Solver's 1.3 thoughts

[Solver]

I bring you “Solver’s 1.3 thoughts”, the sequel to ”Solver’s 1.1 thoughts”. It seems that I have become a Civ game analyst lately, so I’ve been posting some of my thoughts.

I didn’t play version 1.2, but got some time with 1.3. From what I’ve seen so far, it seems like there’s enough material for a new thread. One of the great things about GalCiv2 is that you can essentially play it during breaks from doing other stuff.

1.3 brings all sorts of improvements, of which the AI ones are the ones I was the most interested in. Still, a few words on the other options. The edge scrolling option has been added, which I find a nice change from grabbing and moving the map with the mouse to scroll. Nonetheless, I would also like to see an “arrow button” scroll option added later. That is, make arrow keys on the keyboard scroll the map instead of moving ships. When I play Civ4, I tend to scroll around almost exclusively with the arrow keys, and would also like the option in GalCiv2.

On the bug side, I have noticed a strange one where the minimap will not redraw correctly if a planet flips to me due to influence. Thankfully, it’s a rare occurrence, and it is so far the only bug I have noticed.

One option I was immediately excited to see is allowing the AI to use more CPU-intensive algorithms. I checked that without a second thought, and can’t really comment on it outside of my general AI comments, as I haven’t played any without that option in 1.3. Nonetheless, this is one point I would love to hear clarification from Brad about – specifically, what sorts of algorithms / how many algorithms does this option enable?

So I started a new game with 1.3, Challenging difficulty, which now means all AI opponents are Bright. If I’m not mistaken, that means they use all their algorithms, but do not get any additional economy bonuses (number cheats, in fact). I deliberately played very similarly to how I played in 1.1. I colonized some planets, and then continued to build my economy and research. I maintained almost no military, just a few fighters. In 1.1, I found that the AI sucked horribly, even at Challenging and higher, at waging war. Basically, it would declare war first and prepare for it later. As a consequence, AI declarations of war were actually a warning. By starting to build a military after war has been declared, I used to be able to effectively counter AI assaults every time.

Well… as I was very glad to find out, that was no longer the case here in 1.3. The Arceans declared war on me eventually, and a few turns later, their fleets were on me. A particularly good touch was the fact that an Arcean fighter just went over my territory shooting starbases out of the sky… space. I hadn’t bothered giving them any sort of defensive capabilities, so the Arceans took advantage of that. Torians joined the war, and I got my behind handed to me just like I should have for ignoring military completely until war was declared.

That doesn’t mean everything was perfect, however. I’ve noticed (in this game and subsequent runs) unescorted AI Transports. Pairs thereof, specifically. The AI might dominate the battlefield alright, but leave a couple of undefended Transports just outside a planet, allowing them to be shot down by any one ship. I think I also know the reason for such behaviour. A Transport ship takes 5 logistics slots, so two take 10. At that point in the game, the AI civ’s logistics ability allows for fleets of size 10. Thus, two Transports is all it can include in a fleet. And the flaw in AI reasoning seems to be that it will not put any ships in a square if that square already has a fleet with all logistics slots taken. What the AI should obviously do, though, is put some fighters in the same square as those two Transports, even if it is two separate fleets, as GalCiv2 has no limit on the number of ships you can put in a square.

Also, the AI war preparation, while solid, seems to have another imperfection. From what I’ve observed, the AI will mass its troops somewhere near the border, then declare war and move in. While this is much better than what it would do in previous versions, it would be wiser for the AI to actually move the fleet first and declare war later. Declare war when it can attack – that is, follow a declaration of war with attack immediately. There’s no penalty to crossing borders, there’s no attrition damage, so there’s no reason to declare war before you attack. I would, though, also recommend that such behaviour be limited to “Intelligent” and cleverer AIs. That is, still have AIs that newer players play against declare war at the border, but have the more clever ones declare war simultaneously with attacking.

Still, the new AI behaviour caused me to use some new ship designs. I designed a recon ship with good speed and sensors, and used these at my borders to notice possible troops buildups or approaches. The fact that I found this useful clearly shows that the AI has improved. By the way, a badly needed feature is a patrol order for units (unless I missed it, which is not impossible). It gets annoying moving a ship along the border manually back and forth to observe more space – the ship should be able to do it automatically and stop if something gets spotted.

