Install Steam
login
|
language
简体中文 (Simplified Chinese)
繁體中文 (Traditional Chinese)
日本語 (Japanese)
한국어 (Korean)
ไทย (Thai)
Български (Bulgarian)
Čeština (Czech)
Dansk (Danish)
Deutsch (German)
Español - España (Spanish - Spain)
Español - Latinoamérica (Spanish - Latin America)
Ελληνικά (Greek)
Français (French)
Italiano (Italian)
Bahasa Indonesia (Indonesian)
Magyar (Hungarian)
Nederlands (Dutch)
Norsk (Norwegian)
Polski (Polish)
Português (Portuguese - Portugal)
Português - Brasil (Portuguese - Brazil)
Română (Romanian)
Русский (Russian)
Suomi (Finnish)
Svenska (Swedish)
Türkçe (Turkish)
Tiếng Việt (Vietnamese)
Українська (Ukrainian)
Report a translation problem
Discussions Rules and Guidelines
16 
Report this post
I'm not sure whether I should commit that or not or even make it more significant.
Have been playing quite a lot of "Through the Ages" lately, which kinda spoilt my enjoyment for regular 4x-games and had a good AI right out of the box.
I'm pretty sure that Vic Poonstache is critizicing the AI in the state of 1.6.7, not the initial one.
And I am well aware that a lot of the more current reviews do the same.
The thing is that those people have a very different understanding of what "bad AI" and "good AI" means when compared to people like you, me, muhandai and others who actually like it.
I certainly don't think like that.
I mean there's Google's Alpha Zero, that completely dominated the Go- and Chess-scene.
But of course, a multi-billion-dollar-company isn't really comparable to anything else. So when it comes to the best AI's that were written by single people, then I'd probably think of David Wu's Arimaa-AI, that won the Arimaa-challenge for 10.000$ or "krasi0", who coded a StarCraft-Bot that has almost a 100 ELO-lead over the second best bot and almost 300 over mine on sscait and is also able to beat C+ players on ICCup.
Other worthy mentions might be the AIs for games like "Through the Ages" and "Race for the Galaxy", which also have received some well-justified praise.
However, when compared to what is usual in the gaming-industry, I'd still consider my creation as significantly above average.
I asked you to share your strategic approach that allows you to beat the game in a fashion that outclassed the AI so badly that you deemed it justified to call it bad.
If "Impossible" doesn't provide enough of a challenge to you, then I'm extremely eager to know how you do that and what of the AI behavior you could tell was bad.
Please go ahead and name all those things that other game's AI's have, which Pandora AI doesn't!
If the AI only knew 10% of the game it is supposed to play, then of course it would be absolutely terrible and lose to the aliens.
Which parts do you think I missed or made it handle so badly, that saying that is justified?
A lot of development-time went into like: city placement, building-selection, technology-selection, city-specialization according to make the best use of the surrounding boni, unit-designs, army-composition, weighing of units vs. buildings, tile-improvements, worker-management, tax-management, decisions on how to spend credits, worker-distribution, decisionmaking on how to interact with other empires, target-selection for operations like nukes, how to use field-training, unit-movement in order to use the best tiles for a defense-bonus for that unit, unit-spreading to prevent AoE-Damage, surrounding-cities to get flanking-bonus, moving units in seperate cycles so they can adapt to new information, recognizing when to go for economic/research-victory.
There's even code for properly handling fungus-cultivation against other improvements in the late-game to maximize production without sacrificing happyness.
Basically for all the things that I'm aware of as a player, there's also some code in the AI that tries to mimic how I would handle that situation.
And I'd be rather surprised if I just "missed" 90% of the game.
But feel free to enlighten me of all the hidden concepts that I failed to even notice after more than 800 hours of playing it.
I'm asking for deficiencies in the AI's play, that lower their playing strength and detract from their capability of winning the game and you get back at me with something that basically says. "It should not declare war, when it had been at peace for a long time."?
