The BRUTAL TRUTH About Total War's Warscape Engine & AI 

Obviously its not a very clear cut area. Some campaigns you have decent AI, some you don't. But throughout the series we've consistently seen the Warscape engine's limitations in battles and the AI's inability to be responsive enough to the player in a thematic, immersive and strategic way. I hope Warhammer 3 AI gets fixed soon because it's in particularly a not so great place, and I hope Pharaoh sees some improvement in the series too. And I'm keeping my fingers crossed CA finally develops a new game with a brand new engine that does all the bread and butter battles of Total War justice.

I think the old games hold up so well in part because the battle ai is essentially the same ai used in the new ones.

The extra campaign options to me honestly screams "We can't get a proper non cheaty balance working, so here you go, get all this options and make your own difficulty"

Totally agree! After all these years the AI still can't really figure out a basic flanking maneuver. I know crazy leaps and bounds like having an ambush on a battle map is a bit much, but some kind of strategy other than walk forward would be appreciated.

Glad you mentioned the troy ai, watching troy ignore my allies taking them over so they could keep sailing across the map to attack me killed the immersion

In more than a decade of me playing Total Wars only one I witnessed a commendable AI ambush move. It was in Rome 2 with DEI mod and I was playing as Macedon. My two armies captured the Noricum city over in Raetia et Noricum. I saw one small enemy army near the city and I moved out one of my armies to engage it but it retreated deep into the forest and when I pushed forward again, a full stack army with descent experience was waiting in ambush. That was the one and only time I can remember an excellent AI move. I just wish I could see something like this in Warhammer or in Attila. I don't mind the difficulty but make it look more real rather then just senseless AI buffs and cheats.

There will never be Total War game on the level of Rome and Medieval 2 from CA anymore. I made my peace with that. Thankfully modding scene is alive. Hopefully, one CA crash and bankrupt, some other company will rise from ashes and return to good old days of Total War.

You could add: sending in their forces piecemeal rather than all at once, leaving artillery or ranged units completely vulnerable to cavalry (or flying units in WH), and even when they are the defending army advancing to the player rather than sitting back and making the player attack them

I agree, definitely would like to see the ai switch tactics. As well as actually keeping formations during battles, it sucks to see when the ai has a huge line of phalanx that breaks into 3 different directions when your troops get close. It’s like they use them for skirmishing rather than keeping the battle line firm

This I found to be one of your most interesting videos. I really agreed with you on the subject of unit formations. Blobbing is the biggest problem in with the AI in my opinion. Units in the ancient and medieval time periods fought in formations. Lines, phalanxes, maniples, shield walls, "boar's head", wedges, squares, etc. If these formations were broken, then things tended to go downhill very quickly. The AI in Total War games seems to get its ideas from Hollywood movies where charges and melees immediately devolve into hundreds of individual combats rather than actual historical accounts of battles where units battled units.

Late game AI really struggles in games from rome 2 and onwards. The game breaker I found is how all factions begin to turn on the late game player because of territory size. Even allies and vassals will eventually turn on the player which locks the campaign into non stop wars. Some turns I had 3-4 different battles to fight every time I pressed "end turn".

This has totally made me drop tw all together, paradox games are much more in depth, you don't get to play battles but as a grand campaign competitor is far far far superior. They have troubles getting a simple satrapy function work imagine having multi-nation faction coalitions, proper puppet states, demilitarised zones, border guarantees, etc etc... and that's just the campaign diplomatic options

I hate this engine. It doesn't work for close combat, which is the vast majority of battle encounters, and those 'battle animations' lose their charm after a few times. Then they are just annoying. And I haven't even talked about unit cohesion on the battlefield. Outmaneuvering the opponent is pointless, and every clash of regiments turns them into giant blobs.

It seems like a lot of CA's AI improvements are actually just them removing game mechanics the AI can't understand. For example, Shogun 2 let units climb into forts without needing siege equipment. Armies now provide their own transport ships since the AI was terminally incapable of performing naval invasions the normal way. And in all modern TW titles each faction is limited to a few armies led by a general, which forces units to be concentrated under a decent leader instead of making 50 individual units wander aimlessly every turn. To their credit these changes do work. Shogun 2 had the best sieges. Naval invasions happen now. The AI has concentrated armies that fight big battles and they can't waste 5 minutes per turn shuffling their units around anymore. Nonetheless, simplifying your game because your AI is too dumb to actually play it strikes me as a bit backwards.

