2016 Winter Patch - Combat AI...?

[Disgustipated]

Hmm, not sure why America and I believe Saladin is beating me in tech and civics this game. I did conquer less than usual, but with the medium sized Earth map I had plenty of room to expand. But America does have as many cities as mine, so that may why.

America my friend, who I also had open borders with brought a whole bunchy of units into my territory. Unsurprisingly they attacked as soon as the declaration of friendship expired. They actually managed to pillage 2 of my districts, but posed no threat to my cities themselves. I lost a lot of turns rebuilding those districts. I didn't have walls up initially because it was an interior city. My mistake there.

I could if I wanted to declare on America and take several of his cities, but I am trying to keep friendly with everyone. I only attacked the Aztecs during the first age, and not since. If they declare on me again, I will be prepared this time since I have an army built this time. Before I only had archers/crossbowman, not enough to take cities with.

As for combat AI, looks the same to me. Taking out the Aztecs was very easy, I hit them before they had walls. The only way the AI will have any chance is if the devs program the AI to build ranged archer units and keep them in the cities and program them to actually attack. If they can program the AI to put one archer/crossbowman/field artillery/machine gun in each of their cities and keep them there, and fire upon any units in range, this will be a marked improvement. Of course they still wouldn't be able to take cities. My idea for this is to program the AI to just build lots and lots of units (cheat to give them discounts on unit maintenance) and zerg rush the closest city. This can't be that hard to program. Although maybe some civs like India and China should just keep the defensive stance and not zerg rush.

 

[bladex]

ai has gotten worse IMO now they barely expand at all...

 

[Cruor]

This is just sad... I was going to play now that 2 patches had come out but forget it. I am going to guess the game will be playable in by Fall 2017 maybe? After an Xpac, more patches and a lot of work from modders?

 

[PhilBowles]

Hmm, not sure why America and I believe Saladin is beating me in tech and civics this game. I did conquer less than usual, but with the medium sized Earth map I had plenty of room to expand. But America does have as many cities as mine, so that may why.

In my Sumer game I've been top in techs for a while and I'm having trouble understanding why, as the Germans have had a huge lead in science output for some time (I'm not sure how - we were level-pegging at first contact, they got ahead when they took Krakow, and then somehow went into overdrive with 230 points a turn to my 190 or so). My output is over twice that of any other civ, thanks to ziggurats, most of my captured cities having campuses, and my grabbing numerous Great Scientists (including Gallileo, who rushed me to Scientific Method while everyone else was at best entering the Renaissance).

I'm not sure why but no AI is focusing on GPs in this playthrough - with no Russia there's no run on cultural GPs in particular, but no one is even aiming for engineers or scientists. I was able to get both Newton and Galileo, who I consider the best early-mid game scientists. America is spamming merchants but I'm not sure what the end point of that strategy is - they can only send traders between their two surviving cities or to me. I'm winning the culture race with no good idea why (only in terms of culture output - other civs have more tourism).

I could if I wanted to declare on America and take several of his cities, but I am trying to keep friendly with everyone. I only attacked the Aztecs during the first age, and not since. If they declare on me again, I will be prepared this time since I have an army built this time. Before I only had archers/crossbowman, not enough to take cities with.

I was aggressive in the early game, razing Philadelphia and taking all of Brazil, then destroying Nagoya to get access to a better city spot slightly closer to Kilimanjaro. A sadly counterimmersive AI oddity was that, after agreeing to peace with Japan (I hadn't got any particular interest in their territory, just needed that city removed), Teddy Roosevelt immediately congratulated me on keeping peace on the continent I'd already started three wars on...

Everyone hates me except Gorgo, who's quite impressed at my ruthlessness but is never happier with me than neutral. But the AI is very passive in that regard - no one is attacking me, although from gossip notifications the AIs appear more aggressive among themselves and better at taking each other's cities (Kyoto was even retaken by the Japanese).

