Forum adverts like this one are shown to any user who is not logged in. Join us by filling out a tiny 3 field form and you will get your own, free, dakka user account which gives a good range of benefits to you: No adverts like this in the forums anymore. Times and dates in your local timezone. Full tracking of what you have read so you can skip to your first unread post, easily see what has changed since you last logged in, and easily see what is new at a glance. Email notifications for threads you want to watch closely. Being a part of the oldest wargaming community on the net. If you are already a member then feel free to login now.

A conspicuous absence of talk about the AI. I think the grand total of AI comments that I've seen is a quick mention that enemy troops tend to stay for too long in places where they'll get shot. As I recall, that was a problem mentioned a few months ago. So I'm a bit suspicious about the current state of the game's artificial intelligence.

Why am i not surprised, oh wait always the same Spiel, https://www.google.com/search?q=rtw+ii+ai+meme&tbm=isch&ved=2ahUKEwjhsozEhKriAhWGdlAKHSczAPgQ2-cCegQIABAC&oq=rtw+ii+ai+meme&gs_l=mobile-gws-wiz-img.12...1778.1778..2684...0.0..0.113.113.0j1......0....1.Bb3a8Q8jbMk&ei=lpPiXOGPHobtwQKn5oDADw&bih=512&biw=360&client=ms-android-sonymobile&prmd=imsvn#imgrc=eSkrH4nMT1mVoM

Honestly AI has, at least visually, stagnated in the games market.

FPS and some other formates hide it a lot better, but RTS games tend to show that AI really isn't improving by leaps and bounds like everywhere else. I think the only company at present really pushing AI is Blizzard and they have somewhat nuts resources open to them - so one hopes their developments trickle down into the market in general.


That said I still recall the AI in Medieval 2 (the game everyone loves on). It HATED sieges and most times would retreat from them and draw up battlelines forcing you to either wait them out or move out and attack. Which isn't a bad idea, but its somewhat anti-climatic when you're expecting a siege battle.

Then there's the Empire AI which retreated from their Star fortress as I attacked - even though I'd made a mistake and had NO artillery and NO siege weapons. Seriously I ran up, climbed the walls with no opposition and soon was firing their own defensive cannon at them. They, of course, retreated to the other end of the town on the map and I had to leave the fort and hunt them down. (huge maps are fun but not when it turns into a game of army chase).



Suffice to say the TW AI does decently, but in the end AI is just coming up short in many games. You hear the same issues in Stellaris (although honestly I'm surprised their sector AI is as bad as it can be since it seems to just build at total random).
Heck the Pokemon AI has for generations just random picked what it does each battle turn!

Oh jesus the nonsense you'd never be able to get away with if the Pokemon AI actually made choices based on basic information (like switching out of an unfavorable pairing, or even just using move sets that make sense). You'd never be able to beat the Pokemon League Champion with a Lv. 1 Rattata ever again!

EDIT: Now that I think of it, FGO's ai also picks targets at random, except for some of the challenge boss fights where the AI is coded to kill Casters first because Merlin was making every challenge fight a non-challenge XD

What makes me laugh is when the pokemon AI uses things like sleep over and over even if the opponent is already asleep.


And wasn't there a goldfish who played pokemon for a while and got pretty far though purely random movement.


I mean I get that at its core its a kids game, but still you'd think they'd put a very basic bit of thinking into the choice of what to do per turn. But I figure their main issue is if they did that they'd have to make it so that you auto-healed at the end of every fight otherwise you'd be forever going back and forth to the poke-centre.

It is one time where I'm honestly confused why the AI is so bad when you'd expect it to be good - knowing the status of its opponent; knowing the weak and strong combos and knowing the element line ups (heck pokemon has so many now its actually quite a bit to remember all the various combos that work against each other).

 Overread wrote:
Honestly AI has, at least visually, stagnated in the games market.

