Sept 22 AI Discussion

[isau]

Before live streams existed games in development had tons of small issues the public never saw. Previous civ games have had various issues, this one wasn't one of them for any significant length of time, it's fair to say they know how to fix it. And yes in the scheme of "things that could be horribly broken" AI being ginger about upgrading units is hardly the end of the world. That's a fairly fixable leg of code and no doubt they have a very good idea of what it would take to make it more aggressive.

 

[SupremacyKing2]

I remember back when the new Master of Orion was in early access that it had a similar issue. The AI would never build ships bigger than a frigate (the smallest ship in the game). 100 turns in, the human player would have larger ships like cruisers but the AI would still be spamming frigates. Of course, this made combat too easy for the human player since the AI's frigates could not compete with the human player's more powerful ships. The devs discovered the culprit: the code that determined when it was more cost effective to build smaller ships fast or taker longer to build stronger ships was too strict. So the AI was always building the smaller ships because they too less time to build. As soon as the devs relaxed the calculation, the AI was able to build the bigger ships. Something similar might be going on here. maybe the code is telling the AI that it is better to build these warriors because they can be built super fast in just a couple of turns? Or perhaps, the code for spending gold is prioritizing other things so the AI never has enough gold to upgrade their units? In any case, it should be easy to fix IMO.

 

[CaiusDrewart]

Before live streams existed games in development had tons of small issues the public never saw. Previous civ games have had various issues, this one wasn't one of them for any significant length of time, it's fair to say they know how to fix it. And yes in the scheme of "things that could be horribly broken" AI being ginger about upgrading units is hardly the end of the world. That's a fairly fixable leg of code and no doubt they have a very good idea of what it would take to make it more aggressive.

Upon closer analysis I don't think the problem is that the AI isn't upgrading units, it's that the human player is literally 20+ techs ahead. Someone in the screenshots thread has a shot of Kongo at 51 techs while England is at 30 (and in second place!)

So it's not that the AI isn't upgrading, it's that the AI isn't teching. And it's Prince (right?), so I won't worry too much about that.

 

[Human Crouton]

Maybe this is how difficult they want prince to be. Maybe they want new players to have fun, casual, lengthy games instead of throwing them off a cliff.

 

[malekithau]

I've played plenty of games in previous iterations of Civ at Prince level where I have severely outclassed some of the AI opponents but not to this extent. Lower levels yes so perhaps this was actually a much lower level game?

 

[CaiusDrewart]

I've played plenty of games in previous iterations of Civ at Prince level where I have severely outclassed some of the AI opponents but not to this extent. Lower levels yes so perhaps this was actually a much lower level game?

In Civ IV, a decent player (and let's assume that Pete at Firaxis is decent at civ) could blow past the AI and have colossal mismatches like infantry vs. longbows. I barely spent any time playing Civ V on levels below Emperor so I'm not sure if that's possible in that game or not. But in Civ IV the high levels are still fun and challenging despite how easy Prince is.

 

[Lucius_]

Prince isn't the bottom difficulty though, but the 3rd up. When I first started playing civ, I think I played one game at chieftain then moved up to warlord. It was a pretty long time before I moved up to Prince, then King. King was too hard so I played at Prince level. It wasn't until Civ 5 that I moved into higher difficulties, which again, took some progressing. Moved to Prince pretty quickly but struggled to move up to King and struggled again to get to Immortal.

Looking at this, I don't think my pre-CiV self would have had to do any progressing to get to Prince level. What's the point if the middle difficulty is face-palm easy? That means the lower two difficulties are drooling mouth-breather level with chieftain being literal face roll easy.

 

[cakes]

i think the AI just has too much to do and can't prioritise correctly or something. maybe in higher difficulties with production bonuses etc the AI should be able to handle the workload a little better and will look like a 'proper empire'. i am a little concerned but the game still works even if it has ciV-level competence.

what i am confused by is a lack of options for diplomatic cards. since barbarossa was present in the rome game there was a disincentive to buddy up with city states, but as far as choices for diplomatic cards went it was either a bonus to envoys or a bonus to envoys. do i just switch governments so i don't have to deal with it (and lose my legacy bonus if that is a thing that happens)? do i just risk angering barbarossa because CS bonuses are too good to ignore?

