Suggestion for changing AI bonuses based on difficulty level

[Alamar]

After playing H5 for a while it is clear to me that the AI needs lots of help in order to be competitive in long games. However it's also clear to me that the AI gets far too much help esp. on small and resource poor maps.

Instead of the AI getting fixed bonuses I suggest that it get bonuses based on the actions of the player and the income [in various forms] that the human players have access to.

Example: Let's say you're playing on heroic difficulty level. You start with 10K gold. With my system the AI would start with 10K gold * "cheat factor". If the player got 1K gold every day then I would have the AI bonuses set up to make sure that it gets AT LEAST 1K gold * cheat factor daily. If the AI gets more gold than the player [like more chests found] then good for the AI. If the AI can't keep up with the player then the cheat factor comes into play to make sure that the AI gets at least 1K * cheat factor per day.

Obviously the size of the cheat factor can be tweaked across various difficulty levels. For example on "easy" mode the cheat factor would be less than 1 so the AI may well lag -- but it should lag so no problem .... on heroic mode the cheat factor would be larger than 1 insuring an AI bonus.

The same sort of principle could apply to ore, wood, precious resources, etc. A reasonably fair set of bonuses could be applied to the AI no matter what faction was used so there wouldn't be too much of a problem.

I believe that this could lead to:

1. An AI that doesn't unfairly crush you on small and/or low resource maps.

2. The player can play more naturally because they don't have to live in constant fear of being caught way out of position by a vastly superior force.

3. The AI bonuses would self adjust to the play style [and to a degree skill] of the player. The faster, better, more agressive players just end up making sure that the AI gets more bonuses faster. The slower, less aggressive players don't give the AI bonuses as fast.

4. The illusion that the AI is playing by the same rules that we are.

5. The AI won't be as big a factor deciding MP matches between 2 players on a fully populated 6-8 player map.

Any feedback [aside from the obvious "it won't happen"]?

[Corelanis]

I think that is a really great idea.

[Orfinn]

Im sooo agree here, especially with point 1 and 2.

[PhoenixReborn]

1.) How does the AI cheat factor work now?

2.) My experience has been the exact opposite. I can always win on smaller maps, but on bigger maps I face huge numbers that I can't possibly beat.

Nevertheless, you are right, something needs fixing or balancing. There is or was an "A.I. doesn't cheat mod" on the ubi forums which takes away computer bonuses...

[Alamar]

1.) How does the AI cheat factor work now?

The idea is to make sure that the AI gets bonues that are directly tied to what the human players are doing.

For example on "hard" difficulty level you may decide that the AI needs a 10% bonus in $, resources, etc. So on day 1 if a player gets 20K gold to start with then the AI would get 20K * 1.1 for a total of 22K gold.

Also lets say from from day 1 until day 5 you earn or find 10K gold. Lets say that the AI fairly found 5K gold during that time. Obviously the AI needs some help to stay competitive with the human. Basically you just multiply 10K gold by 1.1 [equal to the cheat factor] and that totals 11K gold. You then need to give the AI 6K gold so it can stay competitive.

A similar sort of formula could be set up for ore, wood, precious resources, etc.

2.) My experience has been the exact opposite. I can always win on smaller maps, but on bigger maps I face huge numbers that I can't possibly beat.

I guess that instead of smaller maps I should have said "less large maps". My experience indicates that the AI's power surge comes when they can afford to buy all their L7s (maybe even upgraded L7s) and I am struggling just to buy out L1-L6. This period of a power surge seems to occur between weeks 3-6. If I can last until week 7 it's usually no problem as I'll out play the AI if my army is only somewhat close.

Nevertheless, you are right, something needs fixing or balancing. There is or was an "A.I. doesn't cheat mod" on the ubi forums which takes away computer bonuses...

The No Cheat mod only took away the AI's scripted bonuses. In addition to the scripted bonuses the AI also gets sizeable hard coded bonuses that are quite large that the mod doesn't effect.

Just to be clear I don't really want to eliminate AI cheats. On heroic mode I WANT the AI to cheat because we're not going to get an AI that can keep up with a human player in the long run. What I want to make sure of though is that the AI cheats in a manner that doesn't degenerate into an unfair scenario during weeks 3-6.

[Campaigner]

OR those lazy bums could program a better A.I! Problem solved

But that's not gonna happen....It'll NEVER be as good as the A.I in GalCiv II anyway....

The funny and sad thing is that this suggestion will maybe be read by the ones in charge but they wont listen to it....not one bit....