FPS and some other formates hide it a lot better, but RTS games tend to show that AI really isn't improving by leaps and bounds like everywhere else. I think the only company at present really pushing AI is Blizzard and they have somewhat nuts resources open to them - so one hopes their developments trickle down into the market in general.


That said I still recall the AI in Medieval 2 (the game everyone loves on). It HATED sieges and most times would retreat from them and draw up battlelines forcing you to either wait them out or move out and attack. Which isn't a bad idea, but its somewhat anti-climatic when you're expecting a siege battle.

Then there's the Empire AI which retreated from their Star fortress as I attacked - even though I'd made a mistake and had NO artillery and NO siege weapons. Seriously I ran up, climbed the walls with no opposition and soon was firing their own defensive cannon at them. They, of course, retreated to the other end of the town on the map and I had to leave the fort and hunt them down. (huge maps are fun but not when it turns into a game of army chase).



Suffice to say the TW AI does decently, but in the end AI is just coming up short in many games. You hear the same issues in Stellaris (although honestly I'm surprised their sector AI is as bad as it can be since it seems to just build at total random).
Heck the Pokemon AI has for generations just random picked what it does each battle turn!

That is true but etw atleast had the excuse of a total new engine, how Napoléon and rtw II turned out soll medicore/ bad at the start even though the Budget was 40% higher is memeworthy.

Eh. Even a "smart" Pokemon AI would probably still be fairly predictable, which would actually be useful in preparing player's for online play since the hardest and most important skill in high-end PvP battles is your ability to predict what your opponent does next.

To steer this back to the topic, I remember what you said in MII, and I generally abused it. The AI handled siege battles horribly. So horribly that a garrison army of archers/crossbowmen could hold off an army twice it's size because the AI would stand outside the wall in shooting range if you sent one unit out to bait them in.

Rinse and repeat and you win every siege battle that actually happens, and some factions had easy access to cheap/early crossbows. Saved a lot of time and money for building large field armies to beat the rest of the AI to a pulp.

Once you learn how the AI makes decisions, it gets really easy to steam roll right over them.

 LordofHats wrote:
Eh. Even a "smart" Pokemon AI would probably still be fairly predictable, which would actually be useful in preparing player's for online play since the hardest and most important skill in high-end PvP battles is your ability to predict what your opponent does next.

To steer this back to the topic, I remember what you said in MII, and I generally abused it. The AI handled siege battles horribly. So horribly that a garrison army of archers/crossbowmen could hold off an army twice it's size because the AI would stand outside the wall in shooting range if you sent one unit out to bait them in.

Rinse and repeat and you win every siege battle that actually happens, and some factions had easy access to cheap/early crossbows. Saved a lot of time and money for building large field armies to beat the rest of the AI to a pulp.

Once you learn how the AI makes decisions, it gets really easy to steam roll right over them.

Granted the wh II ai is equally bad, especially at dealing with flying units if it has non.
I personally found the AI failed the hardest in City fights in etw, i once hold of 2 stacks with 3 units of conscripts 2 line infantry a general and a unit of guerilla.

Overall i find the situation reflects badly on ca.

Eh.

Overread is right. AI across the industry is lackluster, and hasn't really seen much improvement over the past decade. If anything, CA at least tries to push new mechanics and ideas in the genre they're most well known for. Sure all those ideas don't hit, but at least they're trying.

Picking on them for not pushing AI too feels overly critical.

 LordofHats wrote:
Eh.

Overread is right. AI across the industry is lackluster, and hasn't really seen much improvement over the past decade. If anything, CA at least tries to push new mechanics and ideas in the genre they're most well known for. Sure all those ideas don't hit, but at least they're trying.

Picking on them for not pushing AI too feels overly critical.

And what mechanics did they push exactly since etw /ntw? If anything they stagnated until Attila and even there the Innovation wasn't great, but atleast the AI got improved there.
Also sieges have been downgraded heavily over time and so did formation fighting and the melee model overall due to the terribad engine since etw.

It is nice that they are trying to fix diplomacy so that it not Automatically turns into dickplomacy but that comparativly matters little when I can just shoot my way to victory with missile units.

