How is the AI handling the 1.2.5 patch?

[stealth_nsk]

That's a great mental model for thinking about it, thanks!

I'll be honest -- the game AI is the one area I really haven't delved into. I've had to deal enough with AI over the last couple of years to last a lifetime (tl;dr: vector searches are amazing, but the generative side owes its success to good answer templates more than anything). But the way you framed the game AI makes it much easier to break it down into smaller pieces and has me curious to look into it further. Thanks!

Yes. Important thing is for AI to appear playing the game. So things like building things or moving units has to be simulated well, while thinks which aren't visible, like policy cards, don't need to be simulated - AI only needs some believable adjustments to yields, since yields are visible to player.

For the same logic Civ1 had AI build wonders instantly in specific moments and nobody noticed, because before unstacking cities the building queue was not visible.

 

[mielony]

So far, I have played one match. The first two eras were interesting gameplay. Lots of wars, it was fun to play. But the last era was just clicking through 50 turns. I was supposed to win economically, but the game was so boring that I had to switch to a cultural victory. The AI is bad at pursuing victory. With 4 planes, I destroyed an entire civilization.I’ll wait now for the AI mod and play a few games on different maps. If nothing changes with the AI behavior, I’ll take a break from CIV7.

 

[TomSpren]

I'm not sure if this is due to the latest update because I've finally been playing games to the end, the last couple of months, but I find that as long as you keep a big enough military, the AI just lets you run full steam towards your victory condition.

 

[Krikkit1]

I'm not sure if this is due to the latest update because I've finally been playing games to the end, the last couple of months, but I find that as long as you keep a big enough military, the AI just lets you run full steam towards your victory condition.

Well a big military should deter the AI from militarily interfering with your victory condition.... However, they should be Diplomatically (and in other ways) working against it.

A few things
1. the "anti-Legacy/anti-Victory" Espionage options should be introduced earlier and expanded
2. the Victory conditions themselves (not just the Legacies) should be made more competitive....ie perhaps there can only be one civ's world bank office in a settlement and your Banker can replace someone else's (and they need to be in every Settlement with a Factory)
3. Finally make the AI actually pursue the Victory Conditions.

 

[Boris Gudenuf]

2. the Victory conditions themselves (not just the Legacies) should be made more competitive....ie perhaps there can only be one civ's world bank office in a settlement and your Banker can replace someone else's (and they need to be in every Settlement with a Factory)

I think this is the big one.

Right now, there is simply very little competition inherent in any Legacy or Victory Path.
Nothing can stop or even slow down a Missionary or Explorer except to get there first with more of them (both paths, by the way, severely punish Tall play) and once Relics or Codexes are gathered, short of completely conquering every city and settlement in the Civ there's no way to rid your opponent of them.

Similar conditions apply to getting tiles with high numbers for Exploration Science or building Factories and railroads - short of military conquest, it's practically impossible to even slow them down, let alone stop them.

You can raid Treasure Fleets, but in my experience, the map will limit the actual piling up of Treasure Points and resources more than any AI entity or player can.

Basically, with the exception of the Military paths and victory, I have spent too many games utterly ignoring the AI opponents and merrily toodling along all the Economic, Science, and Cultural legacy and victory paths secure in the knowledge that short of a Planetary Alliance of military powers against me there is Nothing They Can Do to stop me - and, of course, nothing much I can do to them, either.

It is probably the major reason I stopped playing Modern Age a couple of months ago: the entire Age became a rote romp to victory regardless of anything that could happen in the game or anything the AI did - too much like trying to play a game of chess after someone achieves Checkmate and an equal waste of time.

 

[Krikkit1]

I think this is the big one.

Right now, there is simply very little competition inherent in any Legacy or Victory Path.
Nothing can stop or even slow down a Missionary or Explorer except to get there first with more of them (both paths, by the way, severely punish Tall play) and once Relics or Codexes are gathered, short of completely conquering every city and settlement in the Civ there's no way to rid your opponent of them.

Similar conditions apply to getting tiles with high numbers for Exploration Science or building Factories and railroads - short of military conquest, it's practically impossible to even slow them down, let alone stop them.

