Game to easy?

[Darkspire]

As I said, have a closer look at the AI for GC1, no cheating on the AI's part, sometimes its like playing over a LAN the AI on that can really surprise you at anything from normal difficulty onwards. Its the only 4x that has ever really done that for me and ive been gaming for 30+ years, with the CPU processing power these days I can but hope that someone makes use of that power with an AI that can really give us all a run for our money.

This is a small snippet from the AI setup for GC2 by Brad Wardell Gal Civ II: Dark Avatar AI Im more familiar with GC than GC2 as GC2 stil needs a polish IMHO but if you read the post you will see where im coming from.

Darkspire

[RaffleSnaffle]

ORIGINAL: Shark7

AI...Artificial Intelligence or Artificial Insanity. Let's take a look at it objectively. Insanity is to repeat a process over and over expecting a different result. Yep, that describes a game's AI perfectly.

Most developers overcome the Artificial Insanity by letting the AI cheat. Since the AI is only going to do what it is told, and keep trying it even though it has failed the previous 6000 times it tried most developers opt to make the AI more powerful with hard-coded cheats.

And how do other games let their AIs cheat?

1. Ignore maintenance settings
2. Increase build speed
3. Increased unit numbers above the caps
4. Increased hit points and damage points

Just examples, the list in infinately long.

DW takes the approach of trying to make the AI behave sanely. Elliotg hasn't perfected it yet, but the AI has gotten a bit better with every patch and expansion. And at least with DW I don't find myself cussing at the screen because of some cheat the AI gets to use that is totally unfair and unbeatable. I prefer Elliotg's method.

Now that being said, the AI in DW isn't perfect, but I do have a few suggestions that would help balance it out.

1. The biggest balancer would be for the AI to build ship designs that equal what the player designs. So if I design an escort with 4 guns, 4 shields, and a torpedo...the AI will mimic or even copy the design. That takes away the biggest player advantage, the ability to build ships that completely over power the AI. In other words, at each ship type, the AI needs to be able to adapt and build ships that match the fire power and defense of the players designs.

2. Hardpoints (and I like everyone else do not really like this idea). Hardpoints would prevent the player from designing over-powering ships. Of course it also takes away the reason for having a ship designer. It is a possibility, but I call it the absolute last resort...I DO NOT ENDORSE THIS PERSONALLY.

3. Adding additional penalties to the live player. Again, not my favorite idea as it really is just another cheat, albeit one that lessens the player instead of bolstering the AI. Again, I don't endorse this.

All in all, I think that suggestion 1 is the best. Once the AI gets the ability to completely match player ship designs, it will balance the game without making the AI a cheater or the hamstringing the player.

I agree completely with your ideas for the AI re balancing and your view on how the AI in this game can be a breath of fresh air compared to straight up AI hacks in other games.

It's cool Elliot hasn't taken the easy way out and just let the AI hack to win, but like you said, it does need work.

[feelotraveller]

Fully agree that the 'AI' could do with some work.
 
But let's be clear about a few things.
 
The current ai DOES cheat.  As one example once you reach the point of them being scared of your huge size and power their development increases to maximum without regard for luxuries.  I've seen multiple contacted cpu players go through this process simultaneously when I first hit this threshold going from 70-80 percent to 105 as quick as possible.  If this was one empire it could be that their luxuries just kicked in.  But when it is 10 simultaneously?  Occam's razor says they are cheating.  [:)]  You can also watch this process happen as you contact new empires.  They will go from their uncheating state to max development regardless of luxuries if you are of huge size and power.
 
Also although it is a bit of a mantra for many games 'better AI' is not a simple process.  What passes for AI is all the algorithms governing cpu player behaviour and their interaction with each other and the player.  So to improve the 'AI' what is needed is a careful (and thorough) critique of where it currently goes wrong and how to make it better.  Things such as the freight transport algorithm as well as the battle target selection not to mention diplomacy and tax (and all the rest...).  If we really want a better AI rather than intoning the mantra it is far better to pull apart one or other cpu player subsystem or their interactions.  When we say 'better AI' what is it we really want, and how oh how do we think Elliot might be able to do it for us?
 
Finally it is worth keeping in mind that the ultimate point of the AI is not to be fair or to give us an ultra-competitive game it is to make it a fun game.  At least fun is what I play for.  [;)]
 

[jpwrunyan]

I think blaming the ai to the exclusion of all else is wrong. The game conditions are exploitable. The game could be improved drstically with some environmental changes (totally agree about characters blowing up the game--that is a good example). There is the entire thread of "three simple things to make dw better" that goes over this.

Ai can always be improved. Thats a given. I would rather see the exploits removed first. Again, see the other thread I referenced. Maybe someome can link it (cant iphone).

[Shark7]

ORIGINAL: feelotraveller

Fully agree that the 'AI' could do with some work.

But let's be clear about a few things.