Warhammer 1 and 2 were vast innovations for CA - they added flying units, super heroes, dragons, massive units, spells, etc... They reworkeda huge amount of stuff.

Even sieges are actually more satisfying from a combate side in many ways because now you're not playing "hunt the army" as you run around a massive map and spend more time chasing the enemy than actually manoeuvring into combat.



I think that the issue is most people have no concept of AI and many games like FPS trick players into thinking AI are far better in them when in actuality the AI is simply having a much easier time because it has far less to think about - its all reactionary mostly, there's no pre-planning or thinking ahead. They can also hide it behind things like pre-designed patrol pathways and pre-design play styles on specific maps etc.... Ergo it appears better than it is whilst RTS games tend to show up the glaring issues far more readily. Esp since most AI are never thinking about the "long game".



Also with diplomacy lets face it - most players will complain unless the AI rolls over to do what they want. And most players are willing to dump an AI ally in a moments notice, heck if the AI self-learned diplomacy from a player they'd just learn to use it to shore up one front or for long distance trade with the long term goal of breaking the alliance and running amock in their once allied lands to take over.

My hope is that things visually settle down and that as a result we see AI get more time to shine and get attention. Blizzard is messing around with it a lot in Starcraft and I think that is where we might well see new ideas and the next generation of RTS AI start to come together. Not forgetting Blizzard doesn't make many actual games; they have the ability to profit without game sales so they've far more resources to put into something like that than a company like CA who has to invest into new games to continue earning. Heck Blizzard has tournaments that bring in hundreds of thousands in prize money alone (or at least raised that in supporting money I think). They've deeper pockets than most.

Warhammer 1 and 2 were vast innovations for CA - they added flying units, super heroes, dragons, massive units, spells, etc... They reworkeda huge amount of stuff.

out of all this only the flying units are actual a new mechanic.

Dmg overtime through spells is comparable to artillery ammo available in ETW for howitzers, damage spells like the flamming skull etc also work similarly to support artilery from shogun 2.
Super heros are either cavalry or infantry units with stats most of the time.
Spells are again the same a temporary buff not unlike some abilities units had in the past like steady or banzai, except this time you can mark a friendly or enemy unit.

Even sieges are actually more satisfying from a combate side in many ways because now you're not playing "hunt the army" as you run around a massive map and spend more time chasing the enemy than actually manoeuvring into combat.

"Kaplan, the players complain about the siege AI, what shall we do?
Let's make sieges just a minescule area where the AI can't maneuvre out of the way."

And this is what is called a band aid, one that even has more shortcomings then before, you can easily abuse the AI plugging gaps in the wall and put archers literally right infront of the unit and kill it without repercussions, infact you can abuse the living hell atm out of the siege AI if you want to.

Also i will give you that, that FPS got a lot better at hiding it's terrible AI, but imo just screwing around with a chaos army that has no flying unit with one flying unit, making it constantly realign and waste stamina also shows that the flying mechanic has from an AI standpoint a very limited interaction capability.
infact just recently i won a ambush battle against a full skaven army because it mindlessly ignored a bunch of flying pirates that killed 3 plague claw catapults whilest just rushing my line.

Also with diplomacy lets face it - most players will complain unless the AI rolls over to do what they want. And most players are willing to dump an AI ally in a moments notice, heck if the AI self-learned diplomacy from a player they'd just learn to use it to shore up one front or for long distance trade with the long term goal of breaking the alliance and running amock in their once allied lands to take over.

the very existence of games like EUIV or Crusader kings would show that this is not the case and there is a lot more to do with an better indepth AI diplomacy system.

 Overread wrote:

And wasn't there a goldfish who played pokemon for a while and got pretty far though purely random movement.
.

That was a twitch plays pokemon thing, and it was one of the earlier gameboy game ones if I remember right.

the very existence of games like EUIV or Crusader kings would show that this is not the case and there is a lot more to do with an better indepth AI diplomacy system.

