|
| Thread: The AI, Should it be able to cheat? |  This thread is pages long: 1 2 3 · «PREV / NEXT» |
|
Stevie
Responsible
Undefeatable Hero
|
posted August 24, 2014 03:10 PM |
|
Edited by Stevie at 16:49, 24 Aug 2014.
|
Indeed, Sal, thanks for pointing that out for me.
The other thing I was trying to say is that JJ's idea of having AI face different army sizes depending on difficulty is still a form of cheating. It's just that you don't make the AI better, you make the AI opponent's worse.
JJ said:
People want a non-cheating AI for just one reason: that they can maintain the illusion that a win against the AI is worth something. If the AI is "doing nothing", but simply gets 10 units each turn out of thin air, the illusion of "playing a game against an oppoent" is destroyed.
LoL. Because playing an even harder AI means nothing.
As I was saying, your idea is still a form of CHEATING! And obviously we don't want "10 units out of thin air", but we DO want 10 units more for the AI to recruit.
You know what, if you think that's unfair, I have an idea: PLAY NORMAL DIFFICUTLY. Since when did Impossible difficulty meant "A better human player" rather than "A harder computer to beat", huh?
The one thing I found interesting tho is that both him and Elvis talk about the necessity of keeping some sort of "illusion". I simply cannot comprehend this argument. I for one never felt the need of such an illusion to make me play a game. Heck, I'm doing Civ5 at DEITY with 1 settler and 1 scout versus 22 civs that start with 2 settlers, 4 warriors, 2 scouts, 2 workers, 5 starting technologies and almost 200% bonuses in every department (growth, income, happiness, army, trade). And I NEVER felt any need of some mystical illusion to push me through the game. And certainly that wasn't the case with Heroes games, or make it any game at all if we're at that. So to me, this argument is simply smokescreen.
And he manages to contradict himself with this:
JJ said:
Special rules are fine also: what is wrong with having to fight against more for getting the same? Nice challenge, without destroying any illusion.
Obvious illusion aside, you clearly have a problem with that "more" when it means fighting an AI with more resources, troops and discounts.
|
|
Storm-Giant
Responsible
Undefeatable Hero
On the Other Side!
|
|
posted August 24, 2014 04:19 PM |
|
|
Ideally the best would be to improve the AI so it keeps up with the player. The problem? In a TBS with so many variables, the player is going to be better under the same rules (and getting a really good AI isn't cheap, we have to be somewhat realistic and not forget Ubisoft won't spend endless money on Heroes).
The idea of the AI fighting less creatures for the same XP sounds excellent to me, definetly worth a try.
____________
|
|
fuChris
Promising
Supreme Hero
Master to the Speed of Light
|
|
posted August 24, 2014 04:49 PM |
|
|
But the illusion is important! A good AI is one that plays like a human but is simply : a,) Better b,)Gains units faster c,)Harder to beat.
a.)
Being Better means it should have a constant adventage compared to the player and the simplest way for that is for it to have an XP multiplyaer(80/100/120/140) OR the player should get less Xp per difficulty level(100/90/80/70).
I cant imagine anything that can take its place in heroes since it lacks maintenance cost and research like the Civ series.
b.)
Gaining units faster means simply being able to outbuild the player. An initial bonus in resources is enough, no need for additional profits from mines and such. On impossible it should have enough resources to build all unit production buildings without any extra resources.
Another option is that the AI should get a bonus to units willing to join it and at reduced cost, helping it to clear the map faster.
c.)Being harder to beat is simply an advanced moveset in battles-which include better use of spells and management of spellpoints- and a faster more expansive playstile makeing it able to rush faster, gain those extra catles and dwellings, makeing further use of its bonuses in resourcees.
Other things like seeing the map could be used but it shoud have to explore as well. Maybe it should have extra knowledge like the things offered by the View Earth/Air spell in H3, but not a full map.
|
|
Sal
Famous Hero
|
|
posted August 24, 2014 05:12 PM |
|
Edited by Sal at 17:16, 24 Aug 2014.
|
I created my own Heroes 3 AI, by customizing all the possible advantages. Of course it does not play better, but the advantages I give him allow to keep with my hardcore skills until the end.