How exactly is changing your mind about whom it is most beneficial to be friends or enemies with weakening the play?
There's no point in not trading with everybody unless you want to go to war with them. This can easily mean 250 turns of peace & trade & good relations and then a sudden war-declaration.
Of course you wouldn't cancel the pacts before-hand to make your victim suspicious and give them more time to prepare.
That being said diplomacy is one thing that was significantly altered between the latest official version (1.6.7) and the inofficial patch, which is being discussed in this thread.
The AI was deemed as too agressive, getting into wars where their advantage wasn't decisive enough or the reason for the war wasn't good enough.
This was not meant to make it easier for the player but to prevent a relatively easy exploit: When you just get a stronger and stronger force but wouldn't declare war yourself, you could still bet that at some point there'd be a war between 2 AIs. And once they weakened each other and moved their defenses out of place you could swoop in and get cities with much less effort than if you were the originial initiator of the war.
In the inofficial patch the AI is way more cautious about starting a war and rather increases their strenght in peace-time some more.
They will also be much less likely to go to war, when there's still aliens running around everywhere, as the 1.6.7 version would sometimes do.
"friends" and "enemies" are not concepts in which the AI thinks.
There's only competitors.
For each of those the AI calculates a value between 0 and 1 depending on that treats them either as competitors with whom it is more beneficial to sign trade- and research-pacts with or competitors with whom it is more beneficial to try and conquer their cities.
It is not trying to "roleplay" real-world interactions. It is trying to win... or at least not lose the game.
And that is the whole point here, isn't it?
You don't want an AI that bases all their decisions around trying to do exactly one thing: Winning.
What you want is an AI, that plays as a side-character in an act in which you are the hero. One that is either your rival or your ally.
But certainly not an entity, that considers you as their side-character in their story where they are the hero.
Well, yeah. They acted exactly as intended. Preventing the runaway from winning in order to preserve their own chance for victory by cooperating with other weaker empires. Not as sign of bad AI in my book.
I'm not sure what you are talking about with "tier 4"... There is no tier 4. The way that unit-strength scales in this game means that 2 units of strength 20 will beat 1 unit of strength 40. Especially if they counter it by using the adequate weapon/chassis-combination, of which doing so is part of the army-composition- and target-selection-algorithms.
Well, there actually is a mod that improves on the 1.6.7 AI by reducing agression, improving-tech-selection and also some changes about tactical unit-movement.
But it does not "fix" the AI if you think that an AI that is trying to win is broken. If anything it tries to win even harder.
It's clearly not a conspiracy... It's just people who are used to and thus were expecting a game that is easy enough to beat without having to invest any time into actually learning how the game is properly played.
In my book a bad AI is one that I can beat first try without any knowledge of the game and it's mechanics.
This was the case when I first played Pandora. Then I worked on it for two years to change that. It is now more challenging with reduced morale and growth than it was before with bonuses to either.
It is no surprise that other companies put only very little effort into their AIs, when the mindset allowing to actually appreciate that is a rare commodity amongst their customer-base.
Let's clear one thing up. I don;t work for Microsoft. I'm not a programmer. I don't use c+ (does anyone use C+ anymore). Your real world examples of AI mean nothing to me as I am not familiar with them. So citing them as an example of good or bad means nothing to me, since I have not the knowledge or experience to refute or confirm your comment.
As to the 90% let's call it hyperbole. As to the bad AI, and "it's an old patch". I'm playing on steam. I just played a game a few days ago. I assume it is the latest patch, as my games auto update, but it may not be. But for arguements sake, lets say it IS the latest patch. Just for this thread.
The reason I posted on this forum, (and maybe 2-3 other forums ever. in the hundred or so games I own), is because I see it has potential. I also see some glaring deficiencies that have been addressed months ago, and some have been rehashed more recently. But since this isn't a critique of "you game bad, me game good. ooga booga". Let's just talk about some specifics.