Aren't those easily broken at times?

Very easily. I'm baffled by anyone holding up paradox games as an example of good or even more than barely functional diplomacy. Its a points-over-time based system with a lot of exceptions and get out of relationship free clauses (country too big, too small, no other possible rivals, hard coded desires for provinces, etc)

Of major strategy companies, they're near bottom (saved from that status by firaxis, who give no rat butts about it )

As for AI, most people don't understand the concept correctly, and don't grasp how massive 'proper AI' would be, or how badly it would ruin games.

Voss wrote:
Very easily. I'm baffled by anyone holding up paradox games as an example of good or even more than barely functional diplomacy. Its a points-over-time based system with a lot of exceptions and get out of relationship free clauses (country too big, too small, no other possible rivals, hard coded desires for provinces, etc)

Of major strategy companies, they're near bottom (saved from that status by firaxis, who give no rat butts about it )

As for AI, most people don't understand the concept correctly, and don't grasp how massive 'proper AI' would be, or how badly it would ruin games.

When you compare with TW the Diplomacy of Paradox is functional. The point was also to highlight that an intersting or improved Diplomacy system is better then non.

Honestly my impression is that most of the diplomacy complaints which aren't "the AI won't do what I tell it to do" tend to boil down to a few things

1) There is rarely if never any actual connection between player and AI actions. Granted few players will go to an AI's aid in war, but similar the AI doesn't tend to do much either, its more convenience thaty ou're against the same foes.

2) Games like Crusader Kings just have WAY more options and depth to the feature. Which is good because their combat element is basically super light. It's mostly a case of balancing your infantry number vs your income and getting the war over fast.
Crusader Kings and TW are both similar in the main map view approach to empire building; but whilst one focuses more on dynasty and diplomacy; the other focuses on army combat. Merging those two elements together would create a very indepth game, but run the high risk that it actually becomes so complex it puts players off.

Sometimes its good to have games that focus in detail on one area; with a TW game most of it is about getting big armies into big 3D battles. A heavy diplomatic system would get in the way.

I've got to say, the diplomacy system is pretty much ignored when I play TW games - get some early game income from trade agreements, but beyond that, I just see enemy diplomats as assassin fodder.
Like with Medieval 2, from my perspective the only really important diplomacy that you can do is with the Pope, when playing as Catholic faction, and that's less diplomacy, more prolonged bribery to prevent getting crusaded against. It's far more about the combat side, or even managing happiness in settlements, than it is about diplomacy, at least from my perspective.

 Dark Apostle 666 wrote:
I've got to say, the diplomacy system is pretty much ignored when I play TW games - get some early game income from trade agreements, but beyond that, I just see enemy diplomats as assassin fodder.
Like with Medieval 2, from my perspective the only really important diplomacy that you can do is with the Pope, when playing as Catholic faction, and that's less diplomacy, more prolonged bribery to prevent getting crusaded against. It's far more about the combat side, or even managing happiness in settlements, than it is about diplomacy, at least from my perspective.

Well, that's because TW games have had a weird to defective diplomacy model. Warhammer's has been almost offensively bad, since alliances are a flat detriment (people can and will drag you into wars where you must break a treaty, ruining your diplomatic ability with everyone, including them) and that mistrust lasted basically forever until recent patches. Plus allies would use your strength in determining there strength in deciding whether or not they were weak enough to confederate with you, which is just pants on head bad, as no matter how weak they are, they're always comparing your strength (X) to their strength of (X)+1.

This one is a lot broader, with all sorts of trades and relationships possible, not just the bare minimum. Even peacefully transferring land, and everything has a value weighted by what they actually need or desire. [2 food to someone with no food is worth more than someone with 20 food]. Its a big shift for the franchise, and from what I've seen, really makes the game more interesting.