I rarely build more units and lost an impressive number in an abortive attack on Washington. I lacked niter so couldn't leverage my technological superiority and - although slow to respond - Roosevelt did so surprisingly capably, turning production from chariot spam to crossbows and actually - eventually - rebuilding the wall I destroyed, as well as making effective use of a pikeman to attack my weakened units (though ineffective use of it sitting in the water next to my catapult for a few turns). The AI still has a tendency to move newly-created ranged units out of the city, and if they're garrisoned inside they won't fire - maybe some bug preventing the AI from recognising that ranged units can fire from a city with walls?

Despite my unit shortage, I've been leading in the domination stakes (possibly bootstrapped by having the highest score), and possibly that's a deterrent to other civs - as it is only Spain, Japan and Norway have cities close enough to attack mine - the Spanish may be strong enough to compete with me, but they're engaged in an on-off war with Greece.

As for combat AI, looks the same to me. Taking out the Aztecs was very easy, I hit them before they had walls. The only way the AI will have any chance is if the devs program the AI to build ranged archer units and keep them in the cities and program them to actually attack. If they can program the AI to put one archer/crossbowman/field artillery/machine gun in each of their cities and keep them there, and fire upon any units in range, this will be a marked improvement.

I'm seeing unexpectedly capable behaviour when the AI has multiple units, including ranged, and certainly it's building ranged units and upgrading better than it did. But AIs just don't seem to produce units often or have cities specialised for production - maybe the slowdown to production times has hurt the AI in that regard because it reaches Apprenticeship late and doesn't prioritise industrial districts or encampments (none of the cities I've taken has had either).

ai has gotten worse IMO now they barely expand at all...

Not my experience - at the start of my current playthrough both of the civs I encountered first had a second city by turn 7 on Emperor - America expanded poorly after that but was almost confined to tundra by my own subsequent expansion. I took an aggressive approach early mainly because otherwise I'd have been cut off from good spots by my neighbours' fast expansion.

I'd taken or destroyed four Brazilian cities by turn 75. There are some odd gaps on the map and their settlement decisions seem off a lot of the time (a Spanish settler bypassing a productive rice-growing site, and Nagoya placed just too far away from Kilimanjaro to make use of it - perhaps because the AI is coded to avoid desert and didn't register that the desert tiles were producing 2 food thanks to the Natural Wonder), but overall the map's pretty densely-settled - my scout exploring the areas around Spain and Kongo has found both civs using all the available space in their heartland, and notifications periodically tell me that every civ (other than America, which can't due to its position) is expanding.

 

[PhilBowles]

I have lost a few high level games to science victories.
For culture they do not compete well but do try a little

At least pre-patch the AI's ability to compete for science victory is broken, since the AI lacks the ability to repair sabotaged spaceports and just builds new ones in different cities, which delays it too far. It can rush through the tech tree well enough to compete, but it can't win a science victory against a player who realises it's coming sufficiently far in advance to destroy the spaceports.

The AI also needs to make better use of its own spies - I've only ever seen them steal random tech boosts, often for techs the AI isn't in any position to obtain (such as Renaissance spies stealing my eureka for Electricity), when sabotage seems generally to be the most useful spy ability (not just for spaceports).

 

[Binkov]

But yeah, late game city conquest is still way too hard with the current gamerules. It's not really the walls, more that the tech scaling of city combat strength puts them enough above the melee units of the same era that they become useless. And since the AI manages melee units way better than siege weapons, it ends up failing to take anything.

So, one way to have a more challenging game would be to have city defense modifiers reduced? Is that possible without SDK? It would also make AI's cities easier to take for the player but perhaps overall effect, if AI finally deemed it worthwhile to go for enemy's cities, would still be a positive one?

But do we actually want an AI that tries to win? If we do, then things like diplomatic penalties for warmongering make no sense - but penalties for being the closest to winning do. If a "play-to-win" AI has decided to pursue a domination victory, it shouldn't matter how many gifts you've sent them or even if you've got an alliance - they might take your capital last if they like or fear you, but eventually those nukes are going to fly. Many players don't want that, they'd rather be able to make a game-long alliance with an AI player.