1- In the tooltips it says the average difficulty (I forget what its called. maybe its called average lol. I dunno, I play a ton of games). It says "the AI has a slight resource disadvantage to the player." or something like that. DING DING DING!!!!!!! That got my attention, because MOST games I've played tend to give the AI a slight ADVANTAGE (or very rarely no advantage). So when I see this I'm thinking, "Awesome, a 4x game. been out awhile, probably had bugs worked out. smart enough AI, to not have to -cheat- on the normal level. This game looks exciting. Let's watch 10-20 minutes of youtube on it, to see if the Lets Plays interest me enough to try it. "
2- I get in game, and indeed it seems as the AI has a disadvantage. How can I tell? Yes I'm a human, as so many smart people point out. "We humans, we smarter than machines. " :) I agree. And I also am aware how long, how many years, and that many people were involved in programming Deep Blue's AI, just to beat Gary Kasparov in chess. So yep, I get it. Ai isn't easy to program, especially in a venue with EXPONENTIALLY more variables than a chess board with 6 unit types and a board with a few spaces to move. I get it. Trust me I get it. Anyway, I can tell because I can easily expand much faster and in better locales than then AI, sooner than them. This was on my first playthrough. Not bad. for someone trying to still figure out the UI, and movement, research etc..... So we're off to a good start.
3- Now the game starts. (this is a summary of every level of the game, I've played them all, and never won on the hardest level, because I get so mad at the AI... as I'll get into) First, the AI SPAMS me "how about a trade pact, how about a research pact, how about this how about this." When I say spam. I mean SPAM!!!!!!. It is nearly every turn. At first this was refreshing, because its un-realistic in my opinion, to play the multitude of other strat games, and the AI never seems to want to get into a mutually beneficial deal. They always just want to stomp you. Whether they make agressive moves or not. So to have the AI, ask me for these was awesome. At first......... Then it gets tedious. They ask and ask and ask and ask and ask and ask and ask. "how about a research?" sure, I'll take one. Then "how about a trade?" sure, I'll take that too. This is cool. "How about maps?" sure why not. Next turn. "how about maps?". Ummmm I havent moved. Next turn "How about maps again. how about maps again, how about maps again." over and over. Feel free to play the game. this is beyond annoying.
4- Check diplomacy status, "ahhh, the orange dudes are Friendly to me. Cool. Must be because we've been trading for 60 turns." uh oh. Green dude, just declared war on orange dude. I got 7 citys, Orange has 5, green has 5. This will be cool. Green hits orange, errrrrrr let me step back. I forgot another awesome AI move. Green dude declares war. Turn is 80 (example). Turn 81, nothing, 82, nothing, 83-89 nothing. Oh cool. 90, he shot over a plane and hit oranges trooper. Turn 91-99 nothing. (seems a little weird the SMART AI would declare war and not do anything for ANY TURNS, let alone 2? 5? 10?). Ok anyway. 20 turns in. Green muscles up, he brings in his mighty t2 tanks. ( I forget the name, and don't care enough to look it up. lets pretend the trooper is t1, the atv is t1.5, the FIRST tank is say t2, the second blue tank, t2.5? maybe t3? The yellow stuff t4. ) Hopefully that is clear enough, so we don't have to debate the specifics of each and every unit, and how they rock-paper-scissors the other Ok. so the Green, uses his Tier 2 tank, comes in. swipes 2 of oranges citys. Awesome. He hit my buddy, now I'm gonna build up a bit more. (though I'm WAYYYYY stronger than him now. )Then I'm gonna do a total human scummy move. move my armies closer, than surprise war. Take those TWO citys. Putting green back down to 5 citys. and orange (now at 3, will be happy, probably, as I just hit someone hes at war with).
