TotalBiscuit thinks the AI is good. and i'm now willing to believe it...

[kaspergm]

While I don't know whether AI is actually good or bad, I do agree that many of the problems we've seen boils down to balance issues rather than necessarily AI issues. Two issues stand out:

• AI builds too few military units, and does not upgrade their military.
• City defense is too low. City ranged attack does a bit (but not a lot) too little damage in early eras.

I do appreciate the fact that city bombard no longer insta-kills units, and I also think it's fine that a melee unit can now attack a city without loosing 90 % of it's health (thus effectively comitting suicide). I also think the new siege mechanism that prevents cities from healing when surrounded by ZOC is great. However, these changes also means that cities need to have a much larger HP capacity, being able to stand attacks for at least 4-5 turns when not faced with overwhelming odds, and that's not what we've seen. Rather, it seems like even reasonably well-defended cities fall and 2-3 turns against rather small invasion forces, and obviously that creates immense problems for the AI.

 

[stealth_nsk]

City defense is too low. City ranged attack does a bit (but not a lot) too little damage in early eras

I believe you're the only one complaining about this. Too strong cities is generally viewed as one of the biggest problems in Civ5.

 

[Josephias]

I believe you're the only one complaining about this. Too strong cities is generally viewed as one of the biggest problems in Civ5.

He has a point though. He is not denying cities were too strong in Civ 5. (On the attack, mainly). And he is asking for only one strenght to be increased: HP. This can be seen also as historically logical: city conquest shouldn't be an army drain, but at least should slow your attack progression.

 

[sonicmyst]

what if the AI gets free military upgrade to their current troops? Or, say, 50% of all purchased troops gets a free upgrade. I know this is some form of cheating, but I cannot think of any way for AI to catch up with human player.

I also remember in Warcraft how AI builds their troops. Warcraft AI always check what the human player is doing and builds troops that effectively counter it. I think it can also be applied in Civ wherein if a human player builds more cavalry units, the AI will just build more anti-cavalry units.

Tech progression of AI players should also have a flat bonus, maybe 30% reduction in cost? That way, AI players will not be burdened in chasing the mini-quests just to boost the tech for a Eureka.

 

[stealth_nsk]

He has a point though. He is not denying cities were too strong in Civ 5. (On the attack, mainly). And he is asking for only one strenght to be increased: HP. This can be seen also as historically logical: city conquest shouldn't be an army drain, but at least should slow your attack progression.

Well, if city has either walls or defending unit, it could defend quite well in Civ6. If the city has neither, I don't see why it should be a significant obstacle for a strong army.

Of course it also depends on how well AI could manage defenses, but patching holes in AI logic at the expense of gameplay doesn't look like a good practice.

 

[Camikaze]

My recollection of Civ5 is that the first major balance patch made the game about half a difficulty harder. I would anticipate something similar for Civ6. At launch difficulty will probably be at a baseline from which it will only rise.

 

[Mustakrakish]

To OP:

While I agree that it's mostly about balance (I don't have any complains about tactical AI), AI priorities etc...

Saying those big ones make AI look bad because they're so good is just wrong IMO. They haven't played THAT good at all. They haven't done anything that wasn't obvious etc. I'm pretty sure I would play the same if not better (because I'm more familiar with the game than them it seems) and I'm no deity player, not even immortal... King is my V difficulty. So pretty much an average player.

 

[Xur]

I saw video where the AI had around 25 apostles and basicly no military units, so I think it's a matter of balancing priorities. At the moment it seems like the AI goes 100% towards one singular goal without much thought to what is happening in the game... maybe I'm wrong, but I have seen plenty of stupid strategic AI mixed with acceptable tactical AI. The one thing I REALLY hate is the forward settling AI that goes to the other end of the world to settle next to you - that's beyond stupid.

 

[jekke]

I suppose it is a bit unfair to ask a topicstarter to explain stuff which is only copied from somewhere, but since there are people here who agree with the message I will nonetheless do it: I kind of understand what you guys mean by 'balance', but what else is AI other than balance, that you are happy with? AI knows how to build units, to move them around and to construct districts in cities? These are all technical stuff as in it just requires corresponding procedures to be coded. AI always knew how to build units in Civ V when Mongols rushed your city with 8 trebuchets and one warrior in renaissance. AI always knew how to propagate religion too, when Haile showed up at your borders with 3 great prophets and 6 missionaries. AI even knew how to build Chichen Itza in their falling capitol just to let the citizens celebrate a golden age before imminent death. It is only the magical balance between those actions that AI could not maintain. A few years later, AI still can do all these things, the technical stuff which even my smartphone can be coded to do (I do not mean graphics here, of course). What I mean is that AI does not have to learn the balance. AI is the balance. This is what we can call 'intelligence'.

Obviously, balance is the problem now. But it is misleading to think that everything else is ok, and this is just a minor thing which will be fixed upon sufficient testing. It probably will, but it is by no means minor, it is the whole AI thing which has not been done properly.