AI ship designs have once again definitely improved, as I’ve seen better mixing of weapon types, and I’ve definitely seen the AI adapting its defenses very well when I only used one type of weapons.

I loved how the various difficulty levels no longer mix different AI intelligence levels together. Previously, I often found it ridiculous choosing a difficulty level that should provide a challenge only to find that some AIs are strong, whereas others are absurdly stupid. This sort of a thing always benefits the human more. When there are some weak/stupid (which is the same, essentially) AIs, they can be exploited. And the human is always better at exploiting, so I’m glad to see this behaviour gone – or rather, made optional.

Warfare is still extremely advantageous, as I commented in 1.1. And the attacker still has the advantage in war. Allow me to quote myself from my 1.1 comments. “Star systems have planets that are within one turn, two maximum, of each another. Therefore, once you have your fleet at a planet, you essentially are threatening all the planets in the system – and if you have enough ships to do so, you can take over an entire star system in one turn. Just need to have enough Transports.”

I think Transports don’t work well enough in the current implementation. The thing is, you can put Transports back on a planet, which makes the planet’s population go up. What this really means is that you can’t have too many Transports in a war. If you built 4, and only needed 2, then you can put the surviving 2 on your newly conquered planets – boom, a gain of population! OK, you had less population while they were in space, but chances are, it didn’t last quite that long. In fact, after you conquer a planet, it has low population. Population grows in percents – that is, a planet with 10 billion will gain a billion quicker than a planet with 300 million. That makes sense. But it also encourages to execute “population relocation” moves after conquests. I just put some Transports on my newly acquired planets, and it gives a great benefit. If no Transports survive, I’ll actually build one or two on my high-population planets and move the population to the newer planets. This results in an overall boost.

That’s the sort of thing which I don’t think is good for the game, however. It’s extra micromanagement, and it’s something that the AI doesn’t seem to understand. This is one of those cases where I would advocate changes in favour of streamlining the game. Specifically, I would make Transports unable to return / land on a planet. Once a Transport is in space, that’s it. It stays there until destroyed. Such a change would serve to improve the game in two separate ways. First, it will get rid of the “silly but beneficial” population-moving aspect. Second, it will make war slightly more of a gamble – if you build too many Transports, well, that’s basically some population you’ve wasted.

Approval still doesn’t matter nearly enough. Very low approval (below 40%) will stall your population growth, but as it stands, it’s still a good idea to ignore approval and tax your people a lot. More taxes = more money = ability to allocate more spending = more research and stuff built. It’s something that should be a strategic decision but isn’t really, because you gain much more from taxing your people to 41% approval, and not enough by keeping your approval high. Having 100% approval is great, but having 80% doesn’t bring enough benefits. So approval still only matters once you gain the ability to switch to advanced governments.

Sadly, the problems that existed in diplomatic negotiation process back in 1.0 still persist. I’m not talking about the AI decision making now, but rather the process of negotiating itself. It’s still tiresomely difficult and annoying to get a good deal. Offer to trade Advanced Trade for Laser II, that’s a deal. Now see how many credits they’ll agree to throw in addition to that. 100bc? Yeah. 110? Yeah. 120? Nay. 119? Nay. 118? Nay. 117, yeah, commence trading! Brad said, if I recall correctly, that there’s a can offer anything for anything system, which makes any “best deal” hard to interpret. I agree with that, but that’s still no excuse for having such a tiresome negotiation system. At least as far as money goes, there should be a “how much you’ll give” button. In the same box where you select how many bc you want them to play.
See, gaining money in tech trades is currently a purely mechanical process. There’s no strategy involved. You have a green deal, but can add something else to it while keeping it green. Clicking on buttons to get the biggest number with which the deal stays green is a completely mechanical process, which is also entirely boring. But from an abstract gameplay perspective, that’s what it makes sense to do, because, abstractly speaking, it’s better to add 118 credits than 110 to your treasury. I know that I can myself be accused of micromanaging too much given how I bother with such details. Maybe. But I don’t even always bother to narrow it down to 1 credit, I often go in tens, like see that they’ll give me 210 but not 220, so I take 210. That’s still not good.