5- Execute plan. Move troops closer. Hide them in my nearest city. Wait for it.... wait.. wait. LAUNCH. Bonzai. I send in my troops. a ROFLSTOMP army, It is looking about power 145, I have a second armycoming up soon, It is power 85. WooHoo!!!!!!!!. I attack. City one down. It is now mine, muhahahahahaha. Errr wait. Did I just read that right? Let me scroll through the 32,000 notifications on side of screen (slight exaggeration, but the Notification UI has a lot to be desired). Wait, Four of my citys were orbital bombed, ok. makes sense. Wait, Orange, has now moved troops into my area. Huh???? Oh wait, Orange declared war. MY BUDDY ORANGE. I just smashed the dude about to wipe him off the map. He NOW DECLARES WAR ON ME???? wtf? Ahh well, it;s ok. Orange is weak. I've been watching his 5 citys, only one is on a coast. And he's now down to 3 citys.. Meh. no biggie. ... Wait. WTF!!!!!!!!!!! where did his 8 friggin dreadnaughts come from? How is it they have MORE POWER THAN MINE. I have max tech for dreads, he is NOT in the next era. He CAN NOT HAVE MORE tech than me. I killed a krakon, got a 25% for naval units. I'm militarist race, another 25. And his dread is 58, mine is 52? What the frikk?????????
This is a very very very common example of play. I mean no disrespect, but there is so much wrong with that. If it's not obvious, nothing I say will help. But like a lemming, I'll try. He SHOULD not have those dreads. I am far ahead of him. I scout the ocean, I scout his citys, I read the econ/research status reports. And I KNOW he shouldn;t have them. He either "cheated them in" (totally understandable even though tool tip said they get a DISADVANTAGE), he shouldnt have higher power. and why the heck is he attacking me?
Why do they all ask for my map turn after turn after turn after turn. Is that not AI? I could give more examples. But I think this is enough. To me, this is POOR AI . (continued)
"We can easily make an AI to stomp any and all Human players, (notice how this is different than most programmers saying, we cant beat humans, they too smart), but we don't. We don't because it would cost us a lot of manpower, and resources in house. And it IS NOT FUN. Who wants to play a game where the computer knows everything (as they have to, since they store the map data etc...) makes perfect decisions based on that knowledge, Knows what the player is doing, makes decisions based on that, and so on. We code AI, to do two things. Make the cheating as inconspicuous as possible. So it doesnt make the player mad, yet saves us programming hours, and MAKES THE GAME FUN. We code in mistakes, we code in bad choices. When we say we have a LEARNING AI, sometimes that means we have it see the mistakes a player makes and we will make the same".
Basically some or all of this was in his youtube video. Others I may have read from him. I dunno. ANyway, if u get some time. check it out.
Key... It is NOT FUN or REASONABLE for your ally to declare war when you are helping them.
"hey look china is attacking japan, good thing USA has the largest military the World has ever seen, lets put a stop to it. " Moves military to the region, tries diplo to get china to back off. threaten china, start prodding china. Move forces in. Getting ready to Put china back to the stone age.... Err wait. Wtf? Is australia and Japan now attacking hawaii? "ummm yeah, they see you're weak right now. so decided not to fight china whos attacking them, and attack you. Oh wait, check it out. China and Japan just made an alliance, they are going after Russia too. Sucks for USA and RUSSIA to be the two most powerful. eh?"
PROTIP There would be no more china, japan, or Australia if anything close to that scenario played out. I woud think the AI, could be along the similar lines.
It is always a bit complicated to explain... I liked Pandaora a lot when I played it but was disappointed in the AI. I made a mod that made it slightly better but with a Mod you could only do so much. Then I contacted the Devs whether I could get access to their source-code to try and improve the AI. After some back and forth they agreed. That was like 3 1/2 years ago.
So working on the Pandora-AI was my primary hobby in the 1 1/2 years following. I left almost no aspect of it untouched.
Eventually my changes up to that date (around 2 years ago) got included in the current official patch 1.6.7.
I've mosty been doing other things, like working on the AI for Dominus Galaxia, a game not yet released. But from time to time I did some more work on the AI for Pandora too. The biggest change is a complete revamp of how the Diplomacy AI works because I considered a lot of the criticism about the 1.6.7 diplo-AI as justified. Especially spamming the player with exchange map-requests and being too trigger-happy.