 

[Ashbery76]

You guys never seem to learn.Devs only work properly on the A.I in the last 4 weeks before release.Good A.I is possible in games as their end cycle A.I are miles better than release.If just requires time.The sad fact is good A.I is not a selling point which is sad as 99% play single player.

 

[isau]

If the problem is science and culture progress it's possible the AI isn't hitting Eurekas. I have a strong feeling the AI is just going to have to cheat a bit there. Or we could settle for AI that just gets a big discount on techs without any need to go for Eurekas. I know some people would be really upset about it not being "fair" but I frankly don't care anymore so than I care about unfair it is that when I hit level 100 in an RPG I encounter creatures that are level 100 but haven't had to farm for it. As long as AI can present a fun obstacle I'm down for it.

 

[CaiusDrewart]

You guys never seem to learn.Devs only work properly on the A.I in the last 4 weeks before release.Good A.I is possible in games as their end cycle A.I are miles better than release.If just requires time.The sad fact is good A.I is not a selling point which is sad as 99% play single player.

I used to think the argument that AI isn't a selling point was crazy. It makes such a huge difference in how much I enjoy the games! Eventually bad combat AI made Civ V unplayable for me. I just didn't find it fun to watch the AI helplessly attempt to maneuver its units around the map, never accomplishing anything. Outmaneuvering and picking off their units offered no satisfaction. That game would have been infinitely better with a good tactical AI.

But then I saw the statistics of what difficulties Civ V players play on, and I realized that like 95% of Civ players never attempt challenging difficulties. AI can hardly be of concern to these people. Then you understand why the designers spend so much time on graphics and presentation (which affects 100% of players) and so little on AI (which affects 5%). It makes sense. Even though I personally would revert to Civ III graphics in exchange for a competent AI in a heartbeat.

 

[CaiusDrewart]

If the problem is science and culture progress it's possible the AI isn't hitting Eurekas. I have a strong feeling the AI is just going to have to cheat a bit there. Or we could settle for AI that just gets a big discount on techs without any need to go for Eurekas. I know some people would be really upset about it not being "fair" but I frankly don't care anymore so than I care about unfair it is that when I hit level 100 in an RPG I encounter creatures that are level 100 but haven't had to farm for it. As long as AI can present a fun obstacle I'm down for it.

My guess is that the AI won't get any special help with Eurekas (and will therefore really, really struggle to get them at the right times), but will get the customary huge discounts on techs. They could also program the AI to always (or almost always) research techs it happens to have boosted in favor of ones it does not. That might result in some pretty random tech paths, but would probably still be more efficient and effective.

Edit: sorry for the double post, meant to combine these--my mistake.

 

[Acken]

In Civ IV, a decent player (and let's assume that Pete at Firaxis is decent at civ) could blow past the AI and have colossal mismatches like infantry vs. longbows. I barely spent any time playing Civ V on levels below Emperor so I'm not sure if that's possible in that game or not. But in Civ IV the high levels are still fun and challenging despite how easy Prince is.

By 1610AD in Civ5 I've already launched into space.

What we've seen of civ6 prince AI so far make it somewhat on par with civ5.

 

[Lucius_]

If the problem is science and culture progress it's possible the AI isn't hitting Eurekas. I have a strong feeling the AI is just going to have to cheat a bit there. Or we could settle for AI that just gets a big discount on techs without any need to go for Eurekas. I know some people would be really upset about it not being "fair" but I frankly don't care anymore so than I care about unfair it is that when I hit level 100 in an RPG I encounter creatures that are level 100 but haven't had to farm for it. As long as AI can present a fun obstacle I'm down for it.

I think some cheating is acceptable, while other AI bonuses are not. I'm fine with the AI getting some Science and Culture bonuses that help them stay on par in higher difficulties. Regarding CiV though, I hated production and happiness bonuses. Production made wonders ridiculous to get unless you have a decent tech lead and it made them able to churn out units at feverish pace, making warring some AI's very tedious.

Happiness, well, ever seen the ideology policy "+33% tourism to civs with less happiness than you"? Your ideology is dominant, they have Civil Resistance and yet are still at 80 positive happiness! In other words, ideological pressure does not apply to the AI. That policy is 100% worthless in single player. If a civ wants, it can found a dozen cities, and never worry about happiness, which the player cannot ever hope to do, at least without happiness religion beliefs and ideologies late in the game.