----

They announced today that the DLC would be 'story chapter' focused, so you'd see various characters at different points in time and different areas. That's honestly a relief, because the same map and same base relationships get old after a while.
It also allows some of the characters in new scenarios- quite a few of the campaign cast were commanders along the northern borders against invaders, but for the main campaign, that period is over. A DLC pack could actually focus on those campaigns, while not touching the 'three kingdoms' setup.

I'm more curious about how they handle mid-to-late game snowballing.

TW games are characterized by that grueling start to carve an empire, interrupted by a sudden shift into auto-resolve grind mode for victory conditions.

This is an issue with 4x games in general, though.

TW isn't as bad as some - you can at least envision having armies in different places in your empire. A lot of 4* I find tend to fall into the terrible pit of "put all your army in one single place to win".

Stellaries suffers from this still; though I think the king from more recent games was Sword of the Stars 1 which almost requires you to do it against the AI otherwise their superfleet will beat any other fleet you send.

It was a huge shame that their second game, which looked like they were going to go for really tactical battles and limited armies with more spread out forces - ergo a really sweet approach to combat for a standard 4* experience - sadly they utterly utterly messed up the launch and the game was a mess. A huge shame because I really think if it had done well it would have changed things for 4* in a big way.

 Overread wrote:
Stellaries suffers from this still; though I think the king from more recent games was Sword of the Stars 1 which almost requires you to do it against the AI otherwise their superfleet will beat any other fleet you send.

This is why I play corvette fleets in Stellaris, cause managing five swarms of tiny fast ships and avoiding the enemy fleet while wrecking their infrastructure left and right is a lot more entertaining than building a battleship stack XD

 Motsie wrote:
I'm more curious about how they handle mid-to-late game snowballing.

TW games are characterized by that grueling start to carve an empire, interrupted by a sudden shift into auto-resolve grind mode for victory conditions.

Reviews have revealed that when one of the warlords reaches a certain strength level, the game will loosely emulate the actual Three Kingdoms phase of the era by forming the remaining warlords into three large power blocs. This is supposed to happen in a far more natural fashion than the "everyone hates you now, including your vassals" stage of Shogun 2. I'm not sure whether the player is required to be the leader of one of the factions, or whether the player could theoretically end up a junior member. Note that historically, this is not what happened. Historically, Wei (Cao Cao) conquered or vassalized pretty much everyone but Wu (Sun family) and Liu Bei (still homeless at the time) before the third kingdom (Shu Han, under Liu Bei) had even begun to form. But if it works, it should help for the game's purposes. My primary concern is how the process ends up dealing with territory mismatches. The bulk of the warlords are in the north and east. And none of the playable warlords start near Cheng'du (where Shu Han historically had its capitol). So depending on how the criteria for formation of the kingdoms are determined, you could theoretically end up with all three kingdoms focused on the northeast of China.

They announced today that the DLC would be 'story chapter' focused, so you'd see various characters at different points in time and different areas.

Three pretty obvious focuses would be -

1.) The Yellow Turban Rebellion - This one would take place before the current game start. The Yellow Turban warlords in the Yellow Turban DLC are hold-overs who survived the big rebellion (which was *huge*), and later started their own smaller, localized uprisings, and shouldn't be confused with the theoretical events of this scenario
2.) Formation of the Three Kingdoms - a scenario start in which Shu Han has finally been formed. Shu and Wu are alternately allying against Wei, and squabbling with each other over Jin Province.
3.) Rise of the Sima Family - The fact of the matter is that the clan that reunited China isn't one that's currently playable. And they never will be with a normal game start. This particular scenario would focus on the rise of the Sima family within the Kingdom of Wei, leading to (unless the player intervenes) the overthrow of the Cao family, and the eventual reuniting of China under the Jin Dynasty.



Edit -

A couple of other random items -

First, CA announced in a blog post that Three Kingdoms is their most "pre-ordered" game. If nothing else, it appears that anticipation for this game is pretty high. There's no qualifier in that announcement with the word "historical", so it appears that it beat out the Warhammer games, as well. I'm rather surprised by that... Of course, this is just pre-orders, and it's entirely possible (likely, even, imo) that it won't do as well against the Warhammer games in total sales.