On the same diplomacy screen, there are still no tooltips, for which I have been asking since my first posts on this game. When moving the mouse over possible techs to trade, there should be the cost of that tech displayed. I can figure out that Laser III is a more advanced tech than Laser II, but is Soil Enhancement or Planetary Invasion more expensive to research? I don’t know the tech tree by heart, so such tooltips would be very nice to have.

Now some more positive comments. You can no longer form alliances at “Friendly” relations, and this is one awesome change. Previously, I would often research the yellow tech branch until “Alliances”, making it my goal to have one or two civs Friendly by the time I get the tech. That was usually not a problem – it was merely a matter of selecting the right civs, taking in account ethical alignment and other stuff. Some trading, and you’ll have Friendly relations no problem. This no longer works, you need Close relations to form an alliance, and that’s considerably harder to get. Thumbs up on that one, absolutely.

However, with relation to alliances, I’ve also noticed a negative thing, that reminded me of vassal states in Civ4: Warlords momentarily. I was playing a cakewalk game as Terrans, where I went ahead of the other civs a lot at one point. Everyone either loved me or feared me, due to my impressive lead in all areas. Now, I signed an alliance with the Drengin. They had, strangely enough, been the weakest civ that game – well, they failed to gain any planets, so only had their two initial ones, and no one else liked them much, given how no evil civs were on the map. And there comes the shocker. The Torians declare war on the Drengin. Uhh, hello, lost your marbles anyone? The Torian military is more than capable of handling the Drengin, but they’re my allies. I declare war on the Torians in response, and send a few of my ships to the Drengin system. Given my extreme technology lead, my “BaltarStar” class ships handled the Torians with ease.

What happened was that the Torian AI absolutely ignored the Drengin allies when considering war. It only considered the actual Torian-Drengin balance of power (which was very much in Torian favour), without considering possible allies. Matters didn’t become any better when the Arceans (who totally loved me) also joined the war vs. the Drengin. I’ve seen AI dogpiling behaviour in 1.3, which isn’t a bad thing, but again, think of the alliances! If the AIs are so afraid of me that they regularly send me gifts, they damn sure shouldn’t be declaring war on my allies. This is an oversight which I hope to see corrected, thankfully, that part shouldn’t be too hard.
I did, though, see several examples of better AI decision making in warfare. The AI has become much better at figuring out which target to attack, or when to retreat instead of attacking. It seems that the AI generally prefers fleets of more fighters than of fewer frigates. Maybe this varies with the specific AI, I don’t know yet. Either way, this is an observation and not a criticism.

Oh, the new combat system. Well, there’s another huge improvement. The attacker no longer fires first, which means that a powerful attack fleet can no longer destroy anything without taking damage. There isn’t a lot to say about this change other than that it’s a great one. It removes some of the tremendous advantage that attackers used to have, while not shifting the balance in favour of the defenders too much. Thumbs up.

1.3 had a few other random moments that made me happy. One was messages from civs that were not trade proposals – such as a warning that the Arceans are getting too much influence. It really brings the game to life when the other civs say something to you – say as in send a message and not a request. That’s good. Another happy moment was when I saw that the AI, at least sometimes, does decent research of the yellow techs – one of my past criticisms has been that it’s usually ignored, giving the human a huge diplomacy and influence advantage.

Speaking of diplomacy. I think it might be a good idea to add periods of time when you can’t declare war against some civ. That would be good merely for the sake of exploit closing and generally adding strategic quality to the game. For example, prevent redeclaring war for 10 turns after signing peace, for 10 turns after receiving tribute (which was asked for), and 5 turns after receiving a gift. Currently, it’s a good idea to ask a civ for tribute just before you attack it. Get the tribute and still declare war. I really don’t see why this should be the case.