These further changes are not included in any official patch. My contract with Proxy-Studios says that they may include my work but not that they have to. So I distributed my own patch by simply uploading the files that need to be replaced. This is linked in the stickied Modding-Thread here.
Exactly this is the case. We shifted the difficulty-level twice between before I was working on it and after.
The bonus that the AI had on Medium in the Base-Game is now the one it has on Very Hard.
"Very Easy" was removed from the bottom and "Impossible" added on the top because the old names didn't reflect the experience.
And it is debateable whether those 2 shifts are even enough to match the players expectations.
Well, "Objection!"
Pandora's growth mechanic is very different from that of other 4x-games. The pop-growth is basically fixed as long as your morale is positive and food-storage above 0.
So expanding quicker, while not being outright wrong in the long-run is not necessarily advantagoues.
The main advantage of quicker expansion is that it is much easier to fight habitat issues.
But it comes with several pitfalls that you might be overlooking here.
The major advantage of having fewer cities with higher population is that almost all of the buildings have percentage-boni and that cities can specialize for certain tasks.
And that's what the AI does. They will usually produce the vast majority of their military in one city specialized to it while other cities are specialized for food, minerals or science. Further cutting the need of infrastructure compared to unspecialized hybrid-cities.
For each new city you have to invest a lot of resources into getting the infrastructure up that makes the population living there as effective as in the well-developed older cities. And by founding them you funnel pop-growth away from the established cities towards the new ones.
Building suburbs and purifiers on tiles that aren't useful for farming or mining anyways can significantly delay the need to expand. Only when all tiles around a city are used for something then having more cities becomes definitely better.
What I'm saying is that having more cities and a higher Economy-Score is not necessarily a sign for doing better than the competition.
I know. That's the reason why I changed it in my inofficial patch. The AI won't ask the player about trading maps or open borders anymore but still agree to them when they find them agreeable when offered by the player. My original philosophy was to make no distinction between player and other AIs but in this case I did because I also was annoyed by the constant requests.
The AI will still ask for trade, research and nap. I think the nap-request should be reworked to only be used when they want to go to war with someone else as the nap only protects for 10 turns.
Vivian Gardinier and Itala Palomino both are female and probably don't like being called "dude"! ;p
The diplomacy-AI and tactical-AI are seperate classes.
The the war-declaration basically just flags someone's units as attackable.
The tactical-AI doesn't even know if it was the one who declared war or the one that got declared war on. It doesn't need to.
So its decisionmaking of whether an actual attack on a city shall be carried out is determined by other metrics that take additional information into consideration.
While having full understanding about that is not important in the context of what Green did to Orange, it is indeed relevant and debateworthy once you start accusing the AI of cheating because you don't fully understand how the strength of a unit is determined and thus are surprised how their troops can be stronger than yours.
Each unit consists of 3 major-components and a potential special. The combination of the 3 major-components determine their base-strength, what other units they get bonuses against and what tiles they get bonuses on.
Calling a unit a certain tier points to a misunderstanding in this regard because there's simply too many combinations to warrant such simplified categorization.
If the chassis of a unit is blue-tier while the weapon and armor are still white-tier, then this unit will obviosuly be a lot weaker than one where all three components are blue-tier. The armors even have three sub-tiers per era so a full-blue-III is stronger than a full-blue-I or -II.
But in addition to the components there's quite a bunch of additional factors that impact the units strength. And you can see all of them when you select one of your units and then hover over one of the enemy, you want to investigate.
Most prominent is the units level. A unit at level 7, built from a city with combined tier 1 and 2 barracks is way stronger than the same unit built from a city with no barracks. Then there's an operation called "Field Training". With that you can get the level of your unis to Level 10. And of course have I taught the AI to use it!
In the gold-era you can even build Level 15 troops, which are more than twice the strength of a Level 1.
Then there's a total of 5 Advancements that improve base-strengths:
3 researchable from the tech tree and 2 researchable after killing certain aliens.