LINK to IMAGE
All advantages are parameterizable, but not chaotically. For example, if the player selects the neutrals growth x2, x3 or even more, on same day level 7 bonus creatures are added to each AI army. This bonus is 2 x player growth selected value. So with x2 growth, AI from castle faction starts the game with 2 archangels. The most important, you can change all settings during the game, on the fly. Feel neutrals became to easy for your big army? Set growth higher, newt week you will face 4000 creatures/stack. Feel AI didn't explore enough of the map? Set his movement to +10 every turn, and soon he will knock at your doors.
Ok, this to say that each of us has different amount of time and interest to consecrate to a game, thus some become quite skilled and need a realistic challenge. Fixed difficulties are only for lurkers, the guy who download the game, play one weekend then goes back to WoW. 200% difficulty is a joke for those knowing the AI tricks and shortcuts. Need more.
For those who want bloody challenges, IMO only super detailed settings and options can satisfy. But my feeling is that they don't really want long life games, because this delays the next sequel purchase.
|
|
JollyJoker
Honorable
Undefeatable Hero
|
|
posted August 24, 2014 05:24 PM |
|
|
|
The sole problem with HoMM 3 AI is AI hero building - heroes end up with too many lousy skills.
|
|
Sal
Famous Hero
|
|
posted August 24, 2014 05:31 PM |
|
|
|
This is where they botched it. Skill tree for AI should be radically different from human, less randomized. A human can win even with lousy skills, but AI without natural combat skills is over.
|
|
flonembourg
Known Hero
|
|
posted August 24, 2014 07:36 PM |
|
|
JollyJoker said:
The sole problem with HoMM 3 AI is AI hero building - heroes end up with too many lousy skills.
I see another problem here! LOUSY SKILLS should not exist!
|
|
bloodsucker
Legendary Hero
|
|
posted August 24, 2014 07:41 PM |
|
|
I may not be seeing the problem right, I know nothing about AI programing, but I think whatever you do to AI you do it both to AI players and neutral armies.
So reinforced neutrals will always be hard for computer to beat, unless the simulation is unexistant and computer simply beats whatever it faces with no losses or close (this was how it worked with Etherlords 1 - you could see a level one hero attack a hard creatures dwelling you had been unable to beat with your level 7 full of tricks major hero and then came out a level 6 hero, then u attacked this level 6 with a lesser hero of level 2 and beat him with ease. It works but it looks stupid). Having good simulator, like H3 one, and letting AI player fight easy battles for full experience seams to me the right way to solve this.
Sal, I think if AI could really be made extremely good, it would had to have very different levels of difficulty or the learning curve would be so hard nobody would invest enouch time to learn how to play.
Imagine an AI playing at League of Champions level; breacking in to treasure area on Jebus at 123 or 124, attacking your castle with more then 20 wyverins and six angels in a hero with all prims above 20 day 126. How many players do u think would persist having their ass kicked over and over again until they manage to be at that level?
P.S. And who would decide what is a losy skill or not? I like water magic and sometimes disable Scouting from my maps, Maretti (who is a champion) said to me he prefered to get scouting to water. I learn archery most of the times I have a chance, good human X human players never choose it, yet it's amazing against the map. And so on...
|
|
OhforfSake
Promising
Legendary Hero
Initiate
|
|
posted August 24, 2014 09:09 PM |
|
|
About saving/loading vs. the AI.
When I play chess againt the computer, it'll let me go back to any position I like, where I can redo any of my moves, and it still trashes me utterly. It can do this, because, like heroes, chess is a Turn Based Game.
There are mainly 2 parts of this game, creeping and battling. saving/reloading during each helps the player to get a better understanding of the game, especially when it comes to creeping where a mistake can make the rest of the game pointless, which is a shame if you've already invested a lot of time.
