I've heard the Naval AI in BNW is really bad. But what about at Immortal difficulty?

ThorHammerz said:

If it was a concern during game balance, then Firaxis would not have given naval units as much movement as they have.

Click to expand...

Or perhaps Firaxis thought that the naval units needed that much move in order to feel viable and fun in a single player game?

ThorHammerz said:

Land units do not have a large pool of movement points, and yet they suffer from the same shoot-and-move issue.

Click to expand...

Camel Archers and Keshiks? Would it actually be fun to have the Deity unit spam combined with using those units effectively?

Potentially you'd argue that if they enabled that then they could reduce the Deity unit spam, I suppose.

ThorHammerz said:

If they did, implementing a "is this a naval unit" boolean check would be trivial.

Click to expand...

And would also lead to inconsistent rules.

There's a reason "Threat" is used in many/most RPGs/MMORPGs -- yeah, it's arbitrary and it might not make a ton of sense that the monster will keep attacking the fighter in plate armor rather than the robe wearing mage in the back...but it makes monster behavior consistent and you can balance the enemies/encounters with that rule in mind.

LordBalkoth said:

O
Snip

Click to expand...

Civ was not designed to be a "free win" game. The programmers (like many others) chose to invest as little as possible into creating a functional tactical AI, and instead give the AI a plethora of flat bonuses.

There's nothing particularly hard to understand about it, and trying to justify sloppy software design with game design theories is not a good way to look at a game.

ThorHammerz said:

Civ was not designed to be a "free win" game.

Click to expand...

Where in the world did I say anything like that?

ThorHammerz said:

The programmers (like many others) chose to invest as little as possible into creating a functional tactical AI, and instead give the AI a plethora of flat bonuses.

Click to expand...

Speaking as a programmer, trying to create a "functional tactical AI" for a game with the world scale of Civ 5, the hex system, and terrain (which includes everything from crossing rivers to mountains making choke points to terrain bonuses and beyond) would be an absolute nightmare. Imagine just a case where you have like half a dozen camel archers on completely flat terrain except for a chokepoint so you need to rotate those camel archers through that one spot (and it's not even simply move, fire, move back to original spot since you have some camel archers that need to move, fire, and move two spots away while others move two spots, fire, and move one spot away (four movement)).

That's already extremely complicated. Now figure out how to recognize those chokepoints in general on maps that are randomly generated, how to position the archers correctly to set-up for the next turn, add in other terrain, etc. I don't see there being any reasonable way to be able to account for all of the potential scenarios. The combat AI is always going to come down to "Get a few decent general algorithms and then give bonuses to make it tougher."

This almost feels like this: https://xkcd.com/1425/

Which is not to say I think Civ 5's AI in general couldn't be massively improved with some relatively simple fixes. Things would include...

1. no massive head start, just overall bonuses as necessary
2. better city placement locations (and not founding cities on snow islands in the middle of the ocean on turn 200)
3. better social policies (actually going Tradition or Liberty depending on the AI's city count goal, actually taking Rationalism, etc)
4. better army composition (and maybe not build 2378326832 empty Aircraft Carriers)

Etc. But none of those are related to the actual combat itself when an army arrives at the "battlefield," they're just the set-up.

I've seen people talk about how the Starcraft 2 AI is much better...but it's really not. It has pre-defined maps with explicitly designated expansion sites. And all it does it expand when it wants, build an army, and then run that army into your front door. It's never going to set up Siege Tanks and lure your army into an artillery trap. It's never going to drop in your base. It's never going to do any of the crazy tricks players do.

I mean, here's a question: why are you upset about the fact the AI can't move and fire...but I've never seen *anyone* complain about how the fact the AI doesn't send mounted units to pillage all your land during war (seriously -- just one Horseman/Knight/etc to each city that it's not sending its main army at is going to be able to freely pillage a ton of stuff (and potentially steal your workers) and be very hard to kill due to healing 25 or 50 HP per turn from pillaging)? Shouldn't a good AI do that as well?

ThorHammerz said:

There's nothing particularly hard to understand about it, and trying to justify sloppy software design with game design theories is not a good way to look at a game.

Click to expand...

What's hard to understand is where you draw the line between "not enough effort into the AI" and "sufficient effort into the AI." Is there a lot of sloppy AI design in Civ 5? Certainly. I even gave a list of four huge things that could be massively improved very easily (and there's certainly more than that).

