4X A little Survey about 4X Games

[tiagocc0]

Yeah, everything is set in stones too so once you learn how to beat the game is just a matter of timing your decisions.

 

[hoverdog]

Now that's a rare find, I loved to play Reunion.

I got it in the same year i got Xcom (i think), through a magazine that sold floppy disk games, best time i've ever had playing games.

Lol at those crafty apes race description on your site, first time i'm hearing about the game, have to say i'm loving the pixel art aesthetics.

It's one of those little phrases that sometimes come up in your head

 

[J_C]

Did the survey.

 

[tiagocc0]

Did the survey.

Thanks!

Here are some results already, 35 took the survey. All questions are optional some some may have skipped one or two.

1. Do you like Space 4X Games?
33 / Yes
2 / I don't mind
0 / No

2. What X do you prefer in a 4X Game?
11 / eXplore
12 / eXpand
6 / eXploit
6 / eXterminate

3. What is the most important aspect in a Space 4X Game?
11 / Research
7 / War
6 / Ship Design
8 / Diplomacy
5 / Other

4. What is the worst aspect in current Space 4X Games?
4 / Research
5 / War
2 / Ship Design
18 / Diplomacy
8 / Other

 

[hoverdog]

4. What is the worst aspect in current Space 4X Games?
18 / Diplomacy

 

[J_C]

4. What is the worst aspect in current Space 4X Games?
18 / Diplomacy

And it is not just in space 4X games, but in every strategy game. The games' AI is not capable of handling the diplomatic situations if they are a bit complex.

 

[tiagocc0]

The problem is that they want the AI to act like a human.
So they plan a very complex system, full of layers. Then budget tells them otherwise and they end up with a half finished AI system full of bugs and complex as hell because they didn't had time to cut the edges.
So anything you try to do with it is like shooting in the dark.

An AI must act like a bot, not a human. After you accomplish that then you may try something better.
Also they think that AI must have a ultimate goal of exterminating everyone, so everything revolves around that.
The ultimate goal should be to survive and not exterminate.

 

[agentorange]

Took it.

4. What is the worst aspect in current Space 4X Games?
18 / Diplomacy

And it is not just in space 4X games, but in every strategy game. The games' AI is not capable of handling the diplomatic situations if they are a bit complex.

Pretty much exactly what I asked for in the "what would like to see in 4x" question. Especially with regards to how it handles diplomacy of war, declaring war, joining war, peace treaties, in most 4x it seems completely arbitrary.

 

[Kem0sabe]

In terms of ingame diplomatic models in strategy games i'm a big fan of the system used in CK2, i don't know how they managed it but every game feels fresh and reactive to the players actions.

 

[tiagocc0]

Took it.

Thanks!

I liked CK2 too, but I haven't played much.


It seems like developers also develop in a way that AIs can communicate with each other (cheat) as to make the game more challenging for the player.
Or they have a 'bigger' AI that coordinates all the others where that AI is 'aware' of a lot of things that the player isn't (cheat again).

While this could increase the difficult, it does so in a pretty retarded way.

AIs should be isolated just like the player is, if the AI is easy to beat because of that then you must optimize the AI instead of cheating to achieve the same result.

 

[agentorange]

In terms of ingame diplomatic models in strategy games i'm a big fan of the system used in CK2, i don't know how they managed it but every game feels fresh and reactive to the players actions.

A lot of that is, I think, down to the sheer amount of factions in the game. In most 4x you're dealing with 4 - 10, but in CK2 you've got pretty much hundreds, since even the lowliest of Dukes of Barons can do something that has major repercussions. And with the marriage and inheritance systems constantly causing further diplomatic complications.

 

[Kem0sabe]

Thanks!

I liked CK2 too, but I haven't played much.


It seems like developers also develop in a way that AIs can communicate with each other (cheat) as to make the game more challenging for the player.
Or they have a 'bigger' AI that coordinates all the others where that AI is 'aware' of a lot of things that the player isn't (cheat again).

While this could increase the difficult, it does so in a pretty retarded way.

AIs should be isolated just like the player is, if the AI is easy to beat because of that then you must optimize the AI instead of cheating to achieve the same result.

Yep, cheating AI's, especially in "overland" strategic management is lame, more resources shouldn't be a crutch developers fall on to tune the AI difficulty.