The two you can get from aliens both work on ships.
Equipped with that knowledge you should easily be able to determine why their ships are 58 strength while yours are only 52.
My guess: The AI had both of the +25% strength-alien-techs stacked and their ships were Level 10 compared to your Level 7.
The theory could easily be checked with the savegame. And even if it is not exactly as I guess, there clearly is an explanation without cheating!
I am 100% sure that there's something missing in your timeline.
Like for example Green and Orange making peace immediately after you declared war on Green to take those cities and most likely also signing a non-agression-treaty.
The AI would never declare a war when it already is partaking in one. Especially not when it is losing.
The AI only cares very little about the past. The past is irrelevant to make decisions for the future. You could even say that considering the past is a weakness!
So all they see in that situation is: "My 2 neighbours are at war, blue is closer to my territory and I just signed a non-agression-pact with green, so it might be a good idea to backstab him and maybe get my cities back."
If you can tell me how this decision is wrong from the perspective of Orange, not your own, then I'll gladly agree, that it was a bad move. But from everything you've written I don't see how this was a mistake and thus a sign of "bad AI".
Load the savegame of the turn before and scan all possible angles they could have come from. Ships have 4 movement, so the possibilities are quite limited. Or even go further and use the "/resign"-command, that allows you to see the entire map.
They did most certainly not come out of nowhere, like you are trying to imply hear.
The Econ/Research-Status-Report is irrelevant, if you are concerned about military-strength? This kind of data is found on the military-status-report and also in the diplomacy overview. There's no way that the 8 x 58 dreads were not taken into account in the strenght-comparison. Load the savegame of 5 turns before, use the aforementioned /resign command and you will see that he already had at least 6 of them!
I can't even tell you how sadding these accusations are to me!
I spent 2 years of my free-time to make an AI of which I think is actually good at the game it plays. And then someone who loses their first game comes here and tells me that the only possible explanation he can come up with is that the AI must cheated! That is exactly why companies like Firaxis don't put any emphasis into the AIs for their games. Because people not only don't appreciate it, no they'll claim it cheats.
So you might aswell make a bad AI, have it cheat, openely communicate that it cheats and people will feel like heros for beating it anyways. Saves development-cost, pleases the crowd and lowers everyone's expactations for your next game.
Believe what you must believe to feel better about yourself, but I'm seriously sad of getting accused that I spent 2 years worth of my freetime for nothing but implementing a cheap "create unreasonably strong units out of nowhere"-cheat. Seriously, play something else and leave Pandora for those who are actually capable of appreciating a competent AI when they encounter one.
As I said earlier. By understanding of how unit-power is calculated this can easily be explained. But then again, that does indeed require to get into specifics. I could even sit down and mathematically determine the exact components, level and boni for a non-Imperium-Dreadnough to get to 58 strenght. But I fear that endeavour would be just as unappraciated as my AI is.
His AI was the last Civ-AI that actually showed some signs of competence so lacking in the newer installments.
I've seen that video long ago and think he was inspring for me to get interested in AI aswell.
I'm not even sure the newer Civs even had an AI-lead-designer since their AIs seem like they were rushed under massive time-pressure by someone who's primary task originally was something else.
I know, right? And that's a statement that I agree with from my personal experience. I mean I would not necessarily subscribe to the combination of "easily" and "any and all" but I think that by putting in enough effort the vast majority can be beaten. And this is more or less what I was trying to accomplish for Pandora.
Erm... most of those things would be considered cheating. Just because the map-data is stored on the computer doesn't mean the AI has access to all of it.
In Pandora each AI has a seperate storage for their own map data only including what they actually scouted. I had a bug in the past where I accidentally let them know when a city that was out of their vision was undefended. It was very obvious and quickly fixed.
But I don't even think that knowledge cheats like that are essential for a strong AI.
I'd say that the management of the economy is probably the most important thing. The order in which buildings are built, the tile-improvements chosen for each tile, the distribution of population in order to maximize the efficiency. These are the parts in which my AI shines and it never gets lazy to optimize every city every turn and can get an edge this way.
