i wish the mechanics of these games would evolve... the unit count thats been nearly the same through the whole series, or the clunky manner in which units move, the lack of variety in maps, the AI which probably doesnt even track its own resources and never initiates diplomacy. why cant these games be better?!
Literally all they have to do to make theses games good is make decent AI
Adam Clark
Every game they bang on about AI improvements but they somehow just make it worse. You would think they would have worked out how to do it seeing how every game since Empire is basically the same
Henry Wright
when you tell a unit to charge and it turns away (with full morale)
Dylan Perez
Why don't these devs use sth like Open Ai but with smaller scale to save fund It would vastly improve the development
Elijah Wilson
I fear that this will end up like Total war Shogun, in that there will be a lack in diversity.
Benjamin Parker
That'll be the norm in a few years desu
Elijah Thompson
You base this on what?
Jose Lee
creative assembly have their priorities wrong and they don't understand why people play their games
Asher Fisher
on the devs not needing to scratching their heads about how to fucking code it The Ai network would perform better and more unpredictable.
Kevin Bennett
how long until we get a game like this with HoI esq logistics and pop cap reflecting the periods with regiment sizes to match
John Adams
never ever look at twg, all those fags want are now more pretty models to masturbate since the rise of fantasoi, saving for a few historychad who still care enough about the campaign mode
Jonathan Hall
Series has been in decline for years now. There's no reason to assume Three Kingdoms will somehow bring it back to its former glory.
John Hill
The sooner it comes out the sooner it can fail and the sooner they can put those resources towards Warhammer 3.
Joshua Carter
My phone isn't powerful enough to run this, don't really care about mobile phones anyways.
Kayden Butler
Or make the MP experience not shit? Like in Shogun 2, per chance? The sole purpose of an AI is to use up resources of your own computer and make the game run like shit in order to simulate a human on the other side of the field.
Please, don't talk about stuff you have no idea about. It's been done before, it has failed before. SupCom 2, Planetary Annihilation, Blitzkrieg 3 all have a neural network AI. You probably don't even know those games exist, let alone how the AI in them behaves. That's not even starting to scratch the surface of why neural network AI and games is a really dumb idea.
Brayden Wilson
Warhammer total war already surpassed the previous titles ‘former glory’. Good bye histcuck we won’t miss you.
Matthew Sullivan
please explain more, I'm interested
Ryder Anderson
warhammer need to fix it sieges, this 1 wall bullshit is retarded
Ryan Price
Based on what? >don't talk about stuff you have no idea about I know user is talking about stuff he has no idea about, that's why I question him on it
Lucas Murphy
I think it'll be perfectly mediocre, it has a lot of good idea's and also some asinine ones.
Nothing about it really 'pops' and makes me say "I want this", it just looks like a bland TW game in a new setting with pretty art. The stuff they don't show off as much, like the gameplay wrinkles each faction has, strike me as more interesting.
James Johnson
It has potential but the amount of shilling I've seen for this game is staggering. Wait a week before buying and check out trusted reviewers. Hint : if they have early access content, they're not to be trusted.
Carson Brooks
the controls for a game like total war are insane for it to actually do anything.... things like MAR.IO only are possible because of sprite sizes and simple controls. to play a game with a battlefield resolution and options selection like total war would require an insane amount of nodes in a neural network and is no where close to feasible yet. they finally got chess and go... a step at a time
Christian Bennett
>the amount of shilling I've seen for this game is staggering It's incredibly niche, I haven't seen it anywhere but the channels of people who already play this shit and on the generals and threads of people who play this shit.
Xavier Adams
ima just pirate it and if its good ill buy it on sale
Jose Wood
I agree it could use some more variety, but it works for the most part. Lets be honest though that CA will never try to improve on it, and it will fall to modders to fix the final end product once the full Mortal Empires map and all DLCs are out. When GW no longer has the ability to hold content for ransom since it is all released then modders can do whatever they want.
Charles Nguyen
the one hope i have is the new ambush mode, only seen one battle of it and having a new objective besides just kill the other player. but outside warhammer II i havent played anything since shogun so is this even a new feature?
Ryan Cook
I quickly read up on SupCom2 and apparently it only uses NN for fight or flight decisions
Chase Flores
>WH3 was delayed for Ching Chong total war I know they want the Chinese market but they couldn't wait another year or two?WH was already a best seller.
isn't it literally just a normal ambush map but if you get a unit to the marker it can withdraw
hardly a massive change
Lucas Baker
its something...
Ryder Anderson
WH3 isn't delayed, yet. If it comes out in 2020 then it has followed the same sort of release time between WH and WH2.
We will probably hear more about it towards the last quarter of this year.
Parker Cook
maybe if were lucky CA will go bankrupt and paradox buys the IP
Jose Phillips
Network AIs are incompatible with game development as a whole. There's a miriad of issues so I'll give you a few bullet-points and you can research all of it yourself.