Peace deals. Ugh. The AI still doesn’t understand them. It seems to be rather stubbornly insistent on not signing peace too soon after a war begins. I fought a war against the Drath. They controlled two star systems. One had decent planets and also had two resources, the other was somewhat crappy – well, mediocre. I declare war and take over their good star system on the same turn. The planets become mine, their mining starbases are gone. Next turn, I go to see if they’re willing to sign peace. They’re not. I’m curious as to why – really, what is the AI thinking? Is it under the impression that I’m chickening out if I want peace on the next turn or what? They’ve just gotten a beating in one turn, sounds like they should be begging for peace so that they can live, but not.

On the other hand, there was that Torian war I mentioned above, the one where I interfered on the behalf of the Drengin. I didn’t contact the Torians for peace for quite a long time. I refused their first peace offer. Now, some time after that, I contacted them, and was able to extort a huge sum of money for peace. However, I hadn’t taken any of their planets or destroyed any of their starbases – I had merely kept destroying ships they were sending to the Drengin system. I’m not sure what happened – maybe the AI gets more willing to pay up after longer wars, or maybe it uses ships killed as a measure of how it’s doing in a war. If it’s the latter, a false estimation of success, then the AI should understand that losing a planet is a bad thing, and losing a whole system quite likely means that they’re screwed in that particular war.

I’m finishing for now. 1.3 is a great update, even better than 1.1, for which I had a lot of praise. As always, there’s still room for improvement. And as always, my goal remains not to criticize or praise, but to attempt to provide objective and constructive insight about the game. Thanks to Stardock and Brad again – it’s a particular pleasure for me to see that a whole bunch of things I mentioned in my previous GalCiv2 posts got some attention in 1.3.

[Solver]

Time for an update! Gah, what have Firaxians and Brad turned me into, some crazy analyst who can spend more time writing than playing .

Getting straight to it, comments on minor civs. I like those folks, even if they are just asking to be eliminated when they’re located close to you – what not with the high planet quality they always have. But they can actually be good trading partners, so eliminating them isn’t always the best idea. That’s good. I noticed, though, a couple of strange things in their behaviour.

First, I saw on a few occasions minor civs build influence starbases. Seems to me that they never should – after all, they don’t have any actual influence as in they can’t flip enemy planets. And, if I understand correctly, their planets also can’t be flipped. It’s possible that I am missing something here about how they are supposed to behave, but influence starbases seem rather pointless.

Second, I had a minor civ trigger a false smart alarm for the AI. The GalCiv2 AI is programmed to, at the higher intelligence levels, recognize troops buildups as a sign that an invasion is imminent, and respond appropriately. So I had a minor civ’s planet in one of my star systems (I controlled the other 2 planets in it). I was fighting a war with a major civ, and preparing a number of Transports at planets in that star system. Just as I was preparing to send them to the front lines, the minor civ saw that there’s a bunch of Transports lined up and declared war. Those Transports were intended for my ongoing war with a major civ, but there you go. This is something that I don’t expect to happen very often, though, and it’s sure impossible to avoid – after all, you can’t have the AI know the player’s intentions.

Which brings me to a thought, I have to try a feint someday. Put some troops near a civ’s planet to make them thing I’m going to attack, thus detracting their attention from their economy and making them build military. Scare tactics.

Next up, influence victory and planet flipping. I got so excited by trying out influence in its various shapes and forms that I am now trying to win by influence all the time. And it is my belief that the 75% threshold is too high. One of my games, I was aiming for influence as the Torians. I wiped out some evil civ militarily, build a load of influence starbases and everything. The remaining civs in the galaxy had excellent relations with me. At controlling more than two thirds of the influence, I saw that I’m vastly more powerful than the others, but can’t win an influence victory until I eliminate some other civ. So I ended up eliminating the Arceans through culture – which allowed me to get to 75%.

My point is, currently, the influence victory doesn’t require just great influence. It requires you to wipe out some other civs, militarily or culturally. Otherwise, in a galaxy with 6 more civs, and you need 75%, you’re talking about you having three times as much influence as those 6 civs combined. Ack.

One option is to simply lower the threshold to 66% (which still takes effort to achieve), but I have something else in mind. I believe that, if you sign an Alliance with someone, their influence should count towards yours, for purposes of the Influence victory only, no UN combination or anything like that. That would be a good addition to the 1.3 change to Alliances, which are now finally hard to get. Then, one option to win an influence victory would be to get to close to 70% yourself and sign an alliance with someone.