The current ai DOES cheat.  As one example once you reach the point of them being scared of your huge size and power their development increases to maximum without regard for luxuries.  I've seen multiple contacted cpu players go through this process simultaneously when I first hit this threshold going from 70-80 percent to 105 as quick as possible.  If this was one empire it could be that their luxuries just kicked in.  But when it is 10 simultaneously?  Occam's razor says they are cheating.  [:)]  You can also watch this process happen as you contact new empires.  They will go from their uncheating state to max development regardless of luxuries if you are of huge size and power.

Also although it is a bit of a mantra for many games 'better AI' is not a simple process.  What passes for AI is all the algorithms governing cpu player behaviour and their interaction with each other and the player.  So to improve the 'AI' what is needed is a careful (and thorough) critique of where it currently goes wrong and how to make it better.  Things such as the freight transport algorithm as well as the battle target selection not to mention diplomacy and tax (and all the rest...).  If we really want a better AI rather than intoning the mantra it is far better to pull apart one or other cpu player subsystem or their interactions.  When we say 'better AI' what is it we really want, and how oh how do we think Elliot might be able to do it for us?

Finally it is worth keeping in mind that the ultimate point of the AI is not to be fair or to give us an ultra-competitive game it is to make it a fun game.  At least fun is what I play for.  [;)]

But honestly, while it may be the AI cheating, its not that bad. It's something I didn't even notice.

The AI cheats that annoy me are the ones that are so blatant no one can miss them. Like certain games where the AI can expand at 10 times the speed of the player, or researches in 1/10th the time. And yes those games are very hard (not challenging) simply because the player is always at a huge disadvantage and never has the chance to catch up.

Imperium Galactica 2 was a great game, but I NEVER managed to beat it on normal difficulty. That games AI was just a bit too hard, it took the challenge out of it and made it an exercise in frustration. I sincerely hope that Elliotg never makes the AI in DW that hard cause it takes the fun right out of the game.

[feelotraveller]

I'm real upset that the cpu players get a bit of a boost when I am massively ahead.   [/sarcasm]
 
However I do get a little tired of people saying that the AI does not cheat in Distant Worlds... I've got this funny obssession with the truth... when it actually does.  (Another little fudging happens when the cpu players sometimes get the boost of an ultra-rare resource without any possible access to such a source.)  Such fudgings are as much a part of the set of systems which comprise the AI as anything else.
 
 

[Algoritm]

I think blaming the AI to the exclusion of all else is wrong. The game conditions are exploitable.

AI can always be improved. That's a given. I would rather see the exploits removed first.

I totally agree with jp on this. The main problem is that where there is an exploit, an intelligent human player will always find a way to turn it to his/her advantage.

One way to fix this, and thus make the game harder, is either to remove the exploit, or make the AI adapt to it.

I will detail a few concrete ways of doing this quite simply (by identifying the exploit and finding a way to even things for the AI):

1) When I first started playing this game and had everything on automated, the very first thing I felt compelled to do was take control of exploration. I quickly realised how exploring would give me an edge over the AI by finding strategic resources, key colonisable worlds, derelict battleships, colony ships, lost independent colonies, super-weapons, world destroyers etc...

the AI always under-estimates the importance of exploration. Look and see for yourself in the race policy folder. Nearly all races (except Zenox) are set at "Exploration Priority ;1,0"

Change that value to "2,0" (after saving all original policies in a separate folder) for all races and you will give the AI a very neat, cheat free, edge.

2) The second thing I focussed on after exploration was research (by the way, I never use the tech trading exploit which I personally find silly as it is). I don't think I need to explain just how important research is in any strategy game.

Again, if you look in race policies, only the Quameno and Kiadian are set "2,0" for "Research Priority". All other races are set at 1,0 or 1,5. Don't be afraid once again to give every race a solid "2,0". The game will not lose it's "unique race" flavour cause the Quameno have both Technocracy and gifted scientists, thus they will naturally remain ahead of.. say for ex: the Atuuk... in science.

3) Thirdly, you guessed it... ship design. And that does not mean simply adding more guns to the design. An intelligent player will always make a design that fits the need.

Unfortunately, it's very complicated to make the AI better at ship design without cheating. For example, if you armed all your ships with ubber latest design titan beams but no long range weapons, you'd be in trouble if the AI was "given" that info and decided to equip his ships with fast thrusters and long distance missile/torpedoes with orders to stand-off.

You'd also be in trouble if the AI realised that your ships had many shields but only a few layers of armour and decided to equip his ships mainly with railguns.

A human player can adapt. Unfortunately, an AI doesn't... unless *coughs* the programmer *cough* found a way to implement that somehow [;)]

I think I'll stop at those three points for now but trust me, there are many other ways of making this game interestingly harder (or more fun).

[Shark7]

Another thing I've noticed that could easily be changed.