>Their success is not currently understood. Making a working AI with them is a matter of a huge lottery. >Their intelligence scales with resource usage. Brutally so. >A neural network solves an optimisation problem. Give me an opponent and I'll fuck him up. Real video game AI is about allowing the player to win, but hiding it. >Building a working AI takes lots of time, lots of resources and people who know very well what they're doing. Game development doesn't have access to gargantuan computation centers costing $500 per second of computation, they can't afford people with PhDs in AI and they are extremely tight on time.
These are just the tip of the iceberg. Look at the Starcraft 2 AI program. It started in mid 2016. It beat the pros in December 2018. It was done by the best people in the field. It's currently only able to beat people without giving them an inch. It can only play Protoss vs Protoss on a single map. Any other kind of matchup or a map and it's practically worthless. That's a multi-million project using the best resources avilable to one of the biggest IT companies on Earth, it took them 2 and a half years to make it play one matchup on one map. The only adaptability observed so far is that newer baance patches didn't fuck it up completely.
>It can only play Protoss vs Protoss on a single map Isn't that how most pro games end up anyway?
Jonathan Cook
all i want is for the battle AI to not to charge me with infantry and flank me with cav
ALL THE FUCKING TIME.
Lucas Murphy
Fucking this. Easily the worst part of the game.
Jacob Foster
what if they raised the unit count and increased map size, then allowed largely coordinated multiplayer games, like a general to give order and like 15 people of lower rank managed the micro for all th units similar to how actual historical battles were carried out?
Liam Robinson
>A neural network solves an optimisation problem. Give me an opponent and I'll fuck him up. Real video game AI is about allowing the player to win, but hiding it.
well actually you can freeze the AIs state as it matures capturing it at less trained state than its more optomized versions and difficulty could be turned into a slider which would simply change the values of the nodes in the network to match longer and longer trained states
Chase Stewart
Could be cool desu
Ryder Brooks
Total War is not a historical simulation. Likewise, all the games that have tried any sort of more realistic hands-off approach failed horribly. Including CA's own Total War Arena.
Watch the campaign co-op gameplay for 3K. That's literally all it takes. You're playing a normal campaign, but whenever you get to fight anything, it's not the retarded AI but your buddy. Likewise, Shogun 2 was extremely good at pulling people into MP with the avatar mode. A simple progression system and a little rough around the edges, but it was successful as all fuck and was a large part of what made Shogun 2 so good (the AI being as retarded as allways, needing nothing more than a yari wall to win every single battle offline).
Kevin Cruz
i got the 2 bridges over a canyon map like 4 or 5 times in a row and just quit playing
Zachary Nguyen
Hahaha No gamer outside of ARMA autists respects chain of command. You don't need to search for examples far and wide, just look at Total War Arena. Heard mentality and "I will deal damage, you tank" tactics. are all you find.
Aaron Collins
That doesn't work because the AI does not just get better and shit. A variant of this could work, where you train XY separate AIs with a different amount of hidden layers involved. However, the training strategy for a smaller network does not work for a bigger one and vice versa. You'd essentially have to create ~3 burtally hard to create AIs all within the span of development of a single game.
Your epoch approach would just lead to the AI going full DERP at absolutely random moments or as a result of completely random actions/events.
Ian Davis
The problem is that they're too frequent. By mid game practically every other battle is a siege.
Matthew Lewis
Not at all, that screenshot alone looks like fortnite. Zoomers have ruined everything I tell you.
Gabriel Hall
>That doesn't work because the AI does not just get better and shit.
but it does, for example look up leela chess, its a deep neural network for chess and it would pause in its training after x time and play rated games and over time as it trained more its chess rating increases. so you can save its state after an arbitrary amount of time and label that a difficulty and the sensibly increase its training time with a difficulty slider
dont care what it looks like so long as it plays well and my controls are clear
Alexander Lopez
But Warhammer surpassed it in gameplay, performance, tactics, graphics, story.
but never AI
Jackson Sanders
>think it'll be any good? Yes. I like the fact that you can play either as "fantasy" mode or realism mode. I might go for the latter, as much as I think wushu duels are sweet.
Adrian Brown
>nuCA hell no
Noah Green
fantasy mode is the warhammer version whereas realism iis every other installment. id be mad if i had no choice in 1 character wiping out 1/2 an army
hopefully yes, Warhammer brought better gameplay and battles, but streamlined everything else
Matthew Johnson
That's highly specific to the network built for that game, then. You can read up the whole write-up of the developers working on Starcraft 2 AI on faggit.
Imagine for a second something like FPS AI. The optimal strategy is to look around yourself like crazy and the moment a player sprite enters your view angle, you go for precise headshots/hits depending on the rules of the game. That's the only real reliable way to play as a machine. You cross that bar and it's over.
Chess is a different game. You see everything, you can try and calculate however far you can get in whatever restriction's been put on you, etc.