But that doesn't mean that everything not perfect in Civ 5 is solely due to being lazy -- sometimes game designers think fixing a certain issue makes the game worse overall...doubly so in a game that's not intended to be competitively balanced for E-Sports.

Isn't the entire premise of giving the AI a competent military/tactical AI is so that you won't need to give them the stupid bonuses they are given to put them at par on Deity?

It's not just Keshiks and Camel archers - you have the same problem with 2-move units such as your regular rank-and-file archers - bazookas. Of course, in a world where the AI can M&S, those would be the most annoying.

LordBalkoth said:

But that doesn't mean that everything not perfect in Civ 5 is solely due to being lazy -- sometimes game designers think fixing a certain issue makes the game worse overall...doubly so in a game that's not intended to be competitively balanced for E-Sports.

Click to expand...

Yes, and that's why I said, it was a conscious design/budget/planning decision not to invest into creating an AI that goes beyond "meh, it sort of works".

ThorHammerz said:

Isn't the entire premise of giving the AI a competent military/tactical AI is so that you won't need to give them the stupid bonuses they are given to put them at par on Deity?

Click to expand...

Define "stupid." Do you mean no bonuses at all and that any bonus is stupid? Because I don't realistically see that happening and I don't think it would be reasonable to expect it.

Think of all the work that went into Deep Blue. And Civ 5 is far, far, far more complex than Chess in terms of figuring out possible moves.

But if you mean that they should be given *smaller* bonuses on Deity, then yes, I think some things could definitely have been done to achieve that.

To go back to Starcraft 2, Blizzard intentionally made the "Insane" AI cheat (and were quite clear about it -- unlike the easier difficulties). The best players still stomped it. Note that the average player can't beat the second hardest AI, though, so that was still fairly difficult...but when it's a turn-based game (and thus you don't have to juggle things in real time) the "strategy" and "management" parts become a lot easier.

ThorHammerz said:

It's not just Keshiks and Camel archers - you have the same problem with 2-move units such as your regular rank-and-file archers - bazookas. Of course, in a world where the AI can M&S, those would be the most annoying.

Click to expand...

If you fixed it so the Archers -> Bazookas move and shoot but the Keshiks/Camel Archers or ships with Logistics didn't move/shoot/move then people would be complaining about that, no? "Wow the AI is so dumb can't even use those units properly."

I'm not trying to start a slippery slope thing here, just pointing out that there's a lot of things related to this which people might still complain about.

ThorHammerz said:

Yes, and that's why I said, it was a conscious design/budget/planning decision not to invest into creating an AI that goes beyond "meh, it sort of works".

Click to expand...

Your exact words were

"The programmers (like many others) chose to invest as little as possible into creating a functional tactical AI."

There's a difference between saying "Eh, it works well enough" and "It's terrible but we don't care, MONEY MONEY MONEY!"

LordBalkoth said:

Define "stupid." Do you mean no bonuses at all and that any bonus is stupid? Because I don't realistically see that happening and I don't think it would be reasonable to expect it.


Your exact words were

"The programmers (like many others) chose to invest as little as possible into creating a functional tactical AI."

There's a difference between saying "Eh, it works well enough" and "It's terrible but we don't care, MONEY MONEY MONEY!"

Click to expand...

Sorry, I should be been more specific. A large number of bonuses to every facet of empire growth, on top of free units and technologies off the bat is rather stupid.

You cannot really compare deep blue with the Civ 5 AI, as it essentially goes through all possible moves and deduces [however many moves ahead it can "look" into].

"The programmers (like many others) chose to invest as little as possible into creating a functional tactical AI."

Click to expand...

I should have been more specific again: They invested little as possible, to make an AI that has some semblance of working/functionality.

As both of us have noted, the AI has many, many, many shortcomings. What we disagree on is simply whether or not making a poor AI was a conscious design decision..... in which I argue it is not - Jon Shafer admitted as much that it was hard for them to make a competent AI to function in 1UPT. They did not "intentionally" make the AI dumb.

I did actually play a continents game with an improved AI mod - I got my arse kicked by an AI navy. I think it was Diety or Immortal.

Basically they had whole fleets sitting off my coast and anytime I tried to get subs out the AI would just chase them down and sink them effortlessly. So be careful what you wish for.