[vicheron]

The computer should just get a percent bonus on the treasures it finds and on its income so on a map like Subterranean Treasures where a certain resource is rare for the first two weeks or so, the AI doesn't get its town fully built at the end of the first month.

[PhoenixReborn]

Thanks for the clarification Alamar, your idea is much clearer to me now.

In addition, there are occasional lapses in combat A.I.

Now that I've bumped up to hard they use higher level spells so that is fine, but occasionally they will make weird moves to block my shooters without striking and then I just clobber them with my guard.

I haven't noticed the misuse of winter ring since the patch...

[Alamar]

I agree that there are some lapses in the combat AI. I know that it sounds odd but I'd rather fix the other AI elements first and fix the combat AI last.

As long as both are done though I imagine that it would be fine.

[omegaweix]

IF the change would work the way you suggest.... it would be simply amazing!

Especially point 1 and 2 are big issues for me!
It happened sooo often, that i crushed two A.I's so easily that it wasn't any fun at all but third and last was so strong that I often despaired after seeing it's ridiculous army teleporting near my main castle... although he obviously didn't claim ANY treasures or ressources in his area (the "neighbours" map is a good example for that one)!

This would surely be a game enhancment to dream of (beside of the all time desired map editor forheavenssake) !

edit:
oh yes, and could this tweaking also prevent A.I. players being blown away by anothe comp in week 2 or three?
I always try to figure out how that can happen, even on very large maps like "dragon pass"!

[Alamar]

IF the change would work the way you suggest.... it would be simply amazing!

I believe that the AI would probably need to improve for my changes to work as intended but I believe that it's possible!

If a system similar to mine is used we should at least see the end of the AI buying out the best troops from 3 castles and throwing them all at you at once.

edit:
oh yes, and could this tweaking also prevent A.I. players being blown away by anothe comp in week 2 or three?
I always try to figure out how that can happen, even on very large maps like "dragon pass"!

I doubt that the mod that I suggested would help too much with that. I imagine that an AI getting totally taken out in week 2 might be delayed until week 3 simply because the AI may not be able to afford to buy out all the troops required to blow through guardian stacks in order to do that ...

[wimfrits]

I doubt that resources are an important factor in driving the AI challenge factor. Amount of creatures, level of heroes and path choosing are a lot more important imo.
But for the sake of argument let us assume that resources are a limiting factor to AI performance.

The problem with a system such as proposed I think lies in the nature of playing on higher difficulties. The higher the difficulty, the more the player needs to be able to find edges. A player might go to great lengths to get that gold mine on day 4 instead of day 10 to be able to build up his town faster and thus gain an edge on the AI opponent. Or sacrifice all of his lvl3 units to be able to get that bag of gold; where easy level players would wait until being able to pick it up without losses.

With a system that adjusts to players resources/income, this effect of edge-finding behavior is neutralized.

In other words, if AI resources are a factor determining AI performance on different difficulty levels, then the ability of the player to prevail on higher difficulty levels partly depends on gaining edges in the field of resources.
If each advantage the player gains is immediately neutralized then resources are no longer a factor relevant to difficulty.

[Jolly Joker]

Alamar, linking AI bonusses with the state of the game human(s) are in, is a bad idea. Why? Because that kills the necessity to go for money and the money dwellings. In short, it gives the player too much control over the AI behaviour. Why should the AI follow the human strategy or being dependent on what the human(s) is (are) doing?
I cannot understand the fuss about the "resource cheating" anyway. I mean, why should things be completely equal anyway? That certainly isn't so in reality. One realm would have more wood, for example, earn more money, whatever. I prefer to imagine an AI player as "rich but lazy and a bit dumb". So where's the problem? Your neighbours are sitting on their arses and earn themselves a golden nose - well that happens, don't you think? Consequently you have to struggle to make up for it. Luckily enough said neighbours are no strategical and tactical geniusses.
Now, the point is that the AI has so much of an advantage on small and resource poor maps. That's true - it gets more difficult. But it's obvious that it is (and the overall balance between factions will suffer as well in MP), so making a small map resource poor is making a difficult map against the AI; you can easily change this by adding more resources.

[Alamar]

I doubt that resources are an important factor in driving the AI challenge factor.

I suspect that with the issues that I'm trying to solve that it is 100% tied to how AI bonus [cheat] resources are structured. Basically I'd like to see the bonuses restructured so that the AI doesn't have the capability to:

1. Blow out another AI or two very early.
2. Finish building all 3 castles [with upgrades]
3. Buy out all 3 castles to attack you

This issue is especially bad in heroic difficulty from weeks 3-6 IMHO.