As I mentioned earlier, "Hans Lemurson" told me about how exactly he approaches this with a concept he called "return of investment"... using actual Excel sheets to determine the values for each building. This is something an AI can do in a blink of an eye. So it was a no-brainer to get it into.
The diplomatic behavior of the AI tries to mimic a player called "Zak0r". He won all of the multiplayer-events for that game in it's early stages and he always found a way to beat my AI even at the highest level... He hasn't tried the most current version yet, though. And reducing the boni of the highest level at some point really helped him to save that claim.
The AI in diplomacy is a backstabbing bastard because Zak0r was one. And he was successful with it despite others anticipating this. Of course all of the rest of his play was also top notch, but his "diplomacy" was what set him truly apart.
Well, that's something you won't find in Pandora-AI.
Obvious mistakes are more or less what I was asking you for to provide to underline your criticism. Mistakes from their point of view, that is.
None of the mistakes the AI makes are intentional. One that I'm aware of is forgetting what it has already seen in the same turn. This is an issue with fast moving units. For example: An unit on their way to an enemy city encounters a strong enemy force on their way. It sees it and wants to run away. But as soon as it moves one tile out of sight it forgets about the enemies and moves back in.
Fixing that would be complicated because it would require to give the AI a temporary memory that exceeds their knowledge of what they currently see.
I even have specific code for the two sub-normal-difficulty-levels to alleviate the negative impact of their morale-malus:
Because they are forced to run a lower tax to avoid stunting their growth even further, the usual army-requirement for safely-expanding is ignored. They will take the risk of doing an unsave expansion because they can't afford the army early on but also can't afford to not expand in order to prevent overcrowding, which worsens their situation further.
That can and indeed sometimes does result in them losing their 2nd city to the Natives. So this might seem like a mistake but it isn't. It's a gamble.
You may have perceived them as your ally. But they didn't love you back. To them you were "A competitor who's military is too strong to take their cities, so we act nicely towards and trade with them so we get at least some benefit out of their existance." that turned into "A competitor who is currently has their hands full with a war against another faction. There probably won't be a better opportunity to try and get additional cities."
To me it is not fun to play against an AI that isn't trying everything (within the rules of the game!) to win.
The 800 hours I have in this game on steam and the many more not recorded here because I played them with my local build are prove of me having fun with the game in a state where the AI is trying to properly compete.
And guess for whom I primarily did all that: For myself! I am not financially tied to Proxy-Studios at all. So my only gain was the opportunity to work on a game that I like and to learn something while doing so.
That means I didn't have to consider the temper of other players. I simply assumed they would like the same things that I liked. Only afterwards I would learn that this wasn't the case and many are more interested in having an AI that roleplays rather than one that tries to optimize all aspects of the game, including diplomacy(!), in a way that would allow them to be better at the one thing that is important in any competition: Winning.
I played of Pandora yesterday, after not having done so for quite a while.
Difficulty: Very Hard, everything else: Random
It was a really small map and I started as Ambassadors really close to Solar Dynasty. I tried to rush them by pumping out some Colonial-Troopers early on. But the rush failed. I had to sign a peace-treaty and try to play catch-up with development.
After I met everyone I realized that my position was really bad. Like my neighbours, I only had room for my 1st city.
Red declared war on me two times but they didn't really attack. I also had no means to bring the war to them, so it was a stalemate.
At some point Imperium and Noxium declared war on Togra who lived on the Island left to me. I tried to get their capital but it fell to the Imperium. In the meantime Divine Ascencion had declared war on Solar.
After the Imperium had taken the Togra-Capital, they immediately declared war on me, while Noxium joined in on the war against the Solar-Dynasty.
It was a really tough war of me trying to hold my city against the Imperium. I used buyout after buyout and depleted all my savings and held the city by the skin of my teeth.
In the meantime the Solar-Dynasty had fallen to the Divine Ascencion.