 

[Firebug]

I suppose it is a bit unfair to ask a topicstarter to explain stuff which is only copied from somewhere

Well, i did write it. But it's copied from my reddit post.

 

[isau]

I'm not too worried because I feel in the end I will be playing a modded version of Civ 6, because modders inevitably are better at balancing the emerging metagame than developers are, for a number of reasons. The Community Balance Patch is noticeably harder than unmodded Civ V BNW and has crushed me a few times. (I still remember jumping when Erik the Viking showed up on my screen and declared war on me even as he was busy taking out the last of Ethiopia).

The game will be unbalanced at release. I simply expect this. The game has too many components to reasonably expect not to find some exploits. This has happened with every civ game, and in fact every strategy game I've played in the past 10 years. The final product will be something different. Hopefully we'll get more balance patches than we got with BNW (I'l confess the final state of balance in that game was highly disappointing). But as long as mechanics are solid I know I'll have a game I'll be spending 100s of hours on.

 

[Camikaze]

I suppose it is a bit unfair to ask a topicstarter to explain stuff which is only copied from somewhere, but since there are people here who agree with the message I will nonetheless do it: I kind of understand what you guys mean by 'balance', but what else is AI other than balance, that you are happy with? AI knows how to build units, to move them around and to construct districts in cities? These are all technical stuff as in it just requires corresponding procedures to be coded. AI always knew how to build units in Civ V when Mongols rushed your city with 8 trebuchets and one warrior in renaissance. AI always knew how to propagate religion too, when Haile showed up at your borders with 3 great prophets and 6 missionaries. AI even knew how to build Chichen Itza in their falling capitol just to let the citizens celebrate a golden age before imminent death. It is only the magical balance between those actions that AI could not maintain. A few years later, AI still can do all these things, the technical stuff which even my smartphone can be coded to do (I do not mean graphics here, of course). What I mean is that AI does not have to learn the balance. AI is the balance. This is what we can call 'intelligence'.

Obviously, balance is the problem now. But it is misleading to think that everything else is ok, and this is just a minor thing which will be fixed upon sufficient testing. It probably will, but it is by no means minor, it is the whole AI thing which has not been done properly.

There are aspects of the AI that are not really related to balance. For starters, there's the difference between the objective value of various strategies, and the value placed on those strategies by the AI. If there is significant objective value in maintaining an army of at least, say, three units to defend against attack from a strong neighbour, and the AI only assigns a low value to that strategic choice (despite being fully capable of achieving it), then the problem is not so much the balance of that strategy against other game strategies, as the value placed on the strategy by the AI. That is, the decision of the AI.

Another area in which balance isn't the same as the AI is in combat. Assuming an AI is able to produce an army objectively sufficient to conquer another civ, their capacity to actual utilise the terrain and move their units so as to respond to the enemy, is not really a matter of balance in the sense that that term is normally meant. The overall strategy may be perfectly fine, but the individual decisions which go into the execution of that strategy may be poor.

 

[Gali]

Totalbiscquit does not put his livestreams on youtube but there is an unofficial channel that does so with his permission as long as they wait a few days to give his subscribers first access and don't monetize the videos. Just search Totalbiscuit unofficial live stream.

The AI in Civ 6 can move and shoot and will hunt down units. i saw a barb slinger hunt down filthy's wounded warrior for 10 turns before it tracked it down and killed it. He sent an unescorted slinger near a barb camp and it was annihilated in one turn. Barbs will now attack out of their camp. Filthyrobot has 5000 hours of multiplayer in civ 5 and views Deity as a joke and he was worried about PRINCE level barbs in civ 6. Not scared but he felt compelled to build military to deal with them.

Remember the AI will get unit strength bonuses for each difficulty as well as civil boosts. And on higher difficulties the AI will probably remain relevant long enough for corps and armies to become active. That and support units will allow it a level of stacking it has never had. Add in air power and the AI swarm will be thicker than ever especially as it can use player roads. With melee units being useful again and a threat to cities AI spam will be more threatening. Heck if the AI pillages properly it won't even have to hit your walled city but start trashing your districts.

If the modding tools even match Civ 5 BNW levels I expect AI mods will be out in a few months that will tweak the AI to use the more efficient strategies found by humans. balance should also be easier in Civ 6. Social policies were a nightmare to balance. Government policies offer far easier balancing.

 

[homan1983]

This thread is wrong on all accounts. The AI is terrible. Robot, probably the best person at civ6 who currently has access to it, still played terrible. He skipped all the tutorials and notes, didn't know how housing and amenities worked, didn't know how the price of districts was affected by number of cities and other districts, didn't know the core foundation of unit roles (i.e. archers don't generate ZOC, heavy cav ignore ZOC etc.), didn't build any early districts and hence was generally behind in science/culture and had no early game great people early. He also didn't pre-plan his cities as he is still learning.