I also think part of the problem with water maps and the AI is embarked units and submarines

-Embarked units are just too easy to destroy. Everyones had a game where you have a city with coastal or lake tiles and the AI moves it units continuously into the water where you can sink them in 1 hit.
They should have just given embarked a better defense against ranged attacks and reduce the penalty to attack on land. That would cover a bit of the AIs mistakes.

-Also submarines are just overpowered but the AI doesn't know how to use them which gives a massive advantage to the human player. Part of the problem is the Wolfpack Bonus - This promotion usually allows you to sink a top tier naval unit in 1 hit. If you have a Coastal city with Brandenburg Gate & Heroic Epic a sub can have a 90% attack bonus.
-Then there is also not enough units to counter submarines. The destroyer is in possibly the worst part of the tech tree and it is far too expensive to build (only 50 hammers less than a nuclear sub). Considering that a Battleship has a range of 3 and can attack cities & units without putting itself in danger the destroyer for its cost is a liability. It should really cost no more than 250 hammers to build (the same as the Ironclad or arguably less since a Destroyer is a small vessel with a streamlined design)
Helicopters also really need to be able to fly across oceans (possibly staying within 5 tiles of aircraft carriers or cities) and should have extra promotions allowing them to see submarines and having an attack bonus against them.

Ring might be the most interesting water map that handicaps the AI as little as possible. There is a narrow channel around the outside that counts as an ocean. The AIs will use it for warfare, but a human player probably would see ships there as wasted.

ThorHammerz said:

Sorry, I should be been more specific. A large number of bonuses to every facet of empire growth, on top of free units and technologies off the bat is rather stupid.

Click to expand...

I completely agree but that's not directly related to the combat AI. I've long thought the Deity AI should start at the same tech level and with the same units as the player, just with more general modifiers (and have it actually take good policy trees).

ThorHammerz said:

You cannot really compare deep blue with the Civ 5 AI, as it essentially goes through all possible moves and deduces [however many moves ahead it can "look" into].

Click to expand...

The Civ AI looks ahead too, no? Looks at all possible ways to move stuff, what cities to attack, what victory condition to aim for, etc. While it might not look as far ahead in general (Deep Blue may compute every possible path for 50+ turns) it also has a lot more to look at for each turn.

ThorHammerz said:

As both of us have noted, the AI has many, many, many shortcomings. What we disagree on is simply whether or not making a poor AI was a conscious design decision..... in which I argue it is not - Jon Shafer admitted as much that it was hard for them to make a competent AI to function in 1UPT. They did not "intentionally" make the AI dumb.

Click to expand...

The only "conscious design decision" I'm referring to is the AI move -> shoot (or lack thereof, rather). Things like moving embarked units next to a city or whatever is just flat out stupid. Ditto larger scale things like building empty carriers, skipping Rationalism, taking Piety/Honor to start, etc.

Redaxe said:

-Also submarines are just overpowered but the AI doesn't know how to use them which gives a massive advantage to the human player
-Then there is also not enough units to counter submarines.

Click to expand...

Indeed to both.

LordBalkoth said:

The Civ AI looks ahead too, no? Looks at all possible ways to move stuff, what cities to attack, what victory condition to aim for, etc. While it might not look as far ahead in general (Deep Blue may compute every possible path for 50+ turns) it also has a lot more to look at for each turn.

Click to expand...

It doesn't really. It especially does not consider all possible ways to move stuff... not in the sense we would imagine it at least (it will consider a series of weights).

The only part that can be argued as a "lookahead" (in terms of military) would be its pathfinder algorithm(s), which is used in the base version extremely inefficiently. The choosing of the points A and B is done based on a series of weights given to each plot ("fertility" for settlers, some system which was designed to represent relative danger; each of these uses weights of each plot & surrounding plot - it does not consider what might happen in the future in the sense of looking ahead).

The policies and technology flavors are indeed propogated down the tech tree (so yes, that can be thought of as looking ahead in that sense), but the flavor weight assignments are often so arbitrary and so out-of-whack that it might as well not be there (non-existent or at most, very limited testing).

Everything else is flavor/weight based and often computed on-the-fly - i.e. the AI will (at most) consider a limited set of current variables. There is no "looking ahead" with regards to the strategic AI - it simply computes what to build/do based on the flavours of the item it is using, and its own current flavours (modified by current grand strategy, which is in turn often set by pre-determined, often hard-coded "checkpoints" - i.e. do i have 4 cities? Is it past turn x yet? Is the amount of land unclaimed on this continent, below/above a certain threshold? Do I have a certain tech yet? It is also set by the AI's own base flavours).