I think the AI simply has to be better at creeping, and it's not an issue.
About pre-map knowledge
When I played heroes, there were spells which allowed me to see the map more or less.
Even then, if the map isn't random, the player could have the option of the AI knowing the map or not (prior experience).
But what I really would like to mention is what I think is the purpose of the AI altogether.
When I was a wee kid playing heroes, the AI were replacements for my friends. Sadly it'd either be too easy (normal diff.) or too difficult (hard diff.).
What I wanted was an exciting and challenging experience. The only way I can imagine for this to happen, is for the AI to be taylor made for each individual player, and the only way I can imagine this happening is if the AI actually learns.
I imagine the easiest ways to create this illusion is to let the AI cheat, so it can match itself up with the opponent, not sending an army that simply crushes him, not having so little army, that the AI will simply be crushed either.
But I think the best way would be to create an AI which doesn't have any advantages the player doesn't have, but I also think it's much more difficult than to let the AI cheat and therefore always have the outcome one wishes for available.
So ultimately, in my opinion, it comes down to how much effort will be put into this AI. It's better to have a cheating AI if it makes for a better game, in my opinion.
____________
Living time backwards
|
|
leiah2
Known Hero
|
|
posted August 24, 2014 09:30 PM |
|
|
|
I also Think the Ai should be stronger, but I don't Think to be bale to see and know excactly wherever your hero and Town is. This is something I found irritating in V and VI
|
|
Stevie
Responsible
Undefeatable Hero
|
|
posted August 24, 2014 10:18 PM |
|
Edited by Stevie at 22:36, 24 Aug 2014.
|
My proposition
The first thing we need to do when things don't go as we'd like them to is identify the problem. And the problem at hand is that the AI doesn't pose enough of a challenge to us humans. On what front does this problem occur? Is the AI weak at gathering resources or building towns? A good AI would hardly have these problems, and even if it did these problems could be solved with more coding. So where then?
Obviously, in combat. Or rather, after combat. Let me explain:
The game between you and the AI is practically over when your best hero defeats the AI's best hero, or rather when the main armies clash. At that point, the AI would be unable to pose any threat to you for the rest of the game. Why? Because after that point on the AI is at a disadvantage in army power that he cannot recover from. - THIS is the critical moment of the game in which you LOSE ALL INTEREST, because no one would see a challenge in beating an AI in a worse state than after beating it in it's best state. That's why you might as well quit, and I'm sure that you like me have done that many times.
So how do you fix that?
First off, no amount of resource bonuses or hiring discount would solve this. Code the AI with more brain power? Well, that would help, certainly, but not to the extent that it would make the AI's 1 Titan beat a human's 5 Titans.
AND - and this is where you all lost yourselves - no amount of messing around with the AI's encounters will fix this either! You can make the AI faces as less numbers as you like, it won't matter at all after you beat his main army.
So what could make the AI interesting again? What could you give the AI to pose a threat to you after his main army bit the dust? - Give him a mechanism to restore his army power faster, obviously!
Now, before you all start crying "But that's unfair Stevie! *mumble, mumble* We need our illusion!", here's your red pill - Pull your head out of the ground and face reality. You cannot have a challenging AI unless the AI CHEATS! So you either have to chose between 100% parameters for both you and the AI - which is what FAIR actually means - and have a crappy game, or you let the AI cheat.
Oh and let me clear another of your misunderstandings. Impossible level means impossible. Yea, you heard me, that's what it means. You can't say "I want impossible level" but EXPECT to win. No, when you set impossible, you expect to LOSE. That's what's fun about it. So with that in mind, let me present you with what I believe to be an accurate translation of difficulty levels:
Easy - either new to the game, or retarded;
Normal - has a working brain;
Hard - requires very good game knowledge;
Expert - challenging, requires perfect skill;
Impossible - brutal game, win 10/10 games on expert then come here;
Now, let me explain my thoughts in regards to the AI army size, and the way they can recover their loses.