It is far too easy to steal tech off the Ancient Guardians. You'd think they'd be better at counter-espionage.. And remember they are far ahead of everyone else tech wise, makes it really easy to steal mid and late game tech early.

[adecoy95]

ORIGINAL: Shark7

Another thing I've noticed that could easily be changed.

It is far too easy to steal tech off the Ancient Guardians. You'd think they'd be better at counter-espionage.. And remember they are far ahead of everyone else tech wise, makes it really easy to steal mid and late game tech early.

haha, remember when tech steal missions used to give you the complete tech, instead of part of it?


also, i have kinda mixed feelings about tech trading "fixes" i think that the race specific technologies feels about in the right place, but the others are too easy to pass around

[Shark7]

Tech trading is too easy, best policy is to refuse it, unless of course you are helping the AI by giving them your tech, which I do from time to time.

Another thing I've noticed. If you set the game to start with multiple characters (governors in particular) the AI has a tendency to leave them exactly where they started. So while I thought giving one of the AI empires 3 Governors and 3 colonies to start with was a good idea, well...the AI left all 3 Governors on the home planet. Governors are the only thing I noticed having trouble though, the AI does pretty good with other character types, sending admirals to fleets, ambassadors to friendly empires, etc. So if you want to give an added race an advantage through governors, make sure you do it with a single starting one and an increase appearance rate for the others.

[Satosky]

I find that customizing the ai empires(such as giving them an excellent home system or small tech advantage) before I start the game helps increase difficulty a good amount. Better than letting it randomly create the Ai empires I think.

[Darkspire]

An example of GC2's AI setup. Can but hope that DW even gets near this standard.

[jpwrunyan]

@akula
Stealing research from the guardians is an exploit, I agree.

But stealing research itself is often an exploit. There is something wrong if I lead the ketarov in successful spy missions with one spy when they have 5. Either counter intelligence missions need a major boost, or the AI needs to spam spy missions (at least when they have >2 spies). Or both.

[Shark7]

ORIGINAL: jpwrunyan

@akula
Stealing research from the guardians is an exploit, I agree.

But stealing research itself is often an exploit. There is something wrong if I lead the ketarov in successful spy missions with one spy when they have 5. Either counter intelligence missions need a major boost, or the AI needs to spam spy missions (at least when they have >2 spies). Or both.

The stealing of the tech itself is not an exploit per se, I just think the steal research mission is far, far too easy in the current game model. That and the fact that the counter intelligence mission is far too weak at stopping spy missions.

[jpwrunyan]

Just so.
If the ai only sabotaged/stole research as much as I do then it wouldnt feel like an exploit. But they dont. I can only assume then that all their spies are on counter intelligence. Their counter intelligence sux.

I demand satisfaction! Sign my petition!

[Shark7]

Well Erik and Elliot do read the forums, even if they don't respond to every thread. I'm sure they've seen this and wouldn't be surprised if spying gets a tweak in the next patch (the expansion is too close to change things I think).

Basically I think between us we've hit on what needs to be done:

1. AI empires need to utilize their spies more for sabotage and steal missions.

2. The AGs need to have a modifier to make it harder to steal from them. Stealing from the AGs out to be about as difficult as stealing from the Orions in the Master of Orion series...it could be done, but it took a maxed out spy and a lot of luck.

Right now I can load up my game with a starting spy and steal a mid game tech from the AGs in 3 months with a 94% chance of success. Its needs a modifier so that my super spy has a 50% chance of success with a year long mission, from my point of view. Basically I limit myself to stealing 3 techs from them then stop: enhanced storage (for the fuel capacity), continental colonization (I don't like researching my natural colony type), and LR Sensors (so I can build monitoring stations). Then I quit because it just throws the balance completely out the window to do more.

[Satosky]

The AG are a reclusive race for reasons that most of us know by now so I do find it a bit strange that they wouldn't notice some random human or teekan spy wandering around and stealing their highly sought after research. The fact they are a planet of exclusively machines only furthers the point.

[Haree78]

I guess infiltrating the AGs is actually just hacking
Would be nice if Empires could get a bonus at counter espionage like other bonuses and AGs have this set very high.

[Shark7]

ORIGINAL: Haree78

I guess infiltrating the AGs is actually just hacking
Would be nice if Empires could get a bonus at counter espionage like other bonuses and AGs have this set very high.

That would fix it.

[Beag]

I wouldn´t compare GC 2 to DW, the first is turn-based (which makes AI programming much easier because you remove a critical factor that gives an advantage to the human - timing), combat is a joke and ship design is much simpler.

As a sugestion for DW 2 I´d say that the strategic part should be turn-based; and combat real time - as long as it doesn´t use Endless Space as its model. It would allow the strategic part to be deeper while still allowing control, and allow the tactical part to be more complex (which nowadays it can´t be, because managing a battle and managing stuff at the same time is impossible).