Second, and only loosely related - I've been watching a Chinese TV series on Netflix ('War of Kings', for the curious). It's set during the fall of the Qin Dynasty, and has quite a few familiar faces if you've seen the Three Kingdoms TV series (for instance, the same actor who played the virtuous Liu Bei in that series plays the Qin Emperor in this one; also, I haven't double-checked, but based on the look of the series, I'm guessing it's by the same company that did Three Kingdoms). Anyway, the reason why I mention it is that I spotted a famiiliar surname early on. So I double-checked the date it was set in, and thought, "Him? REALLY!?" But a quick check on-line confirmed that the guy named "Liu" really was Liu Bei's distant ancestor, and the first emperor of the soon-to-be-founded Han Dynasty.

Voss wrote:
Well, that's because TW games have had a weird to defective diplomacy model. Warhammer's has been almost offensively bad, since alliances are a flat detriment (people can and will drag you into wars where you must break a treaty, ruining your diplomatic ability with everyone, including them) and that mistrust lasted basically forever until recent patches. Plus allies would use your strength in determining there strength in deciding whether or not they were weak enough to confederate with you, which is just pants on head bad, as no matter how weak they are, they're always comparing your strength (X) to their strength of (X)+1.

Now I understand why other AI High Elves would confederate with Tyrion and not me (Averlorn) despite being BFF's with me and not with Tyrion.

The Dynasty Warriors games have for years built a very steady and stable fanbase of their own and almost defined the era for the gamer in terms of visuals and style. CA is riding that coattail heavily with their Dynasty era game. In addition I bet this game has a massive appeal in China (for games of this type in that market); so that alone could prop its sales up massively over some other titles.

All in all if its got more pre-orders than ever before and its beating the Warhammer games (which if I recall right were some of the biggest sellers for CA) then its a very good sign that the game will sell really well once it hits release.

 StygianBeach wrote:

Voss wrote:
Well, that's because TW games have had a weird to defective diplomacy model. Warhammer's has been almost offensively bad, since alliances are a flat detriment (people can and will drag you into wars where you must break a treaty, ruining your diplomatic ability with everyone, including them) and that mistrust lasted basically forever until recent patches. Plus allies would use your strength in determining there strength in deciding whether or not they were weak enough to confederate with you, which is just pants on head bad, as no matter how weak they are, they're always comparing your strength (X) to their strength of (X)+1.

Now I understand why other AI High Elves would confederate with Tyrion and not me (Averlorn) despite being BFF's with me and not with Tyrion.

Atleast they fixed it for chaos, beat up enemy lord, get's trait beaten, make vassal, problem solved, vassal swarm activated.

Two things I really like so far is the presentation of the tech tree as a watercoloring tree, and the character-focused nature of romance-mode.

Pre-load is up for any other fellow Total War-ophile that pre-ordered, like myself!

Can’t wait to jump into a setting that I know feth all about and have read very little of the games new features. Going to be a whole lotta reloads and restarts this weekend!

I thought this was out already? Seems to be all over my steam

 Argive wrote:
I thought this was out already? Seems to be all over my steam

Na just the influencers got it and of these the uncritical only.

 Argive wrote:
I thought this was out already? Seems to be all over my steam

Steam does a lot of pre-release advertising for major titles. Rage 2 started months ago and still hasn't gone away, despite release and a general lack of interest.

Though some of confusion of 3k probably stems from multiple delays, though I think the game has largely benefited from those.

Looks like a hit so far, setting a concurrent player record for the franchise by a fairly large margin on Steam. I certainly didn’t expect that.

I played a few turns as the military centric dude whose name escapes me, he starts in the top right corner of the map. It was “Records” mode and all I really did was feel out the controls and features, as well as get the graphics pumping on my machine. I’m digging it so far, though I will scrap this Records campaign and go for the Romance mode this extended weekend.