Discussion on posting about AI

You may notice that in my looong thread about AI behavior, I did not use the word 'cheat' once. 

Given that the experiment was deliberately ran on "Challenging", there should have not been any difficulty bonuses.  Some of the things I found out (setting on swampland, getting free essence, rushing at no cost, free traits) were definitely not things that should be there on 'Challenging'.  They are advantages that the AI gets because of bugs.

These bugs have to be reported and eradicated.  Players using the word 'cheat' is counter-productive, but not the end of the world.  As you yourself said, it's a serviceable shortcut for "violating the rules in a manner that should not present at a particular difficulty".

Your thread is not one of the ones I was referencing.  I liked it and think it was quite helpful and to-the-point.  I don't think it is the end of the world, just wish people would strive to accurately note what it is their thread or post is referring to.  Repeatedly using "cheat" could make someone...defensive in impede communication. 

To Novalith - your post, sir, is just a piece of elaborate rationalization and apologetic tirade designed to cover a simple fact - the AI does cheat and it is rightfully called cheating.

Can I prove my point? Of course I can. Every game comes with a set of rules (even asymmetric games that pit different forces against each other, like AI wars). These rules should be transparent, unambiguous and clearly understood by all participants prior starting the game. Any violation of such rules is cheating - no exceptions, no fancy talk about that.

What happens if I come to a chess tournament and in a attempt to smuggle an extra rook in an otherwise drawn endgame to achieve victory? I will be thrown out, probably banned, shamed and discredited as a player. I have just spoiled the competitive thrill of the intellectual battle by violating the rules, thus making such competition meaningless. Did I cheat? Oh yes, without a doubt. When a computer program does the same thing, is it cheating? Without a doubt.

Same goes for Civilization, Fallen enchantress, and any other game of similar nature. The manual describes the rules quite clearly - when hiring heroes, the hiring price must be paid, save for one faction that is allowed to hire for free. If the AI hires for free with other factions, it's cheating. When it insta-builds several objects at once, gets buildings and units for free - cheating. Also, the challenging level should be AI algorithms at its best, without cheating. If it's stated so, it should be so - expecially by a game from people like Brad who is so vocal about the "gamer's bill of rights".

It is true that it's probably impossible to build a competent AI for this type of games using the current technology. The AI needs bonuses and other helps to provide enjoyable challenge - but it should be done by a transparent way, by saying "this and this level means this and this bonus for the AI", not by secretly sneaking AI cheats in.

The other option is to build the game around what the AI can do - AI wars is a good example. The AI is given infinite resources, but is allowed to use them in a clearly defined manner (AI progress). That is the correct, clean and transparent way how to do it, I recommend the game to your attention.

Anyway, I believe that the instabuilds and other cheats seen so far are actually bugs, and that the devs will fix them given time and testing. 

That's why IMO people like you are doing the game a disservice by trying to rationalize the bugs and convincing people to accept them, because "the AI cannot manage otherwise". You are clouding the subject instead of clarifying it.

Wow, I think we have a vast chasm of misunderstanding to cross.  Let me see if I have gathered the points of your post correctly and try to respond appropriately.   I will work from the bottom up.

1. You believe I am rationalizing bugs

1a.  No I am not.  The bugs this game has need to be fixed.  It would appear that Frogboy is making an effort to do so.  Anything we point out that the AI is doing that he did not intend it to do is acknowledged as a bug, not a cheat, and fixed.

2. The AI is limited by current technology and may need help.  That help should be "transparent".

2a.  Correct.  I don't disagree that it needs bonuses.  How transparent they are is up to the developer.  If I know it builds faster but not that it builds 2.65 times as fast I am ok with that.  If I am never told that it has an increased crit chance and it does, that isn't cool.  However if that is what it was programmed to do I might be able to live with it.  If the computer is programmed to occasionally rush a project on expert difficulty but when it does it rushes ALL projects for NO cost and that was NOT the intention of the programming then that is a bug, see 1a.  Right now the higher difficulty levels state that sovereigns are buffed and that they receive economic bonuses.  It does not state how much HP they get or if they hit harder.  It does not state that they get x amount of gold extra per turn.  I am ok with the vagueness, you apparently are not.  That is just a difference of opinion. Frogboy stated that he could start a discussion on what exactly the AI does but doesn't want to spoil things for people.  