He sitll took 3 archers and a scout and took every city on the continent until he got bored. Imagine what it will be like when people have mastered the game.

The AI at this state is simply unacceptable. The fanboys need to give it a rest now.

Moderator Action: Please do not use the term "fanboy" -- we regard it as trolling.

 

[Deggial]

... not because outteching or outexpanding the AI was easy, but because of the AI's tactical incompetence. ...

The point is: The actual tactical AI really isn't that bad! It is definitely not Civ5-level of bad and maybe it is even good!

First of all, I haven't seen the major Civ5 bloodcurdling fails (for example watering units while in attack range).
On the other hand, I've seen really smart (or at least reasonable) moves. Slingers were sidestepping (and then attacking) to make room for a melee unit. Sieged cities switched to unit production in order to counter the (overwhelming) attacking forces. And not only this, but the nearby city switched to unit production, too! The fortified unit stayed in the city in order increase the city defense value and didn't suicide against the attacking forces. (While the mentioned unit was melee, I hope, this will be true for ranged defenders, too.)

So, if the obove observations are true, I believe that there is in fact reason for optimism.
Because, as others have written (and I in another thread), unit-production rate and unit-upgrading is very accessible for tweaking and also very responsive to simple boni on higher difficulty levels.

 

[Big J Money]

Could be a direct side effect of roads no longer costing maintenance + trade units auto building them. I could envision times in which the fastest way for a barb to get a captured civilian back to the camp is via a road in neutral territory.

The pathfinding for support units avoiding combat was strangely bad in Civ V too. It seems as simple as always recalculate a route in such a way as to not put you in any hexes than can be captured or attacked next turn. I have to imagine Firaxis tried this though and there are problems with it. I wonder how many AIs function with the concept of a "backup plan". So if for turn there is no safe path, they give up the goal of reaching destination for something else, like maybe trying to hide the unit in the woods or something.

 

[MadDjinn]

I'm sort of amused by this thread.

On the one hand, TB - who I would posit knows nothing of AI design nor what could possibly be built - says the AI is fine. On the other hand, there are a number of posts here saying that obvious shortcomings of the AI are 'easy to fix/change'. Neither of which is true, generally.

the more complex the system, the harder it is to get 'easy fixes'. The AI, and for this case let's only look at it inside it's own game system, is lacking in a number of areas. But that's nothing new for Firaxis AI in new games. It's not that the AI needs to be a deep thinking machine, but rather than sections of it just don't work well inside the game systems -- this part is definitely 'normal' for Firaxis AI.

Ie, declare war -> forget to send units.
a 'simple' fix would be to ensure the diplo AI doesn't trigger a war without the military side actually having units available to prosecute the war.

Except to fix that, first you'd have to have that conflict within the AI (diplo side vs military side). Then have the military side be able to prebuild an army (add conflict with the Economic AI). Add in some forethought of where to build the units, let alone does it need the encampment first due to low strategics, or better... get a builder out to go improve the strategic resource first, then start building units. You could easily telescope out a number of things that would have to happen first. Ofc that doesn't even cover 'what units to build', which falls under the 'what does the enemy have' side of things. Never mind the need to have the military AI be able to group units together, choose to leave others behind for other tasks, and ofc when things go bad, transfer unit between groups based upon need.

 

[Gali]

Filthy robot could go to civ 4 on launch play on Prince and roll the AI in the exact same way. Any deity level player can roll Prince in the next game and roll it easily. The AI is not going to seriously challenge elite players below the top difficulties. For Filthy robot he'll play MP. Haters can go into a rant thread or bring actually relevant discussion and not mere negativity.

The AI is going to need time and modders to reach its potential. SO DID CIV 4. This thread is about the fact that the release state of AI is in a much stronger position than in civ 5. If modders managed to turn that Ai into a decent state they can do much more with civ 6. If Civ 6 has real mod tools on release and not after 2 years the Ai will improve much more quickly.

Especially since civ 6 seems to be in a much stronger position mechanically and structurally than civ 5. It was hard to program diplo AI while fundamentally redesigning the entire diplomatic system. Combat had similar redesigns. Cav was overpowered then melee units and finally range units.

 

[stealth_nsk]

There's a big difference in AI problems due to algorithms, priority values and difficulty level. Values are quite easy to balance and that's actually happening now. AI algorithms could be considered final, we can't expect any significant changes by release. In short - AI algorithms look good, AI priorities look quite bad. The difficulty level is not designed for serious challenge.

But surely, everything depends on people expectations. If someone expected to see AI being able to play tactically equal to huamns... well, those people have their right to complain.

 

[homan1983]

Filthy robot could go to civ 4 on launch play on Prince and roll the AI in the exact same way.

No he couldn't. 3 archers in civ4 won't get you anywhere. Also the AI in civ4 with the AI patches is pretty damn amazing actually.

He is also not particularly good at civ6, and I don't mean this as an insult.