How much AI cheating is "OK"?

[Alamar]

If I'm reading a lot of posts correctly it seems like the AI gets huge resource bonuses on even the MP maps that don't have special AI scripting. I'm not sure of the levels of the bonuses they get but according to some people it seems to be GIGANTIC.

My question is just how much AI cheating is OK?

IMHO a certain amount of cheating is just fine. On the other hand I do want to have the illusion that the AI needs to do the same sorts of things that I do to become strong.

I would think that on "impossible" mode the following combination of "cheats" would be fine for me:

1. If the AI hero isn't customized/scripted/specified in some manner then you should add number of STANDARD starting troops for their alignment in addition to what the specific hero comes with. For example if a "average" Academy hero comes with X Gremlins and Y Gargoyles then add those X Gremlins and Y Gargoyles to whatever your starting hero's army comes with normally. On average this should roughly double the strength of a starting hero.

2. Assuming the AI starting hero is random then make sure that the likelyhood of them getting a STRONG hero [hey we know some heroes are better than others depending on map size and other factors] is MUCH higher than getting a weak starting hero. If the AI starting hero is specified then at least make sure the hero(s) in the tavern tend to be strong.

3. Give the AI some percent bonus [possibility to use one or more sliders to change the amount & types of bonuses?] to many of the things that they do. For example resources collected, output from mines, gold generated from mines/castles, creature stats vs. neutral stacks, hero stats vs. neutral stacks, mana regeneration, etc.

4. The AI should know everything that there is to know about the map. Hero locations, hero skills, army strength, what spells will show up at certain levels of spell towers, location of castles, etc.

To me [certainly on impossible difficulty] it would seem that the above is fair and possibly even necessary.

What does everyone else think?? What sorts of suggestions for improvement do you have? [Please only refer to behaviour on the unscripted MP maps!!]

[ioticus]

Cheating is OK as long as you have the option to play against a non-cheating AI. My problem with Heroes 5 is that is appears to cheat even on Normal level.

[DaemianLucifer]

How about instead of cheating we make an AI learn?I know,it is hard to make a suitable learning system,but a library of knowledge that gets regulary updated is a bit easier.If a player uses area attack spells often,AI learns to split forces;If the player relies on weak scouts to capture mines and cripple the economy,AI sets guards;etc.An AI editor would also enable fans to make AI even stronger.

[Alamar]

An AI editor would also enable fans to make AI even stronger.

I would so KILL for an open source AI / AI editor. As much talent exists with some [many?] heroes fans and as much time as I'm sure they'd put into it I'd be willing to bet that the AI would be light years better than what some people say we have now.

[Jolly Joker]

I don't consider getting more resources and money cheating. You can give a moron as much money as you like, it won't give you a hard time.
My first point here is the following: the AI CANNOT RELOAD! How many of you play single player without EVER reloading? So if the AI goes into a really tough fight and loses half or two thirds of the army AT ANY POINT OF THE GAME it has basically lost (and a human player meeting that AI a week or two later will say, bah, the AI fragged up. A human in sp modus will simply say, drat, play on, reload or start anew.
So the AI gets more money and resources - fine by me. It can play aggressive that way. Hire troops and kick neutrals. Build up a hero. The neutral heroes I met in normal games were nice to play against. They had a good level and a good army and they did give a decent fight. I have to rehire a neutral hero as yet to have a lokk at development - if that's okay, the basics are right.
So the AI simply should pursue the normal game goals. It SHOULD flag mines; it SHOULD kill stacks; it SHOULD chain or transfer troops via spell; it SHOULD fight decently on the battlefield.
And it SHOULD get a boost in money and resources to build and buy. Imagine it would get the ability to RELOAD.

[Joskevermeulen]

I agree with ioticus.
But there will probably be a good reason why there isn't such an option in HeroesV at the moment.

For example it could be that the A.I. at the moment just doesn't give much of a challenge without cheats. So i will add a little modification to your sentence.

Cheating is OK as long as you have the option to play against a non-cheating AI and that in that case the A.I. still stands a chanse to win the game in that case (for example against beginners that are learning heroes).

If the A.I. doesn't have a chanse to win the game because it isn't programmed good enough this makes a very boring gaming experience and it is better to keep it out.

But of the tbs games I already played almost all of them had an option to play against the A.I. without it needing big cheats integrated in the first version of the game that was challenging for a beginner.

I already asked this in another topic but didn't get much replies and the topic is dying away. Is the A.I. any good in the non-scripted maps?
Does it react in an intelligent manner or does it just stand and wait in its castle using cheat bonusses to raise its armies?

[Grumpy Old Wizard]

Any amount of AI cheating is ok as long as:
1) it results in an appropriate challenge for the difficulty level selected.
2) it can be regulated by the mapmaker. The mapmaker should be able to control the bonuses that the AI recieves in order to be able to regulate the flow of the game for his map. We will learn if this is possible when the editor is made available to us.

GOW

[DuRieL]

Any amount of AI cheating is ok as long as:
1) it results in an appropriate challenge for the difficulty level selected.
2) it can be regulated by the mapmaker. The mapmaker should be able to control the bonuses that the AI recieves in order to be able to regulate the flow of the game for his map. We will learn if this is possible when the editor is made available to us.

GOW

I agree, the means are not important, it's the end challenge. Regarding the AI's controllability, I think in H5 a mapmaker will have greater control of the AI than in the past, due to the scripting functions. Though I wonder what kind of scripting interface will we get to see?

[MistWeaver]

Cheating is OK as long as you have the option to play against a non-cheating AI. My problem with Heroes 5 is that is appears to cheat even on Normal level.

I agree with that.

And as for ways of cheating, I think AI can get more income from towns / mines and get a more xp from chests/battles

[Joskevermeulen]

In reply to:
I agree, the means are not important, it's the end challenge.

What do you mean with the end challenge? I hope you don't mean that the end challenge is just to capture the one castle of the enemy in a small map where you just have to wander from point A to B and the others best hero just sits still in its castle. Because in my opinion a good heroes game has to deliver challenges continuously with strong wandering heroes and not only when taking the enemies castle.
About the scripting:
If we have a Huge map with 10 castles and 6 factions/colors were at the beginning of the game you can choose your faction and type of town and the ais are randomly given a faction/town.
Were all the adventure map creatures are randomly created (sometimes normal/shooters/flyers)
I think scripting the A.I. with a mapeditor in such a case is very very difficult if you don't want to give it very high bonusses to cheat with. Because of all the different random possibilities. It is the A.I. itself that has to detect which path is the best to follow according to its environment.
I think the biggest scripting must happen in the basic A.I. It has to be very well written.

I know in Heroes III the A.I. heroes had profiles that you could see in the thievesguild: Explorer/Builder/... and it really acted like its profile. A map could have a very different outcome if an enemy hero had begin as a explorer or as a builder.

[DuRieL]

In reply to:
I agree, the means are not important, it's the end challenge.

What do you mean with the end challenge?

I mean the level of challenge that is intented by the mapmaker for his map. Thus, it doesn't really matter what helps the AI, as long as the intended amount of challenge is there. AI boosts are very map-dependant, in the end.