Planet flipping I have grown very fond of. I love researching the yellow techs and putting two Influence starbases in such a way that they would flip a neighbour’s star system to me eventually. On my luckiest attempt, I’ve flipped four planets with two starbases – those four planets belonged two to each civ, and thus both of them were wiped out. Wow. It costs money, but not that much money. 1000bc for starbase modules, to push them to maximum influence, and possibly the starbase initial construction costs. I would say that isn’t so expensive for a method of absolutely peaceful conquest.

The AI doesn’t respond aggressively enough to this. Yes, I’ve received warnings from them that they don’t like my starbases (though not always). Still, they don’t seem to declare war over it, and I have even managed to maintain Friendly relations with civs up until the moment they’re wiped out by my culture. The AI should probably consider influence starbases near its planets as something more dangerous. To the point of declaring war if they feel up to it – after all, a lot of resources spent on influence also likely means a weaker military. Having a civ send a stern warning about a starbase, but then being able to assimilate them, doesn’t feel right – particularly when you manage to do that to different civs, several games in a row.

On a related notice, it would be great to see this strategy added to the AI’s arsenal. It will place influence starbases in its systems if these systems are at the risk of being overran by foreign influence, but I haven’t really seen offensive influence starbases – after all, they have to not only be close to a rival planet, but equipped with a considerable number of modules.

By the way, what’s up with the AI preferring to build several Economy star bases per a star system? I usually see something like 2 Economy starbases placed close together, instead of 1 with good upgrades. According to the documentation, the bonuses stack, but wouldn’t it make more sense to first fully upgrade the first starbase before building a second? Two weak starbases seem a worse solution. However, the AI definitely loves its mining starbases – I always see it going for resources only, and see its mining starbases equipped with additional modules at a good pace. That’s nice.

I finally got around to playing with the governments some. Previously, the situation wasn’t quite right for that. The idea is very neat, particularly that of the Senate. However, there’s a problem. You need control of Senate, which is determined at the elections. The chance of losing control is the one and only drawback to running an advanced government. Elections are determined by your approval rating. In theory, this should force you to pay more attention to your approval when running an advanced government. In reality, you can keep watch on when the elections are coming up, and lower the tax rate dramatically the turn before the elections. That should let your party win the elections alright. I’ve also used this when the Senate disapproved of a war I wanted to start. I lowered the tax for high approval, and could declare war easily. I think the optimal way to handle this is by having mean approval between the elections matter. Thus, when new elections happen, what matters is your approval throughout the entire time since the previous ones. The having high approval for one turn won’t help you if you are running low approval for the rest of the time.

Another thing is special planetary attacks. You know, Mass Drivers, Tidal Disruption, whatever. As the game progresses, you became unable to take larger planets with traditional invasion technique – at least not without losing a bunch of Transports first. So you have to resort to some of these special tactics. These, in turn, have a disadvantage in the form of loss of planet quality and/or planetary improvements. However, it turns out that there is a problem – these negative effects do not happen if the invasion fails.

Basically, here is what you can do. There’s some planet which you can’t really conquer just yet. You estimate you need three Transports with special tactics – with just a little experience, those estimates are pretty easy to make. You send in two Transports, choosing a powerful special tactic, such as Mass Drivers. You lose, taking out many of the defenders. Then you send in another Transport or two, using a non-damaging tactic – either the standard invasion or Information Warfare. Yay! Planet taken, no damage to planet quality or improvements!

This ain’t good. By sacrificing a few Transports here and there, you can take planets without having to face the negative effects of special tactics. If you had used enough Transports in the first place, you’d still likely lose at least one, so your Transport losses don’t increase by much with this tactic. However, you do get enemy planets intact – and in the long run, it’s definitely more than worth it. After all, capturing a class-20 planet is so much better than having it reduced to class 11!