The way I see it, some players play for victory, and they would like AI to play for the win. Some players play for roleplay experience and flavor, and expect the same from the AI. Neither of the two groups wants a pushover opponent.

Since we have so many difficulty levels, in the ideal world immortal and diety levels would make AI go more for the win. Those players would are not so big on minmaxing and want roleplay can have all the fun they want on emperor and lower levels. Of course, i gather it would be next to impossible to mod two different AI logic types, dependant on difficulty level.

Nukes don't need to fly, nor total conquest needs to be done for anyone to win. Having one's production damaged enough so he's not a threat is enough to ensure victory.

Furthermore, playing roleplay and playing to win are not mutually exclusive. Some AI may have a trait that tells it to go for military conquest. Maybe just one of several within a game. But if it does have that set as its goal, then it should try its best to pull it off. And warmongering, ranged combat/movement/fire rules and city defense strength seem to be in the way of that being possible. So the most efficient way to deal with those would be to change the rules. At least until SDK is released.

 

[Manifold]

Siptah : CS are unable to take cities I think...

I have seen a CS how he conquer and raze a city with 1 population and without walls. It was not a capital.

 

[Siesta Guru]

So, one way to have a more challenging game would be to have city defense modifiers reduced? Is that possible without SDK? It would also make AI's cities easier to take for the player but perhaps overall effect, if AI finally deemed it worthwhile to go for enemy's cities, would still be a positive one?

Yep, that would definitely help. Especially if it also includes a decrease in city healing. There are a couple of changes possible to reduce combat strength. Sadly none that can affect the scaling of it proportionally. The main change in city combat strength is that it bases its strength on the highest melee combat strength of any unit you can build. On top of that, there's some increases based on walls, population, districts and the garisson. All except the scaling based on the unit can be changed, and you can also change the strength by a flat amount. But it's the scaling based on the highest melee unit that's the worst and you can't change that.
Like in renaissance, there's a time period where conquest gets nearly impossible for AIs, because their cities get massively buffed due to teching into musketman. No weapons are now adequate to take down cities until they get bombards or musketmen with siege towers. But since that takes hooking up niter, they end up being on the field so late (if they arrive at all), that people already get cavalry/infantry, buffing the cities right out of reach again.

 

[PhilBowles]

Thoughts following my first full post-patch playthrough:

- Tech/production scaling is much improved. My game went on to, I think, 1924 (turn 345) - at that point I had only started the first Future Tech and was researching the final (non-Social Media) civic. There's still too little worth producing in the late game beyond military units, but that's an issue for an expansion adding later-game content rather than a patch. The district structure simply makes build orders too linear, with each containing exactly the same four tiers that provide their resource in exactly the same way.

- Though the AI supposedly has coding to make it better at building forts, I never saw a fort.

- Naval invasions are disastrous. The AI can swim, but it won't use ships to escort its embarked units. Worse, if they're combat units it seems to perceive them as being able to fight - I've had embarked melee units swimming around my missile cruisers etc. instead of retreating.

- Leading AIs seem to be good at upgrading and moderate at producing a good number and mix of unit types, but for some reason (maybe only AIs coded to be aggressive do this?) weaker ones seem to just give up. Immediately before I won the space race America had crossbows, catapults and pikes massing near my borders - formed into corps, so presuming the Americans were reasonably close in techs to their civics, they should have had better units (they'd also had access to niter and to oil, though never developed the latter). The Aztecs were barely better - sending catapults along with their cavalry.

- I never saw any AI aircraft; given that I was being praised for my mighty airforce of one bomber, I presume there genuinely were none.

- I saw some moderately good efforts at city defence - still no actual garrisoned missile units, but artillery or missile units parked near enough to threaten my attackers and force me to call off attacks on Washington and Hamburg. I'm really starting to think AI tactical behaviour may be less at fault than the fact that AIs simply aren't producing enough units (somehow I kept being criticised for a small standing army while leading in domination score for much of the game), as they need more units than a human does in an equivalent situation.