Let's call this the AI's Kingdom Army Strength:
Easy - 100%
Normal - 100%
Hard - 150%
Expert - 150%
Impossible - 200%
1. We can make the AI outright produce more troops than base growth + fortifications. Meaning that Easy and Normal levels won't be affected at all, while Hard, Expert and Impossible will produce with 50% and 100% more.
This tho, seems a bit right off the bat. But there are advantages to raw troop growth injections because it wouldn't be too hard to implement. And you can keep track of a constant more easily. However, if left as such, this acceleration of growth cannot be stopped or regulated.
Or
2. We can make the AI's growth accelerate to that threshold by dividing it. Meaning that when the Kingdom Army Strength of the AI would be less than the 150%/200% of your Army Strength, the AI's weekly growth will receive an injection of troops until it achieves that threshold after which it stops. This division could be done in 2 ways:
A) A fixed amount of troops - which means that the respective growth is limited by a maximum, like there won't be more than say 5 Titans per week - unless week of titans, but that's an exception;
B) The amount of troops is divided into percentage per week, rather than a fixed number of troops per week - which means that depending on how big the players's army is, the AI could get well above 5 Titans per week if the percentages require it - say if your army strength consists of 50 Titans, and you have to achieve 150% of that in 4 weeks (1 month), that would mean 75/4 = 18 titans in a week, so an injection of +18 Titans added to the weekly growth. Now I know that sounds a lot, but 50 Titans sound awfully a lot more! Without such an injection that AI wouldn't have ANY way of recovering.
This however would require some amount of coding... And even so there are quite some variables at play. Like multiple towns, neutrals joining your army which would automatically increase the opponents growth, things like that...
After the threshold is achieved - AI has 150%/200% the Army Strength that the player has - there will be no other troop injections, meaning that the AI's growth will revert back to 100% until it needs to reach that threshold again.
In fewer words - Accelerating the AI's growth then decelerating.
It's a pain! But I'm trying to come up with solutions, been brainstorming about this all day and it still looks like quite a mess. Tried to put it in the best shape possible. Please appreciate.
|
|
alcibiades
Honorable
Undefeatable Hero
of Gold Dragons
|
|
posted August 25, 2014 06:35 AM |
|
|
leiah2 said:
I also Think the Ai should be stronger, but I don't Think to be bale to see and know excactly wherever your hero and Town is. This is something I found irritating in V and VI
Agreed, that is indeed exactly what I call cheating and should be avoided.
I don't mind AI starting with more resources on higher difficulty levels. Yes you will face early charges, but that's kind of what you sign up for by picking the higher difficulty level. The AI facing weaker neutrals also could work fine, again that would not be cheating but just loading the stacks in favor of the AI, which is what you sign up for on higher difficulty levels.
H3 AI just picked any new skill that was offered to it if I'm not mistaken. That is, of course, very poor AI programming when the game is shock full of crap skills (in itself a problem). AI should be able to select certain skills, preferably lock into a pattern after the first couple of skills are opened - for instance, if Hero unlocks Fire Magic early this should add a positive favor for Sorcery, to use H3 language.
|
|
Dave_Jame
Promising
Legendary Hero
I'm Faceless, not Brainless.
|
|
posted August 25, 2014 08:28 AM |
|
|
You know Stevie your idea somehow reminds me of my creature cap purposal I made a long time ago to compensate for that evil thing called creature pool. But at the same time I remember one Heroes game where beating the Oponent's main army was not the end of the game, and certainly not the point where you have won and could walk away from it.
The game had a nice charm, was inbalanced as hell and is often overlooked. The game was Heroes II.
Now I will tell you why, when you defeated the oponents main hero you did not have the "You win feeling" yet.
1: No Primary castal/No town conversion: Defeating the opopnents main hero did not mean that we killed his entire army. The fact that no castle had a primary status (H3+) and that a castle could not be converted (H6), therefor the troops produced there were an anrmy of its own, resulted in creating small forcec that would eventualy grow, and could be later used as a backup force, once that evil player took our main Island. Also it often made you considere "Do I really need those archers? They keep falling like flies and there is a large groupe of gargouils in this Vilage.