You can raid Treasure Fleets, but in my experience, the map will limit the actual piling up of Treasure Points and resources more than any AI entity or player can.

Basically, with the exception of the Military paths and victory, I have spent too many games utterly ignoring the AI opponents and merrily toodling along all the Economic, Science, and Cultural legacy and victory paths secure in the knowledge that short of a Planetary Alliance of military powers against me there is Nothing They Can Do to stop me - and, of course, nothing much I can do to them, either.

It is probably the major reason I stopped playing Modern Age a couple of months ago: the entire Age became a rote romp to victory regardless of anything that could happen in the game or anything the AI did - too much like trying to play a game of chess after someone achieves Checkmate and an equal waste of time.

I think the Legacies don't need to be As competitive (maybe a tiny bit more for some than they are..say get an occasional bonus Codex for being the first to a Mastery). But the Victories Absolutely should be.

 

[UWHabs]

I think it's hard to make the victories too competitive without turning things into a grind.

I wonder, though, if maybe the answer is more about giving you some flexibility in claiming victory, and then basically having a greater fight for them. Like, I know they set up the victories so that you have this grand final state, so that you know it was your action that ended the game. But ending on a wonder finishing isn't really any more exciting than finishing when some other counter finishes.

My suggestion would be to rework all the victories so that at some point, the victories would unlock (maybe where they are now), but when you declare a victory, the game starts a 10 (20?) turn counter, and you basically have another challenge to complete in that time. So the econ victory maybe triggers when the Banker finishes, but what happens is that from that point on, you have to get like another 200 RR Tycoon points. But the caveat is that every tycoon point the AI puts out counts against yours. So basically you have to set yourself up so that you know you can win, but if you fail the challenge, then it doesn't count as a win on your record.

Similarly, the military victory could be that once you declare, you have 20 turns to get another 20 or whatever military points, but you get a big negative if you lose a settlement. Maybe also the AI would get like a free unit for every urban population in each of their settlements or something, or it would unlock the ability to draft units from your cities. So again, you could declare as soon as you finished project ivy, but you have to be ready to continue from that point and fight for the victory.
Space you could either bring back the lasers from 6, or maybe the spaceship component system from back in civ 2 days, was it? Something that maybe if you launch early there's a chance of failure, or maybe the AI gets like a massive boost to their space program when you declare.
Culture you might have to change it around more, since at some point the relic system would basically run out for you. But maybe the culture victory unlocks a unique culture tree where you get like a repeatable civic to give you a special relic, but the AI could also get them to remove one from your inventory.

You could probably have the victories unlock at the current place when you trigger the victory, and then fold in the special bonus piece into that. So whenever you declare the econ victory, you get the Great Banker, and for each civ that he goes to, it gives you 50 points towards the tycoon victory. Maybe Project Ivy isn't the end cap, but basically it's worth like +10 points to that victory. etc...

I dunno, something so that you have an actual race against the AI, and it's not just about who gets there first, but about getting there first while also being comfortable that given some bonuses, nobody else will catch you.

 

[vorlon_mi]

Well a big military should deter the AI from militarily interfering with your victory condition.... However, they should be Diplomatically (and in other ways) working against it.

A few things
1. the "anti-Legacy/anti-Victory" Espionage options should be introduced earlier and expanded
2. the Victory conditions themselves (not just the Legacies) should be made more competitive....ie perhaps there can only be one civ's world bank office in a settlement and your Banker can replace someone else's (and they need to be in every Settlement with a Factory)
3. Finally make the AI actually pursue the Victory Conditions.

With respect to 3), I have experience with this, which I described over in https://forums.civfanatics.com/thre...atch-1-2-4-ai-and-sandbox-reflections.700041/
At least 1 AI player pursues RR tycoon victory in each of the last 3 games I've finished. Sometimes they're behind me; this time, Harriet beat me.

With respect to 2), I agree.
I suspected that she was sending her Banker around, but I couldn't see it. I couldn't see the evidence that it had visited the cities where I was sending my Banker. Being able to overwrite/replace would be cool; having a victory tracker like we had in Civ6 would also be helpful. In that game, no one was close to accumulating the artifacts for Cultural victory. It would still be helpful to know if anyone has started the World's Fair. We can tell how many players have completed which Space projects. We can tell if an AI player is accumulating military points as well.