I terms of combat i have more mixed feelings about it, any kind of AI will always be easy in battle if it doesn't have overwhelming force, Total War handled this by giving the player moral penalties and bonuses to the AI troops, not the method i prefer, but on the other hand i have little to suggest to improve on it.

In 4x games it's mostly down to technology, the side with the better lasers and shields usually wins even if the numbers are lopsided, but all things being equal the player has no contest.

A lot of that is, I think, down to the sheer amount of factions in the game. In most 4x you're dealing with 4 - 10, but in CK2 you've got pretty much hundreds, since even the lowliest of Dukes of Barons can do something that has major repercussions. And with the marriage and inheritance systems constantly causing further diplomatic complications.

I love the sheer insanity of CK2 familial relationships to be awesome. You could end up with Genghis Kan being the new pope if you screwed around Europe with enough enthusiasm.

 

[tiagocc0]

It's difficult to see an AI that actually preys the player, attacking the most vulnerable part and increasing the intensity of the attack with time.
Even in Eador the AI usually strays from ripping you apart and for no reason, I haven't played the last difficult setting, but maybe there the AI does its job well.

 

[DwarvenFood]

filled your survey, bro.

 

[tiagocc0]

filled your survey, bro.

Thanks DwarvenFood!
With yours we reached 55 responses!

 

[Zewp]

An AI must act like a bot, not a human. After you accomplish that then you may try something better.
Also they think that AI must have a ultimate goal of exterminating everyone, so everything revolves around that.
The ultimate goal should be to survive and not exterminate.

This. So much this. AI often isn't 'aware' of its own survivability. I've played countless 4X games where a small, underpowered AI faction would declare war on other factions without 'thinking' about it. An AI should do things that is in the interest of survival, not trying to wipe out other factions. Extermination can be a goal, but it should not be the goal.

 

[spectre]

Done the survey. I think the best way of doing a space 4x game is to identify what made MOO2 tick and expand on it.
An important thing to consider is the AI. It's fine for it to use cheats to give it an edge (because, hey, let's be realistic here), but the important thing to keep in mind is that you never want the AI to give an impression that it plays by a different set of rules.
This is the golden rule that applies to each and every strategy game worth its salt that was ever made.

That said, all in all, what I'm expecting of a space 4X game is basically MOO2 with updated graphics that attempts to expand on the MOO2's bare bones.
Presentation wise, I thought Endless Space did a pretty good job with how the planets and systems were handled with regards to descriptions and visuals.
Unfortunately, the gameplay failed to deliver in many respects.

 

[MetalCraze]

Overall MOO3 > MOO2
But the interface is horrible clunky shit. It's what brings it down.

MOO2 is 'bare bones' indeed. It's very short, there are not many victory conditions (2 and if you refuse the Emperor one you have only Conquest forced on you). And the late game is always very micromanage-ish with you building same set of buildings on every planet. There's no point in planet specialization because you can build everything on any planet given time. GC2 and MOO3 did it right by limiting a given planet. Tactical combat is a gimmick.

Good 4X design is to leave you running the empire, not handhold every planet in it - although you should be able to do that if you wish (instead of mandatory and of course not horribly dumbed down like in Endless Space). That's where MOO3 (and SMAC - talking about cities and resources - to a degree) succeeded.

MOO3 just needs to be remade with a good interface.

Of course the best 4X is certainly SMAC.

 

[tiagocc0]

The AI can already make thousands or more analysis in a few seconds. I don't see why it should cheat.
It's easy to make an AI that will kick your ass, but since they want the AI to act like people they avoid the best solution and go after complicated systems that act dumb on purpose.
Ok, easy difficult is done, but now what about hard difficult? Cheat, since this is the only way for it to still look like it is human. Oh, the irony.

But I want to take another approach.
Instead of trying to simulate humans I want to make a very hard AI that uses all features of the game to it's maximum. Given it has enough time to calculate everything.
Then I cut the time it has to calculate for easy difficult, it is still intelligent, it just can't check too much possibilities (can't think too far into the future).

Of course for the hard difficult I will have to cut it too, since it would take too long to fully analyse the game, but it will have more time to think than on easy.

That's why you see turns taking longer on chess when you raise the difficulty while turns take the same amount of time on games like this even on max difficulty.