- Diplomacy remains a weak point - though Gorgo ignoring warmonger penalties gave some sense of character, all of the other AIs basically behaved identically, differing only by virtue of relative strength. The two civs that competed with me seriously for victory - Germany and Kongo - both adopted the same basic approach (though Kongo was much less effective for some reason despite having a giant empire) and I ended up in a three-way space race. Everyone but Gorgo seemed to hate every other civ due to warmongering.

- Most civs declared war on others at some point and the wars were inconclusive - gratifyingly they did result in cities changing hands or being destroyed. I didn't see the old behaviour of AIs engaging in suicidal wars because they hated you for still owning their cities (in fact I was never even denounced for occupying anyone's cities, though I held American cities for most of the game). Suicidal wars still happened for no reason at all that I can discern - Harald declared war late, when he had a city adjacent to mine which was soon no longer his. Turns out longships and musketmen match up badly against mech inf, battleships and bombers. The even more backward Aztecs also declared war unprompted, but they were nowhere close so I just ignored them.

- I was getting a lot of notifications from civs that had denounced me demanding gifts, for no very clear reason and when they weren't in a position to go to war anyway. Speaking of which the denunciation system still needs to be fixed - having AIs auto-denounce every time a denunciation expires to prevent the player denouncing them leads to very uninteractive diplomacy, particularly since you can't usefully make any deals with them and buffs from favourable trades are so low that you can't improve relations unless you happen to align perfectly with their agendas.

- The AI used its spies better than it had previously - still (as in Civ V) mainly hitting the capital whether or not that was of strategic relevance, but it's as well for me that my capital wasn't my production hub as it was sabotaged repeatedly (my true production city only once). I didn't however knowingly run into counterespionage (although my spies did keep failing on 74% chances ti disrupt rocketry in Mbanza Kongo).

- Firaxis may be aware of the science victory 'exploit' and have taken steps to fix it. This is the fact that the AIs can't repair districts, so you can indefinitely delay completion of the space race by a rival by sabotaging all their spaceports. The fix I noticed is that a civ in the late stages of a science victory just spams spaceports - and still doesn't repair the ones that are destroyed. I came very close to losing after noticing Germany's progress too late - if one of my spy actions against Germany rather than Kongo (not having noticed before that there's a tooltip that tells you how far a civ has progressed when they've reached stage 3, I put too much effort into stopping Kongo reach the third stage when they were two techs short of ever completing the ship) had failed I'd almost certainly have lost. It wasn't helped by a less than obliging Jadwiga - Germany was on another continent and partially encircled by Poland, so without open borders with Jadwiga I couldn't even find all the German cities, including one with an active spaceport and two others that production could have shifted to. This fix - if such it is - is not ideal (if you ever manage to sabotage all a civ's spaceports and they have a spaceport in every productive city, they'll simply never be able to recover), but it beats the pre-patch system since it does improve the AI's odds of beating you in a close race.

- The AI seems somewhat able to switch civics to support a particular strategy - Germany's science shot up at one point and down again as soon as they'd completed the key techs - but they aren't using Great People or Wonders effectively. Very few Wonders were in demand in that game and those mostly an era behind - I had no trouble getting key wonders like Ruhr Valley and Oxford University, and although the patch suggested AI prioritisation of the Arsenal was improved it remained unbuilt at game's end (I hadn't needed it). I was beaten to Petra, which I wanted only because I had an ideal site but was late starting it, and the Oracle - which I ended up capturing, a move which may have been vital to my victory as I faith-bought a lot of GPs. I don't know if it's possible to make the AI recognise GPs' individual attributes and select them on that basis, but there was certainly a point where Germany would probably have had the resources to buy Carl Sagan and win on the spot - as it was I was able to get him and leap two and a half stages through the space race as soon as I built my spaceport. It may be dependent on the specific civs in a given game (and I destroyed Brazil early), as Russia has been part of most of my past games, but there was very little AI prioritisation of GPs of any type post-patch, where I've found getting the choice GPs a real fight in earlier games - no one ever had many engineer points and Germany had about 19 scientist points at its peak to 39 of mine. I wasn't trying for any cultural GPs but still seemed to end up getting most writers.

