, and even if it would pay only a tenth it couldn't make use of it.

So your saying that its ok to give AI bonuses on normal because it gets handicaps as well?I thought processors and AI algorhitms were advanced enough so you dont have to base the difficulty of a game by different degrees of cheating/handicaps.

[PhoenixReborn]

So how did Fortuna decide?

[Shauku]

This failure to retreat thing sounds like a serious flaw.

This is still one of my biggest concerns with the AI. Why oh WHY do they never flee with their level 15 Hero. I doubt any half reasonable player would kill a level 10 Hero just to get a few more Pixies down with their last 3 Squires.

[Humakt]

This failure to retreat thing sounds like a serious flaw.

This is still one of my biggest concerns with the AI. Why oh WHY do they never flee with their level 15 Hero. I doubt any half reasonable player would kill a level 10 Hero just to get a few more Pixies down with their last 3 Squires.

They certainly have fleed from me, at least in the HoF campaigns.

[DaemianLucifer]

They certainly have fleed from me, at least in the HoF campaigns.

But on what difficulty?

[Blaze85]

In custom games they never flee or retreat. This is my biggest problem too with the AI.
It's surely difficult to make the AI flee when it hasn't got enough creature to survive my attacks until it's next move. Maximum three lines in the program code as i see.