3. You give the example of bringing extra pieces to a chess match as your human example and playing by different rules than the manual states as different examples of cheating.  

3a.  I would call your chess example cheating because that is a no-brainer.  You knew the rules and chose to disobey them in explicitly state so.  You are a cheater.  As for the computer...well let me try explaining it a different way.  I tell you I don't like your chili because it is nasty.  Does that help? Nope.  You know I don't like it but you cannot tell why.  If I say it is too salty you can add less salt, too sweet then less brown sugar, too spicy less jalapenos.  If I tell you that it is too plastic, and there is no plastic in it well then I am just wrong.

If you say "I don't like this game because the AI cheats" but what you mean by cheating isn't clearly defined then how are they supposed to fix the problem?  If you say the AI is bugged, it is doing something it shouldn't (like amassing 20000 gold by turn 25) then they look for the bug.  If you say the bonuses the AI get are not transparent, or shouldn't be there at all then they can modify them, or make the tooltips more verbose, or make a post stating everything the AI does on each difficulty and provide you with clarity.  If you say that you hate how the AI plays by different rules then the AI can be reprogrammed or the rules can be changed.  Perhaps some of the things that may have been changed to save on processor power could instead be made a toggle.  

You can label any issue you have with the AI as cheating if that is what you chose to do.  What I was stating in my post is that to simply say it cheats clouds the issue.  How you saw my wanting people to explain there issues in a clearer fashion as "muddying the water." I honestly do not know.  I suppose that at least in your case I failed to explain my intentions clearly.  I shall see if maybe I can edit it to be clearer

You also seem to imply that I think the game is perfect and that I believe people should not post their problems or opinions if they have them.  Again that is not the case.  I didn't say that just simply asked for clarity.  If you did not intend your post to project that implication than perhaps it is me reading too much into your tone.  

v/r

Novalith

I tend to agree with Kanamura_CZ specifically in regard to transparency.

I don't in principle mind the AI playing to different rules. However the rules it is playing to need to be clear and transparent to me the player.

Without transparency how do I know that something I see the AI doing that differs from the rules that I am aware of is not a bug, in fact without the information to be aware that this is WAD I have to assume it is a bug.

With transparency I can see that an AI rule exception is WAD, if however the consensus opinion is that this exception is making the game unfun and skewing the challenge unfairly then the designers can decide if it should be changed.

Without transparency I can see there being increasing frustration with the amount of perceived bugs and the willingness to search and report to forums to find out if the game is WAD or is "broken" shouldn't be taken for granted from an individual player or the playing community (that is what open betas are for).

I agree the AI does NOT CHEAT, what happens is that it gets to SOME UNFAIR ADVANTAGES and HANDICAPS. Now I'm not against advantages and handicaps and even to the point of UNFAIR as long as the DIFFICULTY setting is high or highest I don't even see UNFAIR advantages and handicaps. I know the AI can only be programed to do so much in a game. The more stuff there is to do in a game the lesser the AI is going to be. Take a game like Warlords IV vs this game or the Total War game and many other MONSTER games and I already know the AI in those games aren't going to be as good as the Warlords AI. What I do want though are ADVANTAGES and HANDICAPS to be given out to them based on difficulty and the higher the difficulty the more it should get. I NEVER EXPECT the AI to PLAY a BETTER game between EASY and IMPOSSIBLE though as there's only a few tweaks that could/would make this happen anyways and they would never be significant enough to make the AI a real challenge without advantages and handicaps.

SO yes players need to stop saying the AI is cheating because an AI can't cheat it doesn't know how to cheat hell for the most part it doesn't know how to play. lol

I think there are some bugs that, intentionally or not, result in the AI cheating.