2: Only 6 Units in a castle: Lower number of units and almost no requirements in building schems made it easy to build a working second castle that producec units. The bigest limit was not your developement, but your resources, and here is what motivated you to go on, while for the PC this was an easilly adjustable advantage.
3: The fact that the game was inbalanced: Not good for Multiplayer when the factions are balanced more around the fact how big the map is, rather then combat strenght, but still this help the game to keep interesting even in later game, with the easy building trees, you never knew if the oponent does not have a Warlock town full of Dragons somewhere in the background that could burn the small remains of your army undead army to ashes.
All these and few more were examples how the game kept you interested. The problem is, if you noticed, THAT THESE ARE ALL DESIGNE CHOICES, there is not much that the AI could influence on itself since the core mechanic of the game is set in such a way that it ables a . Sure the AI was in some parts much better in H-II, and in some not that much. But the core fact is, that the AI was not the main tool to keep me interested in the game.
I'm not saying your idea is bad, I actually like it, but I think this particular problem could be overcomed by good Game designe rather then arteficial adjustments to the AI game mechanics
____________
I'm just a Mirror of your self.
We see, we look, we gather, we store, we teach.
We are many, and you can be one of us.
|
|
JollyJoker
Honorable
Undefeatable Hero
|
|
posted August 25, 2014 08:42 AM |
|
|
Even there I disagree - the got good spies on you.
I mean, let's face it, if you play a "normal" game, no AI can be a challenge, so will basically play your game, win and be done with it. "A challenge" can mean a lot of things, but in the end, the campaign map challenge - explore a map, unchallenged, then beat the boss in some town - is not the challenge you want.
A challenge is, meeting a neighbor or rival, when you are still developing, and having to ALTER YOUR PLANS accordingly.
Now, sneaking into opponent's lands and steal away their town from under their nose may be funny once in a while and especially on many-player maps, but isn't good regularly.
Conversely - if you moved too far away from home, leaving it too lightly or not at all guarded, and AI opponent makes a beeline to your town, not only did you probably risk too much: you can also RELOAD.
Remember: you played on high difficulty FOR A CHALLENGE, and CHALLENGE obviously means, you have to tread carefully.
Now, sure, having an option to invest in "counter-espionage" in a nice mechanism to "avoid being scrutinized by AI opponent , would be fine, but:
The thing is, that the AI can only do what it is programmed to do; it has no inherent capacity to make conclusions from facts other than those the programmers defined in the manner, If A and B and C, do D. Programming the AI is like writing a manual for someone who can do only one thing: counting off the "triggers" ("pre-defined things happening or not-happening"), compare them with a priority list, then do what it is programmed to in that case - no buts or reasoning, and nothing else, except some artificial randomizers, like If (X) do (Y) or (Z), randomly.
You can all start thinking about how YOU would write that manual (you can do it in regular language) - teaching someone without ANY ability to make their own conclusions how to play the game.
The first thing you will notice, is the fact, that you have to KNOW, what is important and how to act and react, if you want to do it, and that's only where it starts.
Think about the inticacies of logistics in the game - buying heroes, troop sharing, chaining, maximizing the amount of possible actions each turn. Do you really think the AI can cope?
Bottom line is this: the AI doesn't play the same game than a human player - it can't. It must EMULATE progress in a sufficiently natural way that passes for halfway regular gameplay, but if you think about it, AI battling AI ... makes not really much sense. Does it really make sense, on the other hand, when the AI, like the player, struggles for the last 200 Gold needed to get that all-important build in on day 7? NO ONE NOTICES, when it's a Wood short - no one notices, if it gets that one wood out of thin air either.
You just cannot expect the AI to struggle with the game same way as you.
So the bottom line is - the AI as such is a cheat, so of course it cheats, and there's nothing wrong with it either.