Gotcha. I assumed it would do something like say "From my scouting I see 10 possible city sites that I like and I want to expand to the three best ones I can get with the goal of doing it by turn 50" and then try to reach that goal (unless something prompts it to change).

Redaxe said:

I did actually play a continents game with an improved AI mod - I got my arse kicked by an AI navy. I think it was Diety or Immortal.

Basically they had whole fleets sitting off my coast and anytime I tried to get subs out the AI would just chase them down and sink them effortlessly. So be careful what you wish for.

I also think part of the problem with water maps and the AI is embarked units and submarines

-Embarked units are just too easy to destroy. Everyones had a game where you have a city with coastal or lake tiles and the AI moves it units continuously into the water where you can sink them in 1 hit.
They should have just given embarked a better defense against ranged attacks and reduce the penalty to attack on land. That would cover a bit of the AIs mistakes.

-Also submarines are just overpowered but the AI doesn't know how to use them which gives a massive advantage to the human player. Part of the problem is the Wolfpack Bonus - This promotion usually allows you to sink a top tier naval unit in 1 hit. If you have a Coastal city with Brandenburg Gate & Heroic Epic a sub can have a 90% attack bonus.
-Then there is also not enough units to counter submarines. The destroyer is in possibly the worst part of the tech tree and it is far too expensive to build (only 50 hammers less than a nuclear sub). Considering that a Battleship has a range of 3 and can attack cities & units without putting itself in danger the destroyer for its cost is a liability. It should really cost no more than 250 hammers to build (the same as the Ironclad or arguably less since a Destroyer is a small vessel with a streamlined design)
Helicopters also really need to be able to fly across oceans (possibly staying within 5 tiles of aircraft carriers or cities) and should have extra promotions allowing them to see submarines and having an attack bonus against them.

Click to expand...

I kind of disagree with submarines being overpowered because other players can easily get them up on time to defend themselves.
I kind of agree with what you said about how expensive destroyers have become because Ive had privateers and rarely ironclads that i have had upgraded to destroyers and their upgrades were really expensive even with military tradition. Military tradition reduces unit upgrade costs but it was still expensive to upgrade a few units. I imagine how much more expensive it wouldve been if i wouldve upgraded without military tradition.

LordBalkoth said:

Gotcha. I assumed it would do something like say "From my scouting I see 10 possible city sites that I like and I want to expand to the three best ones I can get with the goal of doing it by turn 50" and then try to reach that goal (unless something prompts it to change).

Click to expand...

In that sense, every visible, non-owned plot is essentially a "possible city site". Only the best tiles (computed each time the AI wants to evaluate potential city sites... i.e. when it finishes building a settler) in the goldilocks zone of a range of 4-6ish tiles of the nearest city are considered, unless expanding offshore to another continent (this result is usually never met however... evident in island/naval maps where the AI rarely expands beyond it's home island/continent, unless the next island is literally 2 tiles away).

In the base game, you will also notice many stupid things happening with AI settlement patterns as well: cities being clumped too close for its own good, and the tendency to settle 1-2 tiles away from the coast if it is not a "coastal" (read: England, Polynesia, etc) civ, which is identified somewhere in the XML.

All in all, you could argue that the idiocy of the AI was pre-designed from the conceptual stage - it's simply that developing AI costs money, and when profit margins need to be as big as possible, why hire extra programmers in AI when you can toss the job to a junior-middling programmer who can build something that "sort of works... sort of".

reddishrecue said:

I kind of disagree with submarines being overpowered because other players can easily get them up on time to defend themselves.

Click to expand...

The point is that you *can't* defend against submarines. If I have 20 Nuclear Submarines and you have 20 Nuclear Submarines...I will still win massively if I attack first. Either I wipe out 20 of your subs in the opening volley (one-shots) or I wipe out half your subs in the opening volley (two-shots).

ThorHammerz said:

All in all, you could argue that the idiocy of the AI was pre-designed from the conceptual stage - it's simply that developing AI costs money, and when profit margins need to be as big as possible, why hire extra programmers in AI when you can toss the job to a junior-middling programmer who can build something that "sort of works... sort of".

Click to expand...