They bug me because I've put a lot of time into the AI to ensure it doesn't cheat.

The cheats I'm aware of and am investigating are:

1. Report that the AI was able to build on a swamp (that shouldn't be possible)
2. Report that the AI wasn't charged for recruting champions (I've fixed this -- I assumed that when I called recruit, it charged me like it would a player, it did not).
3. Report that the AI wasn't charged properly on higher levels for rushing production (I've fixed this).

Now, realistically, none of these things is going to make the difference between winning or losing. Especially in a game where the humans are engaging in massive massive self-bonusing (custom factions, custom sovereigns, save/reload).  I don't begrudge people doing those things. It's their game.  

While challenging is the highest level that the AI plays at without getting bonuses, I tend to think of Hard as the first level where the AI and an experienced human are playing on equal terms even though the AI gets some bonuses.

In most games, the AI cheats at every level. I don't think most people realize that. I can't think of a single major PC strategy game where the AI, even on "normal" isn't playing a significantly different game than the player.  That's really not a big deal btw. The goal of an AI is to provide a fun game.  It's just that Stardock games tend to have dedicated AI development so we expect more out of them.

One of the biggest problems about transparency with regards to AI rules is the players ability to manipulate the loopholes in the rules.

I do not doubt that transparency does help identify bugs and determines whether the AI is playing by its set rules. But if you knew exactly what the AI will do, then you lose some of magic of what the AI is going to do if you do X. Once this is all revealed you lose your title as player and move into the world as developer.

The only way I can describe this magic is through movies and film. You lose some of the magic when you know how it is done, where it was filmed. What aspects were needed to make the effect show up on the screen. You lose the immersion of the movie. Same applies to gameplay.

I think this part of the reason that the game is so random is to make it entertaining not only for the players, but also the developers. I was happy with my ignorant bliss that the AI was doing quests and simulating a battle just like we were doing, but then I found that the AI rules don't do that and instead have a seperate roll to determine success or failure of the quest. I was fine not knowing this, but I guess if I discovered this I would cry foul not knowing that it was working as designed. Now I would like a different application where the AI does a battle auto-resolve on the quests with battles or an avereage battle if the quest contains multiple battles.

There needs to be some explanation on what the AI is doing, but there does not need to be full transparency on the AI rules. That is akin to asking to see the code, which I don't think any player needs to see.

This is where we run into the problem of how much transparency is enough.  I have burned out on games beta testing them because I had to learn everything the AI could do and how in order to test if it was being done correctly.  The other thing is you cannot un-see things.  So once you pass your point of transparency comfort then it is already too late. 

This is one of those things where there is no right answer despite what some people might believe.  What is good for you may not be good for me.  It is personal preference and up to the developers to decide where the line is drawn.  

I am the one of the people who use the term 'cheat', and I think you already found the answer yourself unintentionally.

We all aware that, there is no way, AI can compete against humans properly (hey, even abstract games like Chess, you need a super computer to beat a human, as far as I concerned.), so it is inevitable that AI needs helping hands to fight well on harder difficulty setting.

What makes me, probably many others is that AIs violating 'rules' to beat the game are really upsetting and sometimes just breaks game theme itself.

I don't care about AI getting massive gold boost, it needs since it wastes so much units for stupid things. It also needs more HPs because they are poor at designing/using units and just gotten beaten by players so easily. But none of these bonuses do not alter/violate the 'rules' they game provide. They have to do quests to get items, need to build a city and improvements, and fight monsters then destroy lairs so the land can be safe enough to be settled.

Whether they have free golds falling from skies, and normal units as tough as my mid-leveled heroes, I am ok as long as they are all have to follow the game mechanics (a.k.a 'rules') and play the very same game as I do.

The problem is they don't. Essentially they don't actually 'build' improvements (since they are done instantly) and they don't clear monster lairs while monsters protect them from player army.