Starcraft is different from both of them. So is Dota2, poker, those auto-driving cars,... If all neural networks behaved the same and you could tell exactly what to use to get it to behave as "it should" we'd be living in an utopia right now. But they absolutely do not. That's why even Google has to waste millions throwing random darts on the board for methods to get them to do anything worthwhile.
Carter Martin
>performance, tactics, graphics
kek
Bentley Edwards
Friendly reminder that CA has more employees than Rockstar North yet insists they're a small, indie, family-owned company when things aren't going their way but will proudly proclaim they're AAA when things are looking good. They falsely believe they can work on a dozen projects at once even though most of their staff are PR tranny diversity hires.
Sebastian Peterson
warhammer runs better, the monstrous unit types and magic are super varied in tactical options, and it looks better.
John Wood
pretty much any neural network based AI ive seen has shown logarithmic developmental progression.
>and it looks better The lighting they used for WH2 makes it look awful, Shogun 2 is objectively the nicest looking game they've made.
Xavier White
user...
Jeremiah Barnes
No, TW has been trash since Rome.
David Foster
havent played it but thrones of britannia looked best imo, heard its an awful experience though
Ryan Kelly
Attila, and by extension ToB, looks very realistic and probably has the 'best' graphics but it's very dull. Shogun 2 is the nicest to look at imo because it's so colorful and stylistic.
Angel Russell
That's the nature of optimisation problems.
The problem here is, AIs for video games like SC2 are searching a much, much bigger "area" with tons and tons of local minima. Once you detect they've gotten close to something, you examine it and judge it it's okay or if you need to throw a rock at them and get them searching elsewhere. This is built into the design of the network itself for the SC2 bot. And of course, the local minimum could be something like "I can perfectly defend against rushes but can't play the game if the opponent doesn't go for one." Making it worthless. The problem here is the opponent and unknowning factor. When you're trying to walk, your goal is perfectly defined. If you're playing chess, you can theoretically know exactly how the game will go for a while and then decide on some action. With Starcraft, it is much more complicated and the only weight they have is that the AI won the game in the end. They literally do not incorporate what it did during the game into the final learning score and their snapshot for the pro games was handpicked. If they let it train again, it could very well scre up in those same games.
Wyatt Evans
Friendly reminder that you're pulling things out your ass
Joshua Powell
The unit count actually got smaller over time I remember 10k armies in rome being easily achievable
My biggest problem with three kingdoms is that I believe CA should finally stop trying to make games with Top graphics that 10% of the playerbase can max out and start working on how to make the battles absolutely massive without crashing everyone's PCs
Owen Baker
i think the problem isnt crashing PCs from unit numbers but the inability to micro 400 regiments
maybe if they had an AI assist to control battles but outside that youd need a lot of people playing together
Joseph Adams
The AI in Planetary Annihilation is actually good, one of the few things about the game that they did properly.
Logan Russell
Sorry, I meant Unit size Rome and Medieval have the same unit count as Three Kingdom and Rome 2 yet they have double the unit size Warhammer is the worst offender since you can have an entire army.... comprised of 1000 Atleast warhammer have justification since its fantasy
Nolan Scott
Auto-resolve being so godly from Rome 2 onward also didnt help to make the games feel shallower
Cooper Kelly
I was extremely skeptical at first, mostly because they focused on the Romance mode for all of their early marketing, which wasn't going to be my favorite game mode anyway, but the more they've been revealing the more I've become convinced that it's going to be a great game.
It's definitely not going to be perfect and there's a 100% chance that it'll suffer from the same issues as most TW games (the first and foremost being shit AI), but I think it stands a good chance of being Shogun 2 tier good, or even better.
Jason Rodriguez
In the newer games you can have 2 stacks on the field at once for 40units under your control and 40 for the enemy. I'm pretty sure you couldn't do that in Rome and Medieval but I might be wrong
Nathan Watson
>CA [...] insists they're a small, indie, family-owned company Proofs?
Joseph Morales
You could only control one stack But in the older games you could have more armies on the field, in the newer games if you have more than 40 units they will trickle into the Battlefield once a "slot" opens
Aaron Carter
40 units being under your control definitely is too much They should revert to pre-rome 2 reinforcements where the reinforcing stack would be AI controlled Unless you were a god of micro , the AI couldnt possibly do a worse job than you with the extra units
Gavin Jenkins
Why? 40 units aren't that hard to control especially because like Says, the only thing the AI does ever is March their infantry at you and send cav to flanks. I can't count how many times I've moved spears to the flanks and just had the enemy cav run the entire length of my front line getting murdered by archer fire.
Easton Young
40 isn't that hard to control. The biggest problem with it is the stupid unit card order jumping around. If the shit would just stay in the order you put it then even if the extra units trickle on it isn't a big problem.
Which also reminds me that it is really annoying having to reorganise the unit cards at the start of every battle. Would be nice if they had the ability to keep the organisation/group in place unless changed.
John Diaz
>performance, tactics, graphics Fantasykeks are retarded and this comes from someone who's second favorite Total War is Warhammer II (Mortal Empires)