Amount of creatures, level of heroes and path are a lot more important imo.
But for the sake of argument let us assume that resources are a limiting factor to AI performance.

I agree that the above are more important for the long run challenge and will make for a "better" AI.

However my suggestions weren't done to make the AI more competititve per se ... they were made to tone done some of it's outlandish economic behaviour.

The problem with a system such as proposed I think lies in the nature of playing on higher difficulties. The higher the difficulty, the more the player needs to be able to find edges. A player might go to great lengths to get that gold mine on day 4 instead of day 10 to be able to build up his town faster and thus gain an edge on the AI opponent. Or sacrifice all of his lvl3 units to be able to get that bag of gold; where easy level players would wait until being able to pick it up without losses.

The way I see how hard or heroic difficulty SHOULD work is that the AI should always be at least 1 step or two [as opposed to 20 steps] ahead of you in order to provide the proper level of challenge.

With a system that adjusts to players resources/income, this effect of edge-finding behavior is neutralized.

Which is exactly the effect that I was going for!!

No matter how good the player is, how many edges they use, no matter how many exploits [unintended edges] they abuse, etc. the AI will at least be able to keep up with them in some respects to provide more of a challenge throughout the game.

Certainly this would be something that would be good for heroic difficulty right?

In other words, if AI resources are a factor determining AI performance on different difficulty levels, then the ability of the player to prevail on higher difficulty levels partly depends on gaining edges in the field of resources.
If each advantage the player gains is immediately neutralized then resources are no longer a factor relevant to difficulty.

This is a good point.

The thing is do you want an AI whose potential is capped at X level and if the human plays better than X they will always win OR do you want an AI that has its bonuses adjusted in such a manner as to theoretically be able to give you a good game no matter how well you play.

[Alamar]

Alamar, linking AI bonusses with the state of the game human(s) are in, is a bad idea. Why? Because that kills the necessity to go for money and the money dwellings.

Actually my proposal does NOTHING of the sort.

First you will need money and other economic structures to keep up with the AI. My proposals are done to make sure the AI gets at LEAST as much as a human player. There is NOTHING in my proposal that keeps the AI from making MORE money, resources, etc. than you do.

Second you will need money and other economic structures in order to buy enough creatures to clear out neutral guardians.

Third there is no reason to expect that if you can outplay the AI with minimal resources [as you suggest] that you can't outplay the AI with increased resource levels.

In summary it pays for the human to make as much $, resources, etc. because in my proposal the AI is guaranteed to make as much or possibly more than a human would anyway so mathematically the way to keep the AI's advantage to a relative minimum is to maximize their own income.

In short, it gives the player too much control over the AI behaviour.

Actually the player has ZERO control over the AI behaviour!!!!

If an AI wants to pick up treasure, take mines, kill creatures, build in its castle it is perfectly free to do so. If the AI is more successful in gathering resources than a human player then GOOD FOR THE AI as it gets to keep EVERYTHING that it gathers.

The AI is only given a BONUS if it starts lagging behind human achievements [on in case of the cheat factor is close to the same achievements]. This insures, esp. on heroic difficulty, that the AI is ahead of the human player in economic terms without EVER being overwhelmingly ahead. [UNLESS the AI does an overwhelmingly better job than the human player on its own.]

Therefore a human player isn't limiting the AI in any way whatsoever from performing legal actions.

Why should the AI follow the human strategy or being dependent on what the human(s) is (are) doing?

The AI, IMHO, should be free to persue any human or non-human strategy that it cares to. As far as I care it can use a non-human strategy as long as it is successful and plays by roughly the same rules that humans do.

There is nothing in my proposal that would prevent the AI from using any strategy that it likes. If it does a better job than a human then good for the AI.

Also the AI does not depend [in terms of rely upon] what a human player does. There is nothing in my proposal keeping the AI from taking mines, taking tresures, building in it's castle, etc. Only BONUSES are dependant on human behaviour and those bonuses are only given out IF the AI needs them to stay competitive.

I cannot understand the fuss about the "resource cheating" anyway.

Hmm ... [possible flame deleted prior to original post]

The fuss is that the AI gets such HUGE bonuses in the current version that it could easily steamroll you right out of the game if it chose and you wouldn't have much of a chance.

If the AI bonuses were restructured in such a manner as to where they could still be ahead of a player [to provide challenge] while not being overwhelming wouldn't that be a step forward???