Keep in mind the *only* part of the AI that I am defending is the lack of move -> shoot. Because I think it would simply not be fun for the player in many cases. The AI doesn't complain about the player being able to do crazy things with some units

LordBalkoth said:

Keep in mind the *only* part of the AI that I am defending is the lack of move -> shoot. Because I think it would simply not be fun for the player in many cases.

Click to expand...

I have to agree with this. I read here on civfanatics how terrible the AI is all day and I guess the clue is in the name, we are indeed fanatics. But the average (civ) game player? I think the AI is fine. We forget that Civ has a lot of things to juggle because we know how to juggle them all well. But newer players trying to get to grips with everything need help.

LordBalkoth said:

Keep in mind the *only* part of the AI that I am defending is the lack of move -> shoot. Because I think it would simply not be fun for the player in many cases. The AI doesn't complain about the player being able to do crazy things with some units

Click to expand...

Why this feature should be included or not in the game can be debated: I am simply stating/refuting against the argument that this is by intention; it is part of a larger systemic problem on the AI at large, as opposed to "this is not fun, let's keep it out".

LordBalkoth said:

The point is that you *can't* defend against submarines. If I have 20 Nuclear Submarines and you have 20 Nuclear Submarines...I will still win massively if I attack first. Either I wipe out 20 of your subs in the opening volley (one-shots) or I wipe out half your subs in the opening volley (two-shots).



Keep in mind the *only* part of the AI that I am defending is the lack of move -> shoot. Because I think it would simply not be fun for the player in many cases. The AI doesn't complain about the player being able to do crazy things with some units

Click to expand...

That already depends on player skill now because most abandoned units get easily killed in high seas particularly by submarines. Another disadvantage that submarines have is that theyre visible by the local 2 tile radius city target shooter. If submarines werent that visible then maybe submarines couldve been OP. Another thing is that units getting stuck inside the city because of surrounding enemy ships wont allow new submarines to shoot while inside the city. Unlike visible ranged ships, submarines cant shoot while theyre inside a coastal city. Submarines need to be in water to shoot.

ThorHammerz said:

Why this feature should be included or not in the game can be debated: I am simply stating/refuting against the argument that this is by intention; it is part of a larger systemic problem on the AI at large, as opposed to "this is not fun, let's keep it out".

Click to expand...

You're speculating that it wasn't intentional. I mean, the AI also doesn't all declare war on the player if it looks like the player will win a Science victory. Is it becaue the AI is bad or is it because that wouldn't be fun?

reddishrecue said:

That already depends on player skill now because most abandoned units get easily killed in high seas particularly by submarines.

Click to expand...

Talking about a simple case of my entire fleet versus your entire fleet (or at least a fleet where we each have 20 submarines in an area).

LordBalkoth said:

You're speculating that it wasn't intentional. I mean, the AI also doesn't all declare war on the player if it looks like the player will win a Science victory. Is it becaue the AI is bad or is it because that wouldn't be fun?

Click to expand...

Aren't we both speculating? I'm just making my conclusions based on how the .DLL is coded (its quality, or the lack thereof).

If it sounds like a fart, smells like a fart, and feels like a fart - it is, without needing to use some theory attempting to explain it otherwise, a fart.

As for winning the game, they will declare war on you if they hate you enough for other things (it's like ideologies in which by itself, it will not, but can definitely be the "final drop of water"), if it`s the same VC they themselves are going for. You gain an extra "game-threat" consideration when considering how desirable it is for the AI to go to war. This is modified by the VictoryCompetitiveness setting for each Leaderhead. Quite strange tbh.

x2Madda said:

I have to agree with this. I read here on civfanatics how terrible the AI is all day and I guess the clue is in the name, we are indeed fanatics. But the average (civ) game player? I think the AI is fine. We forget that Civ has a lot of things to juggle because we know how to juggle them all well. But newer players trying to get to grips with everything need help.

Click to expand...

The AI is fine to be 'stupid' on lower levels, but adding in move and shoot at higher levels would be fine as long as their 'cheatin' was reduced to compensate. I would much rather struggle against a smarter AI than a 'cheatin' AI. A truly superior AI would only need to get smarter on harder levels.

It's just so lame when the AI constantly drops it's units in the water for me to one shot or sends it's siege into my territory alone or forgets to bring melee units to actually take a city that it easily took to zero defense.