In 1.6.7, I think that would have led to Divine Ascension and Noxium declaring war on me. But this is not what happened here: Instead they declared war on the Imperium! And the until-now-inactive Terra-Salvum joined them on that and apparently even got the majority of Imperial cities.
I was kinda relieved... but what were my realistic options here? I continued to build up my city some more, get new Techs and troops but it didn't take too long for the Imperium to have fallen to three foes at once.
So Divine Ascension declared war on me and there was no way to stop their already partially "golden" naval-forces. So I quickly joined the 3 Factions that had been eliminated before me.
Now in 1.6.7 the Onslaught would have continued. But the 1.6.8 diplo-AI is smarter than that and determined that whoever would declare war on one of the other factions of similar strength would just allow the remaining faction to pick their target after both are weakened. So there was no further war and Terra Salvum eventually won with a Science-Victory.
And guess what: I enjoyed that game! I don't have to win in order to do so and feel like I'm a part of the story. Seeing how the AIs all played appropriate according to their situation and faction-traits still made me feel good about it. Pandora is always a hard and punishing challenge. It is made for those who like hard and punishing challenges.
I get flashes of the Terminator franchise (Skynet in particular).
In my current campaign (4-way skirmish), I barely managed to pull through the Mezari skewer and their infestation because my island had the majority of the invasion. With careful strategies, small wars, bluffs and other strategems somehow paced my way to the top (although currently in war with both the 2nd-strongest faction "divine" and weakest faction "ambassador").
Really, I enjoy the feisty drama that takes place in the game. Dog-eat-dog AI.
I know that Pandora's AI is like this, and so I don't mind it. In other genres, I also enjoy the opposite ideal, which is allies unto death, where I'll heroically stick with a weak ally and not betray it, no matter the temptation or perceived benefit to me. That mode of thought works best in a setting where loyalty/karma/alliance is explicitly modeled as a first-class value, as in some RPGs or 4Xes. (In other words, I'm just being pragmatic, after all, shunning reward of one kind for reward of another ;)
One game of Pandora is kind of too short to admit both extremes in diplomatic behavior.
Just uploaded it and put link into Mod-Thread.
I did not expand before my captial was at size 34 because I was right in the middle of the continent and was surrounded by alien hives. And while I managed to keep my army alive I constantly had some of them healing so it took quite a while before I was in the position to take out hives.
But thanks to not having to deal with useless water-tiles I could keep up with fighting the unhappyness due to building suburbs and purifiers.
My second city was close to some mountains, which my capital lacked and this really helped with the production so I could get my military strong enough to ward off enemies.
It worked: I was not involved in a single war.
The others also were relatively peaceful. There was a campaing against Togra led by the Divine Ascension in which Vermillion and Preston partook. I thought about participating too but since I already had the science lead and would have had to walk my army really far for that, I decided not to. In the end Vermillion got all three cities of Schreiber.
There was not really a point in taking any risks. So I just maintained a strong army-production while focussing on science. At the end I had mayb 5% advantage in science, was at 70% military of Noxium/DA but was almost last in economy since I only had 5 cities while most of the AIs, especially Noxium and DA had significantly more.
Heid had been the weakest throughout the entire game. I think he must have had some serious losses to the aliens early on. But noone declared war on him until like 5 turns before the end (at that point Vermillion declared first and Terra-Salvum followed).
It is hard to say whether the AI could have done anything better. Their play looked solid and appropriate for the map-size. Had the game gone on any longer I would have had a lot of trouble due to having fallen behind in land-grab and neglecting-fungus-tech in order to get through the tech-tree faster.
But it doesn't matter. I won. Next time I'll try impossible! :D
The fact that your city was not coastal significantly helps with Aliens 2 and with defense against faction war, since factions can't easily find / reach you when they're on another continent/island.
Also, land-grab is a meta where both coast & fungus act as defense - yet to pull it off succesfully myself though. Generally the land-grab becomes a problem well before I can complete the terrain-makeover.