Those very noticeable bonuses to the AI, irk the hell out of me and entire game systems being ignored by the AI just makes me rage.

 

[Acken]

I think some cheating is acceptable, while other AI bonuses are not. I'm fine with the AI getting some Science and Culture bonuses that help them stay on par in higher difficulties. Regarding CiV though, I hated production and happiness bonuses. Production made wonders ridiculous to get unless you have a decent tech lead and it made them able to churn out units at feverish pace, making warring some AI's very tedious.

The AI gets very little bonus for producting wonders.

 

[isau]

I think some cheating is acceptable, while other AI bonuses are not. I'm fine with the AI getting some Science and Culture bonuses that help them stay on par in higher difficulties. Regarding CiV though, I hated production and happiness bonuses. Production made wonders ridiculous to get unless you have a decent tech lead and it made them able to churn out units at feverish pace, making warring some AI's very tedious.

Happiness, well, ever seen the ideology policy "+33% tourism to civs with less happiness than you"? Your ideology is dominant, they have Civil Resistance and yet are still at 80 positive happiness! In other words, ideological pressure does not apply to the AI. That policy is 100% worthless in single player. If a civ wants, it can found a dozen cities, and never worry about happiness, which the player cannot ever hope to do, at least without happiness religion beliefs and ideologies late in the game.

Those very noticeable bonuses to the AI, irk the hell out of me and entire game systems being ignored by the AI just makes me rage.

I agree with you on some of this, mainly the AI getting free Happiness because that made it kind of impossible to harm the AI in interesting ways.

The Eurekas though I'm not so sure. I wouldn't mind if AI just paid 35% less for techs all the time and if they scored a Eureka they got the final 15%. To give a small reward if you somehow thwart them but not totally hamper them. I just don't ever see the AI being smart enough to be like "I need to kill a unit with an Archer specifically to get a Eureka." That's almost surely player-only skill levels. And I don't think it's worth the program cycles to try to teach those kinds of things.

 

[dturtle1]

Upon closer analysis I don't think the problem is that the AI isn't upgrading units, it's that the human player is literally 20+ techs ahead. Someone in the screenshots thread has a shot of Kongo at 51 techs while England is at 30 (and in second place!)

So it's not that the AI isn't upgrading, it's that the AI isn't teching. And it's Prince (right?), so I won't worry too much about that.

Sounds like Prince... My current game i have a 20+ lead in techs as well, though this later than 1610, however i dont exactly rush science on prince . I used to play alot of Prince, at first i convinced myself that it was the "fair" difficulty, I realised though it was only because i wanted all the Wonders I now Play King, Emperor depending on the type of game i want to play. Maybe i should try an immortal or higher difficulty....never tried

Good players should be able do that on prince no worries. I am not exactly a Hardcore Civer. (I think the highest i have ever completed was Emperor) but even i absolutely roll-stomp all over prince. I am sure they will tweak the A.I and difficulty settings as we approach launch but from everything i have seen, it does look comparable to Civ5, maybe they might need some help with Eureka's or an increase in the flat boost to make up for poor Eureka skills.

 

[homan1983]

Everyone is saying the problem is the AI is far ahead, but what is more worriesome is the Rome vs Greece game where the greek player is ahead in points and cities (4:3) but simply doesnt upgrade the warriors.

One problem may be that the AI built lots of warriors since it was ahead. But because it had so many units it decided against building its unique unit. And then a lack of iron meant it couldn't upgrade the units.

 

[stealth_nsk]

Everyone is saying the problem is the AI is far ahead, but what is more worriesome is the Rome vs Greece game where the greek player is ahead in points and cities (4:3) but simply doesnt upgrade the warriors.

Upgrades are expensive in the game and we don't know if Greeks had access to Iron.

 

[ElMasBandido]

At the start of the Kongo video there is a notification that Norway has just built Stonehenge.

1610 AD. Turn 233. One of the first available wonders is just being constructed by AI...





No one knows who they were or what they were doing
But their legacy (of awful AI) remains
Hewn into the the living rock of Stonehenge