The only question is for the QUALITY of the cheating!
|
|
Stevie
Responsible
Undefeatable Hero
|
|
posted August 25, 2014 10:43 AM |
|
|
@Dave:
I understand. I've never played H2 so I wouldn't know how things work there. My proposition came with H3 and H5 in mind. And I have one question for you - What kind of game design would make an AI's 1 Titan beat my 5 Titans?
@JJ:
I didn't expect you to agree with me not one bit. You're the type of guy who'd only agree with himself.
That aside, I really don't understand why you're trying to show me how an AI works, that the devs design some parameters in which an AI acts. I know that full well, yet, how does that help us here? As I said, identify the problem, and find the solution. The PROBLEM as I see it is that after the main armies clash and you win, the game ends, because the AI doesn't have a mechanism to recover!
How do you answer to THIS! I know the AI could be this and could be that. That's all irrelevant after you beat him in it's best state.
So you either keep the game balance and give more towns to the AI that you have for the player (yea, great idea of game design) OR, you make the AI CHEAT so that it poses a threat for you longer than just one decisive battle.
|
|
VOKIALBG
Honorable
Legendary Hero
First in line
|
|
posted August 25, 2014 11:06 AM |
|
|
No.
But it won't harm anyone if its optional.
____________
|
|
Sal
Famous Hero
|
|
posted August 25, 2014 11:15 AM |
|
|
Stevie said:
@Dave:
The PROBLEM as I see it is that after the main armies clash and you win, the game ends, because the AI doesn't have a mechanism to recover!
How is this a problem? This is the goal in almost every video game: once the principal resources one player gathered are exhausted (armies for example), game is over. Commonly we call that "Win a game". It helps people to keep some social life as well.
|
|
JollyJoker
Honorable
Undefeatable Hero
|
|
posted August 25, 2014 11:20 AM |
|
|
Dave is right.
In HoMM 3+ games (except 4) there is no way to recover when losing your Capitol and main hero, and what's more, there shouldn't be one either, except a direct recapture, hero having fled (and isn't dead). Letting the defeated recover like some revenant is flogging a dead horse - you mop things up and finish things, and if that's no fun, something is wrong with the game or the AI or both.
|
|
artu
Promising
Undefeatable Hero
My BS sensor is tingling again
|
|
posted August 25, 2014 11:20 AM |
|
|
Stevie said:
The PROBLEM as I see it is that after the main armies clash and you win, the game ends, because the AI doesn't have a mechanism to recover!
Actually, that's not a problem specific to the AI. If it's not a very large map and if you're not fighting against a tactical genius of some sort, the big battle that determines the fate of the game is always there, even when you play with humans.
The thing is, if people divide their force into many heroes, the player who doesn't and makes one terminator hero wipes them all out with it. So, everybody focuses on one undefeatable hero and expendable side-kicks scouting or bringing +250 gold etc etc.
Some people tried writing mods that limits the amount of units a hero can carry to turn this situation around. So that everybody has many heroes through out the game and it doesn't always come to this one, big decisive battle between two mains but usually it always does.
|
|
Stevie
Responsible
Undefeatable Hero
|
|
posted August 25, 2014 11:58 AM |
|
Edited by Stevie at 11:59, 25 Aug 2014.
|
I don't know how it is for you guys. But for me that's a problem. In Heroes 5 the game was over when I defeated the invading hero in week 3/4. The game was OVER. There was no real point in continuing, I had the upper hand and I knew I would win. That's a problem, when the game just started and you already lost interest in it.
Now sure, you can make the AI take a more balanced route and "play it safe". But again, catch his main hero and it's over. The rest is just a tedious moping up of the map which won't EVER be as interesting as that big battle. Hence, the halfway quitting.
With Heroes 3 I LOVED it when I felt a constant pressure of the AI being able to overpower me at any second. I never taught that after beating his main the game was over. The only moment I felt the game ended was when I defeated the last hero he had and captured the last town he owned, and THAT's what I want BACK!
|
| |
|
|