I mean, why should things be completely equal anyway? That certainly isn't so in reality. One realm would have more wood, for example, earn more money, whatever.

In reality other nations/factions would more readily be able to get wood, iron, etc. However for HoMM I prefer the simulation of reality be left up to the map maker [give a wood bonus here, a sulfer bonus there, a gold mine here, an ore mine there, etc.]. I will trust the map maker to insure that the map is fun & fair.

However, for balance reasons, in HoMM I prefer each faction to start out on roughly equal footing.

I prefer to imagine an AI player as "rich but lazy and a bit dumb". So where's the problem?

Right there is EXACTLY the problem!! Your view of the AI as rich but stupid is fine for you. My view of what the AI should be is TOTALLY different and totally incompatible with the "rich and stupid" view.

In my "perfect HoMM" the AI would be able to pass the HoMM equivalent of the Turing test [be indistinguishable from a human] and would be a very strong player without any bonuses at ALL.

Obviously we don't have a perfect AI. Because the AI isn't perfect we will need to suppliment it with certain bonuses that would probably include some form of resource bonus to make sure that it performs as close to a human like manner as possible. If having an AI that performs similar to a human is a good goal then what better way is there other than to tie AI bonuses to what a human player is doing????

Now, the point is that the AI has so much of an advantage on small and resource poor maps. That's true - it gets more difficult. But it's obvious that it is (and the overall balance between factions will suffer as well in MP), so making a small map resource poor is making a difficult map against the AI; you can easily change this by adding more resources.

I can't believe that you're actually suggesting that.

Aren't you tired of having to go to so much effort when you make maps to make sure that it is balanced for all factions and the AI???

Wouldn't you think that it would be FAR better if the bonuses that the AI got depended largely on whether you had a "rich" map or a "poor" map?? Wouldn't auto-adjusting those bonuses be a good thing? Wouldn't that result in a MUCH better game and a MUCH larger variety of maps??? A system like the one that I have proposed takes care of those issues AUTOMATICALLY.

The only real downside that I see to my system is that the AI should be strengthened to understand the proper value of taking mines & resources, reinforcing hero armies, being able to retreat properly if needed, etc.

[Jolly Joker]

Alamar, don't you know what you are suggesting anymore? This is a quote of your post:

"Instead of the AI getting fixed bonuses I suggest that it get bonuses based on the actions of the player and the income [in various forms] that the human players have access to. "

If that isn't linking the AI with what the human player does I don't know what is. So I'm not going to answer to the rest because it seems you don't know what you are suggesting,
No offense, of course.
Alamar, I don't want to bragg or somethihg, but when I (or you or everyone else having played as much Heroes than the hardcore players) plays against a newbie or a relative newbie and that newbie will past the Turing test, but still lose miserably. What *I* want is a challenge and I don't care whether this challenge comes from the AI having doeble as much money and resources than me or starts with double as much towns or whatever. I. DON'T. CARE. Because if I play a human opponent said opponent plays different than me - and if it's worse it's not much of a challenge even though it's human thinking behind it.

[Alamar]

Alamar, don't you know what you are suggesting anymore? This is a quote of your post:

"Instead of the AI getting fixed bonuses I suggest that it get bonuses based on the actions of the player and the income [in various forms] that the human players have access to. "

Obviously you did NOT read my entire post or any of my replies. Let me highlite the critical words for you:

"Instead of the AI getting fixed bonuses I suggest that it get bonuses based on the actions of the player and the income [in various forms] that the human players have access to. "

I went on to elaborate EXACTLY what I was proposing and I believe that it would be reasonably clear if you had read my posts on the subject.

To make it easy for you let me try to summarize my proposal:

Any BONUSES [cheats] that the AI gets will be tied to how well their human opponnent(s) are doing. If the human is making more income than the AI then the AI will get bonuses to offset this so it can stay competitive for longer. If the AI makes more income than the human then good for the AI -- IT DOESN'T NEED MORE BONUSES!!!!!!!

My theory is that any additional bonuses above and beyond what it needs to be competitive will lead to situations where the AI is far too powerful and this creates too many imbalances in maps. This is what I'd like to eliminate.

If that isn't linking the AI with what the human player does I don't know what is.

Once again only the BONUSES are tied to human performance. The bonuses are only applied if the AI needs them to stay competitive.

Any other POSSIBLE behaviour that the AI wishes to engage in are NOT tied to ANYTHING that the human does.

So I'm not going to answer to the rest because it seems you don't know what you are suggesting,
No offense, of course.