Overall I found myself finally getting into Civ VI - I'm not sure how much of this is attributable to patch changes, as I hadn't played it extensively before this and had only played two prior full games.

 

[bladex]

Not my experience - at the start of my current playthrough both of the civs I encountered first had a second city by turn 7 on Emperor

yeah well not everyone plays on Emperor

 

[PhilBowles]

yeah well not everyone plays on Emperor

If the AI isn't expanding fast enough to be a challenge you should probably play at a higher level, in that case (I presume it's not a level higher than Emperor where they aren't expanding). Expansion behaviour shouldn't be affected by difficulty, other than the fact that the AI gets production boosts (and at the highest difficulties, extra settlers) that let it settle more quickly.

Civ VI is easier than Civ V, but from my most recent experience not by nearly as much following the patches as it was at release.

 

[DizzKneeLand33]

I don't understand the second city by turn 7 on Emperor. There isn't supposed to be a bonus settler on Emperor, is there? If so, that would be nice to know, if not, then they had to pop one out of a hut or something -- it's not just better game mechanics. Or am I missing something?

 

[PhilBowles]

I don't understand the second city by turn 7 on Emperor. There isn't supposed to be a bonus settler on Emperor, is there? If so, that would be nice to know, if not, then they had to pop one out of a hut or something -- it's not just better game mechanics. Or am I missing something?

I was surprised as well, especially as there were two civs in that situation (though one could have been a little later) - even if the AI could get settlers from huts that seems particularly unusual. If they get bonus pop in their first city it's not impossible if they build a settler before anything else, though.

 

[nokmirt]

This thread does not make me optimistic about playing again.

 

[spfun]

I don't understand the second city by turn 7 on Emperor. There isn't supposed to be a bonus settler on Emperor, is there? If so, that would be nice to know, if not, then they had to pop one out of a hut or something -- it's not just better game mechanics. Or am I missing something?

On Emperor and Immortal they get an extra settler, Deity they get 2 extra settlers. I hate it and hope they can improve the AI enough so they can remove it from Emperor/Immortal.

 

[V. Soma]

The main change in city combat strength is that it bases its strength on the highest melee combat strength of any unit you can build.

I find that weird. The city bombards - shouldn't it be based on a ranged unit...?

 

[Jabulani]

On Emperor the AI with 3-4 mediocre cities placed in a bad locations and with few units of an ancient army can reach 700-800% more science than what it would be allowed, that insane amount of science comes just from cheating bonuses.

I wonder why Ed Beach and his boy band have done that, Sid would turn in his grave if he knew what is going on.

 

[isau]

I suppose on the plus side, I modded the game so that borders expand faster and the AI is way more competitive. Constantly grabbing spots I want and their border intruding... grr. Not that they defend their spaces well when I invade. But it's something.

The core of the game is great. C'mon Firaxis... fix a few game rules (rate of border expansion! encampent unlocks! value of sea tiles! movement!) and you can correct a lot without even touching the AI. Even if AI still needs work. A LOT of the issue is just how the game rules themselves need minor tweaks.

 

[player1 fanatic]

yeah well not everyone plays on Emperor

And they should have city in up to 7th turn. They start with extra settler for the reason.

 

[Siesta Guru]

I find that weird. The city bombards - shouldn't it be based on a ranged unit...?

Honestly not sure if the city ranged damage strike is based on range/melee, I just know that it's defensive combat strength is based on melee units. It took me quite a while to figure that out too. Which is the main problem, a walled city with high combat strength takes basically 0 damage from any non-siege weapons.
It may actually also be affected by naval units,



For those of you who hate the non-aggressiveness of the AI, you may want to try out v9 of AI+, it's getting quite a bit better already. There's still no fix for district repairing etc though, there doesn't seem to be a way for me to impact that.