Unfortunately, the obvious solution also has a flaw. That is, if special tactics are made to damage planets regardless of the invasion’s outcome, that is also a possible exploit. Send single Transports (one troops module) at enemy planets, use Mass Drivers, fail, but considerably weaken the enemy planet at the cost of just 500 million population. Launching such attacks against key enemy planets would also be a very powerful tactic. On the other hand, you can’t invade a planet unless its orbit is empty, so you wouldn’t be able to just send Transports all over the place. And the AI puts a very high priority on maintaining some ships in the orbit. So maybe it wouldn’t really be much of an exploit. I’m unsure, though it would be interesting to test.

Speaking of exploits, what’s up with selling the AI obsolete techs? If you put Laser II and Laser III in the same deal, the AI smartly recognizes that and tells you that Laser II is useless if you’re also selling Laser III. Great. But what you can do is sell the AI Laser III for whatever, and then, separately, sell Laser II for something else (you can at least get a sum of money out of it). Why? Many older techs are indeed useless if you have the newer ones. There are exceptions, but in the situation above, the AI should see it already has Laser III and refuse to give absolutely anything for Laser II.

On a positive notice, I recently noticed the Terrans research the mass driver tech branch well, being ahead in it of everyone else. In a smart move, they refused to trade those techs point blank. I hope it wasn’t a fluke but instead the AI actually trying to maintain a monopoly on these techs it’s well ahead in.

I complained, in my first 1.3 analysis post, that the AI is sometimes weak at figuring out when to sign peace. I’ve got a suspected reason for it today. It seems that the AI, when deciding about peace (and which side should pay for it) uses the comparative ratings of itself and the civ it’s at war with. That is, say, you’re at war with the Dratha. Later, if the Dratha have more population and economy than you do, they’ll think you should pay for peace – possibly even if you’re kicking their butt. I don’t have access to AI code so I can’t check anything of course, but this is really the only explanation I can come up with for some of the AI peace-making decisions I have seen, in wars against me and other AIs alike.

A good measure of who’s winning/losing the war would be just that, relative increases/decreases in various numbers since the war began. When a war starts, the AI should memorize its population, industrial and military strength, the number of planets it has, and such info. Then, when considering peace, gains/losses relative to that should be considered. If the AI loses three planets and 20 billion population in a war, then it is losing, and should understand so – even if its research or industrial capacity outstrips that of the enemy. At peace time, capacity matters. At war, it’s the actual losses or gains you have that matter.

And I just became even more assured that forced peace periods are needed in some circumstances. Being able to declare war the turn after you sign peace just isn’t right. If you’re close to exterminating an AI, you can sign peace, take all / nearly all their tech and money for peace, and then just redeclare war and finish the job. At the same time, if you pay a superior AI for peace, you can’t even be sure if you bought yourself any time at all – which is why some forced peace period seems like a good idea.

My final point for now is regarding the whole life support thing. It would be better if life support ranges scaled to galaxy size. Playing a Huge galaxy, life support technologies become important, and there are indeed many places you can’t reach. Playing a Medium galaxy, I put a low priority on those technologies, and have never used a life support module on a ship. Scaling would simply make life support range relevant to any galaxy size.

Oh, another possible tweak would be to make influence / economy starbases not extend your range. Only planets and military starbases. Not a major thing, but I’d have a good feeling about such a small tweak.

For now, that’s it. As always, thanks for reading!

[Verrucosus]

Thank you for putting so much effort in analysing and describing the changes. I don't have any substantive comments, because I'm not playing GalCiv at the moment (too much time spent on Civ and ... work), but I will get back to it and then your essay will help a lot. Anyway, I just wanted you to know that you are not posting behind closed doors.

[Solver]

Thanks . I know there are people here and there reading!

[dearmad]

I said this over at the official forums, but AFAIK "Bright" isn't par AI, Intelligent is. One more notch up. Once you play at that level, then you can fairly crit the AI. What you have above is a crit of a crippled AI. Sorry, but, kinda valueless.

[Solver]

Bright's on par economically, but not with optimal algorithms, indeed, Intelligent is that. I've played there on a pretty wide array of settings, though, including Normal, Bright and Intelligent.

And it's not all criticism either .

[dearmad]

crit is short for critique, in this case, not critisize.