They do not follow those rules.... It is like trying to play Chess when your opponents are playing Janggi/Xiangqi. Clearing monster areas to regain the area control is one of the most important aspect of this game that makes it different from other 4X games. Without it, this game is slightly more boring, leaner version of Civ game with magic.

I'm interested in the omissions from this list.

Is the AI insta-building multiple city improvements not a bug?

Is the founding of 30 populations cities by the AI without expending a pioneer not a bug?

If the answer to these questions is no then I find it hard to reconcile with statements that the AI is playing by essentially the same rules as the player.  If the answer is yes I am wondering why you don't note them as being investigated (nor is there mention in the upcoming patch notes that they have been 'fixed').

These 2 issues can be game-breakers and have been documented as occuring on very low difficulties (novice/beginner respectively).  The assmuption is that they are therefore happening at all harder difficulties as well.  Perhaps you could put our minds at rest?

What? not only free, re-usable pioneers, but they also got free 30 population on top of that? With so many glaring things I could not even look through it. *facepalm*

I noticed a memory leak that is causing AI units to gain traits they should not be able to get. Heroes are getting Unit_Design traits. The debug was sent when it crashed a few turns later. Hope it helps you nail down the leak. This is technically AI cheating, but something that is relatively easy to fix.

Like many people already said and said again outside of this Stardock forum repeatedly, the game engine is really, really, really bad IMO, and I am being very gentleman about this.

Please hire more game engine guys instead of just 1... You guys are going to use just slightly modified engine for GalCiv3, and at that point, the game engine shouldn't be issue no more.

It only happens rarely but when it does it can be a game-breaker.  As far as I can tell it is a condition that the pioneer be part of an army.  And yes I reported it and Frogboy has the saves.  Which is why I am concerned that he does not mention it.  I remember at least one other poster mentioning seeing it too, so it's not just me...

(One game on beginner I had a Magnar opponent with 6 cities before turn 50, 4 of them founded with 30 population and by the same pioneer - I checked after abandoning the game and watched the 'nefariousness' happen.  NOT the easy game I was playing beginner to get.   )

If those things are happening they would be bugs. But we've not seen it yet. In all the bets, it never came up. once we reproduce it then we can address it.

I'm a bit skeptical on some of the claims. Only pioneers can found new cities, for instance.  Cities start out with whatever their default population. The ai can't affect that. 

Other things aren't AI bugs but i engine issues. The building In the swamp is an example.  Most of these kinds of things are pretty rare.

You might have some sort of bug with CorePlayerAbiliites. I notice that many memory leaks give bonuses to the abilities in this file. Anything from starting city population to number of unit design traits can increase for human and AI like at times. I know just saying it doesn't help that much, but maybe thinking about it in that context will hep later on when you reproduce it?

I'm surprised Frogboy.

I've posted saves both the turn immediately before the occurence and the next turn where the save shows what happened for me (in case your hitting the end turn button on the first save produces something different).  Load them up and look for yourself.  I assumed that you already had.

You responded in the first thread and had just previously asked for saves showing the problem in the second.

AI founds 30 population city https://forums.elementalgame.com/435274

AI insta-builds 3 city improvements in 3 cities (of the 5)  https://forums.elementalgame.com/435826/get;3268259

I hope that I am not wasting everyone's time by making an effort to improve the game by spotting this AI behaviour and providing these saves. 

Frogboy, I have also seen the 30 pop cities founded without consuming the pioneer.  I was playing ridiculous at the time, so I assumed it was just a difficulty AI bonus.  It was a sovereign + hero + pioneer stack going around and making cities.  It made at least two, and it still had the pioneer at the time I  killed everyone.

Knowing that the AI gets to build a bunch of free buildings every time I need to save/load ruins the game for me.

Yes, it's a bug, but I have no problem saying that the AI is "cheating" in this case.

Thanks for looking into these issues.  Personally, I feel the difficulty levels have become harder lately so I wonder if some of the bugs are the inadvertant results of more recent changes (new code not playing nice with old code).  This is pure, unadulterated, idle speculation and really me just thinking "out loud".

In any case, good luck with the bug hunt.