Actually I believe that you choose to misinterpret what I'm proposing so that you don't have to reply in a meaningful fashion to my suggestion.

And I am offended by your post which is obviously aimed at belittling me AND my suggestion.

Alamar, I don't want to bragg or somethihg, but when I (or you or everyone else having played as much Heroes than the hardcore players) plays against a newbie or a relative newbie and that newbie will past the Turing test, but still lose miserably.

Yet again you are unwilling or unable to listen to what I have to say. To quote myself from a prior post:

In my "perfect HoMM" the AI would be able to pass the HoMM equivalent of the Turing test [be indistinguishable from a human] and would be a very strong player without any bonuses at ALL.

The key word above is AND ... the AI needs to act in a manner that a human could act AND it needs to be a strong player. Please at least try to talk about the SPIRIT of my proposal and posts instead of what you did in your last post.

I understand that you DO NOT CARE how AI bonuses are structured as long as it gives you a competitive game in the end. I believe that to a certain degree that it is fine.

However I DO CARE if the AI bonus structure is set up to where certain types of maps are virtually unwinnable against the AI. I DO CARE if the mapmakers have to go to enormous trouble to create a map that is balanced both for AI and human players.

I find it incomprehensible that you are fine with a system of this nature when fixes are possible that would allieviate this issue. I recognize that my proposal isn't perfect but it goes far further towards addressing these issues than anything else I've seen proposed.

If you'd like to propose something that is reasonably well thought out I would love to hear it. If your idea achieves my goals better than my idea trust me when I say that I'll embrace your idea.

[Jolly Joker]

Alamar, seriously.
If the game registers that the human player is ahead and "reacts" by giving the AI a bonus it's already too late. I mean, how do you expect to make a competitive AI when half of the human players (probably more) are "not competetive" for a hardcore player. Moreover, the competitiveness as such, or the ability to "play a good game" is nothing fixed or reproducable. It depends on what you face on the map, what town you have, what hero you play, what you must achieve and so on. It's the ability to do what is called for to make the most of it.
It is clearly impossible to program an AI to be competetive.
What you want is a competetive AI that would get a hard or heroic opponent with the bonusses, and I doubt that is possible.
Another thing. any harder difficulty than normal implies having disadvantages. I mean what would be the "hard" part in a game, if your AI opponent would face the same situation than you (that was done in Heroes IV and it was the worst that could happen for the game)? Answer: It is NOT harder. In fact, it gets easier, because the shortcomings of AI play will be more pronounced.
So any difficulty harder thn normal implies that the AI faces another situation than the human player. However, starting money alone is NOT enough. The reason is simple. Monster stacks are harder, and that must be offset as well. So while the AI should ALWAYS start with at least normal starting condition it would have to get another bonus as well, for example a daily income or resource bonus, added creature growth and so on.
That's the definition of harder than normal difficulty levels. You cannot seriously expect that the AI outplays a halfway experienced player without any significant bonusses, can you?

[DaemianLucifer]

I cannot understand the fuss about the "resource cheating" anyway. I mean, why should things be completely equal anyway? That certainly isn't so in reality. One realm would have more wood, for example, earn more money, whatever. I prefer to imagine an AI player as "rich but lazy and a bit dumb". So where's the problem? Your neighbours are sitting on their arses and earn themselves a golden nose - well that happens, don't you think? Consequently you have to struggle to make up for it. Luckily enough said neighbours are no strategical and tactical geniusses.

This should actually depend on a map,and not given to the AI on every map.In reality no one just receives those resources out of the thin air,they are just blessed with richer mining locations,or they know how to exploit those locations better.

Another thing. any harder difficulty than normal implies having disadvantages. I mean what would be the "hard" part in a game, if your AI opponent would face the same situation than you (that was done in Heroes IV and it was the worst that could happen for the game)? Answer: It is NOT harder. In fact, it gets easier, because the shortcomings of AI play will be more pronounced.

Why?Why cant harder mean just AI playing better?I dont remember that in chess harder means giving you less figures.

[Campaigner]

My sense of justice doesn't like that "self adjusting bonuses" if I play better than the computer. Feels useless to sacrifice troops and stuff to get that goldmine very early if the A.I will get the same amount of gold from thin air....so no, I don't support this suggestion. I'd rather have % bonuses as we had in Heroes II (but lower, 10% and 25% would be enough).

In essence, If I sacrifice troops to get important resources so I can get all my lvl 1-6 dwellings on week 1 then I want a reward for it. I should have the appropriate extra stuff that I fought for so I will be ahead of the enemy.