 

[timmychompchomp]

In my experience, for whatever reason, I have crushed the AI in my 2 games post patch. The AI seems more passive and less capable from a military perspective. It is quite possible that I have improved my play given that the updated UI allows for better city development. I think asymmetry in the map might be helping me dominate as well, meaning I got lucky with my geography and resources. I never reroll.

 

[TomSpren]

In my experience, for whatever reason, I have crushed the AI in my 2 games post patch. The AI seems more passive and less capable from a military perspective. It is quite possible that I have improved my play given that the updated UI allows for better city development. I think asymmetry in the map might be helping me dominate as well, meaning I got lucky with my geography and resources. I never reroll.

I'm having the same experience as you are.

 

[Leucarum]

I think there are definitely more gaps between when it comes to yields the leader who are haves (e.g. Franklin and Confucius) and the have-nots (e.g. Bolivar and Napoleon) in this patch. Firaxis intentionally didn't adjust leaders with yeild bonuses but there is definitely a pretty wide gulf among the AI's performance when you have a leader that has significant boosts to certain yeilds with the changes.

 

[ggalindo001]

I had originally voted "worse".

I realized that I had been playing with the RHQ Artificially Intelligent mod coming up into 1.2.5 --- so the base AI is "worse" than that.

However, likely better than previous vanilla releases. Which is an improvement.

I will say that I am finding the AI on Diety a little more aggressive in declaring war and in rushing units to the battle front. However, use of commanders and proper tactics is still very much lacking.

 

[MutilationWave]

I had originally voted "worse".

I realized that I had been playing with the RHQ Artificially Intelligent mod coming up into 1.2.5 --- so the base AI is "worse" than that.

However, likely better than previous vanilla releases. Which is an improvement.

I will say that I am finding the AI on Diety a little more aggressive in declaring war and in rushing units to the battle front. However, use of commanders and proper tactics is still very much lacking.

In my last game another civ rolled a commander up onto the island where I had two settlements and an IP I was befriending. To my surprise it popped out four units and began killing IP units. I attempted to block for them but was unsuccessful. The AI killed all their units and sat a unit on the IP center. I was very impressed. Then the units stood around, never dispersed the IP. When I suzerained it, they got kicked out of the borders and left.

 

[ggalindo001]

This is the stuff that just irritates me -- the AI just doesn't follow-through. to compare with Old World, where the AI becomes pretty ruthless.

 

[Boris Gudenuf]

This is the stuff that just irritates me -- the AI just doesn't follow-through. to compare with Old World, where the AI becomes pretty ruthless.

The problem is not that the Civ VII AI isn't ruthless, but that it is hapless.

More specifically, it has not been programmed to play the complete game.

In this case, as a good example, I have seen the AI take settlements and raze them from the AI Civs (and my own Civ, on several occasions!), so the AI 'understands' the process of taking and eliminating Civ settlements. From the example in this thread, though, it has not been programmed for the difference in razing/eliminating an IP settlement, which does not use the same process.

Multiple this little difference by dozens, and the result is a hapless AI that does many things almost well, but not well enough to compete and at times maddeningly incomplete.

 

[uppi]

The problem is not that the Civ VII AI isn't ruthless, but that it is hapless.

More specifically, it has not been programmed to play the complete game.

In this case, as a good example, I have seen the AI take settlements and raze them from the AI Civs (and my own Civ, on several occasions!), so the AI 'understands' the process of taking and eliminating Civ settlements. From the example in this thread, though, it has not been programmed for the difference in razing/eliminating an IP settlement, which does not use the same process.

Multiple this little difference by dozens, and the result is a hapless AI that does many things almost well, but not well enough to compete and at times maddeningly incomplete.

I have seen enough examples of the AI dispersing independent powers that I can confidently state that it has been programmed to do it. It just has recently become more hesitant to actually do it. I suspect that they tend to not disperse when someone is trying to befriend it and they like that player and don't want the relationship to deteriorate. Or it might just be my rationalization and it is just a bug. Problem is, the AI cannot explain itself, so sometimes it is hard to say whether this is working as intended.