EDIT: Resuming, they feel that the time between each turn is so important that it dictates how the AI will work, this is really an issue on RT games, but it's retarded on TB games since it's turn based and we have much better computers now then when MoO was released.
So I don't mind sacrificing time to get a better result.

 

[hoverdog]

I agree, but the time for the AI to calculate his turns cannot be too long, or it kills the flow of the game. Heroes of Might and Magic V is main culprit here; granted the whole engine was awfully slow, but when you play a large map with 7 AIs and each takes half a minute to play his turn, you spend more time on staring at the screen than on playing the damn game. That's why I've only played multiplayer HoMM5 a couple times.

 

[tiagocc0]

But they had a dumb AI plus took a long time.
Also it must have taken the same amount of time for even the easy difficult.

If you will feel the time when you play on very hard, and if you are playing there it probably means that you want a challenge and you are willing to sacrifice something for it.

I surely wouldn't mind playing MoO2 on a difficulty where it does not cheat but would take longer.

 

[spectre]

If you can pull this off... more power to you. I always wondered if it is possible for the AI to fully grasp the intricacies and work the system in the 4X game. I've yet to see it done, keeping my fingers crossed for you, but allow me to remain skeptical until I see it.

When it comes to AI design, you may find it useful to examine how it was done in Chessmaster 10. I think that you may be well familiar with how chess ai is programmed, but Chessmaster is a good example of how to make work.
It can be set to full kick-ass mode, but what was always the most interesting to me were the various quirks that could be programmed in, giving it an edge/penalty in various stages of the game, over or undervaluing different types of play.
You may find it of interest especially within the framework of an 4X game, where you will want to have different races which all play differently.
I don't think it's fun to see all the AIs play with the same machine-like efficiency, it's the little quirks that make it fun and while it's fun to see Ghandi going trigger happy with nukes in Civ, the impact is only there because you generally see him as a harmless pacifist most of the time. It can be argued how much of this can be preserved when we're talking about Hard and Impossible difficulty levels.

 

[tiagocc0]

This is why acting like human ruins it all again.
If it acts like a bot, when I set that a race is pacifist or aggressive then it really means what I set, instead of it falling on random numbers and so their traits would only affect it occasionally.
If it is a bot it will act just like I told it to, it means that a race will act like it should, it's predicable and it must be. Or the game won't make sense and every race feels the same.

The map must be randomized, not the AI behavior.

Like I said the ultimate goal must be to survive instead of exterminate, unless is a racial trait.
So playing its best doesn't mean they will kill everything, but instead that it will be very hard to kill.

 

[Horus]

The AI can already make thousands or more analysis in a few seconds. I don't see why it should cheat.
It's easy to make an AI that will kick your ass, but since they want the AI to act like people they avoid the best solution and go after complicated systems that act dumb on purpose.
Ok, easy difficult is done, but now what about hard difficult? Cheat, since this is the only way for it to still look like it is human. Oh, the irony.

But I want to take another approach.
Instead of trying to simulate humans I want to make a very hard AI that uses all features of the game to it's maximum. Given it has enough time to calculate everything.
Then I cut the time it has to calculate for easy difficult, it is still intelligent, it just can't check too much possibilities (can't think too far into the future).

Of course for the hard difficult I will have to cut it too, since it would take too long to fully analyse the game, but it will have more time to think than on easy.

That's why you see turns taking longer on chess when you raise the difficulty while turns take the same amount of time on games like this even on max difficulty.

EDIT: Resuming, they feel that the time between each turn is so important that it dictates how the AI will work, this is really an issue on RT games, but it's retarded on TB games since it's turn based and we have much better computers now then when MoO was released.
So I don't mind sacrificing time to get a better result.

There are no AI in real life so we have to make do with scripted moves.
Chess is simpler than 4x games where you have to consider too many things even for just to make computer move their units.
So sadly as games gets more complex ai gets worse.

 

[tiagocc0]

Chess is simple, but chess programs are made to beat geniuses.
Everything in the end must be scripted, but the problem is that most games choose to fuzz data whenever possible and use too much heuristics to make it as fast as possible and end up with a system that is too difficult to script.

Thus poor AI that cheats is the only way to release the game on time.

EDIT: Like if I was hired to do the AI and after 6 months of work I find out that my system sucks I wouldn't discard it, I would most likely try to fill the holes with whatever I could and hope to make a better system on my next job.
That's the reality of most games.