 

[vorlon_mi]

In my experience, for whatever reason, I have crushed the AI in my 2 games post patch. The AI seems more passive and less capable from a military perspective. It is quite possible that I have improved my play given that the updated UI allows for better city development. I think asymmetry in the map might be helping me dominate as well, meaning I got lucky with my geography and resources. I never reroll.

I'm finishing Antiquity with Hattie, one of my first 2 games post patch. All 4 Homeland opponents have declared on me; yes, I started it by taking one of Ben Franklin's somewhat-forward settled town on the coast. Ben and Himiko allied up, as did Amina and Friedrich. They are actively attacking, though not particularly *effectively* attacking. Himiko even went from Friendly to Hostile in one turn. What did I do to make her so mad?

Need to go read the thread on Hattie as a leader. I'm struggling to keep up in science; I lost the race for wonders 3 times.

 

[Evolena]

yes, I started it by taking one of Ben Franklin's somewhat-forward settled town on the coast. Ben and Himiko allied up, [..]. Himiko even went from Friendly to Hostile in one turn. What did I do to make her so mad?

She had the choice to end the alliance or join her ally in the war, and chose the latter.
Or didn't she join the war?

 

[Lebo44]

Then the units stood around, never dispersed the IP. When I suzerained it, they got kicked out of the borders and left.

AI often exhibits strange behavior due to numerous minor bugs. One recurring issue is its tendency to remain inactive for few to several turns, as if some hidden condition temporarily prevents it from acting. This pattern was especially noticeable during Notque's streams, when he worked on improving/debugging the AI. I think it might acting differently if human player is around. If its not around, the AI doesn't sense 'danger' and tends to freeze.

 

[Boris Gudenuf]

I have seen enough examples of the AI dispersing independent powers that I can confidently state that it has been programmed to do it. It just has recently become more hesitant to actually do it. I suspect that they tend to not disperse when someone is trying to befriend it and they like that player and don't want the relationship to deteriorate. Or it might just be my rationalization and it is just a bug. Problem is, the AI cannot explain itself, so sometimes it is hard to say whether this is working as intended.

You are quite right, and I overstated the case.
On the other hand, you are also quite right that the AI's behavior on many occasions and in many game instances defies logic or strategic analysis.

While I have also seen numerous IPs 'disappear' from the map, especially in Antiquity and Exploration, which IPs disappear is a mystery.

Hostile IPs right next to an AI Civ are still harassing that Civ all through the Age, while other hostile IPs some distance away are neutralized by an expedition or made friendly with Influence.

A hostile IP that wipes out a settlement is not counterattacked, but instead that entire direction is abandoned - possibly a strategic decision, but not consistent because on other occasions the AI Civ throws armies at the IP despite having a major war with another Civ going on another front.

And in its use of Leaders and deployment of armies the AI is, at best, mediocre. Very, very rarely does it use Leaders to bring up entire armies at once (I never saw that before the last patch, have seen it exactly 3 times in a dozen games since) while usually the Leader is wandering around alone like some very expensive Scout, and with about the same life expectancy. The AI does not consistently use ranged units - although it does build plenty of them, unless it has a Unique unit of another type, in which case everything seems to be subordinate to spamming the Unique. Rome and Greece are grand examples, fielding waves of Legions or Hoplites with almost no ranged units at all. Regardless of how good the individual unit is, this kind of army is easy meat for a mixed force of adequate melee infantry backed up by good ranged firepower.

The AI does prioritize cavalry in Exploration, even Leaders without any cavalry bonuses attached field armies almost entirely on horseback - but the AI does not seem to understand combined arms at all, even as defined by the game, and this makes all of its forces relatively easy to defeat by any human player who is paying attention.

The result, unfortunately, is and has been for over a decade a game in which combat is, after you learn the basics, pretty much mindless and not much fun at all. Refighting Little Big Horn as the Lakota is only fun a couple of times before you look at the starting situation and wonder "why bother?"