You articulated my major complaint with the CA series, and why I haven't purchased a CA title in about 4 years.

I hope CA is starting to train ChatGPT AI on what players are looking for in these types of games. They should just expose GPT to multiplayer battles to study how humans play with each other.

Empires. Watch enemy cavalry charge diagonally across my entire line of musket men, heading towards my cannon, only to stop, hesitate, get shot again, and flee. Or watch the enemy send half their army to each flank, with none remaining in the centre... and then have each half of their army arrive at a different time, easily allowing you to massacre them. Empires is so large, and the battles are all the same, that taking over the world becomes a really, really solid chore.

The only game that i can say that both campaign and battle ai work well is 3k. The campaign ai work like you expect.CaoCao and Yuan Shu, Sun Ce will always scheming and backstab you whenever they can, LiuBei, TaoQian, KongRong will stand by your side in time of need. And LuBu will most definitely quack you in the back( now that surprise😂).I have a blast playing like a cunning leader, making friends and grabbing land without fighting. Always find myself coming back to 3k for in depth diplomacy system

I'm telling you, CA is gonna stick with the current iteration of the Warscape engine until the inevitable launch of a Warhammer 40k Total War gamein a few years. They're gonna straight up try to make tanks, cover, squad tactics and nukes work with this engine before they develop a new one from scratch.

So true, this is why I mainly play Multiplayer, the games have good mechanics and players actually use them, they have real strategies and aren't easily predictable. Ive played a bunch of Total war games and all of them have bad AI, that's why I don't play campaign. Like your having a huge siege with 20 stack armies, your fighting on a cool looking map but then the AI just rushes all their troops into the gate and it just becomes a waiting game, since all you have left to do is sit and watch your units massacre the AI. Recently I decided to download Divide and Conquer and I have barely played it, I expected to be fighting for my life as gondor, instead when mordor attacked osgiliath, they just rushed my one gate and didn't even react to getting shot in the side by my archers. The same happened on cair andros.

Oh a big big issue i have with autoresolving any battle is damage to literally every unit i have, even if the balance is far superior in my favour. I have lost many of these too, even with thousands of extra troops too. I refuse to do it in any battle now which is just tedious.

Again you hit the nerve that i feel longtime fans feel about the new games. Great video once again!

This was a really interesting topic to talk about. Completely agree. 3K is indeed the best total war game when it comes to both campaign and battle AI. The bad thing is that both warhammer 3 and upcoming Pharaoh share 0 similarity with it. Warhammer series, while being hugely profitable, is holding the series back. Each game of this trilogy is built on another with content compatibility in mind. Therefore, the core they share is the same one intoduced in 2016 with WH 1. Pharaoh a.k.a Troy 1.5 is also child of warhammer 2. As a result, the series is stagnating. I am happy that wh 3 is the end of the trilogy and CA can truly move on.

It's weird that computers have been playing chess, which has (apparently) more piece placement combinations than stars in the sky, for decades - and yet CA can't make cavalry that doesn't charge a spear wall, or an army comp that isn't 60% skirmishers!

I would argue the AI has come a long way to, but its also one of the points that have advanced slowly. I remember the line formations that would push and retreat in Medieval 2 and Rome for them to run back allowing you to charge in the enemy soldiers backs. I remember the extremly poor decision making in older games compared to more modern ones. Building empires without scripting them would rarely happen. Will it be perfect... probably not. Coding AI is the most difficult thing game developers do today. Getting it to comprehend and make correct moves between a million or more combinations and also long term strategizing is a challenge even with millions of dollars in budget. Even our cutting advanced AI are still wonky at best. I think it have made some progress. Especially the decisionmaking have seen many improvements and the long term gameplay have seen some betterment. I hope they keep at working on the AI maybe we will be able to have one that is fun to fight and also clever. Its easy to make a computer overpowered but making it play toe-to-toe against a human with no buffs or bonuses will probably not be possible for a long time. Its an industry problem that will not see any quick fixes.

For the warhammer series i think any AI is doomed because there are just too many different systems to be able to actually compete with an actual human player. Historical games (expecially in more recent games with less unit variety) should be able to be more competent.

I wish they collected all the multiplayer battles of the games and trained an AI with it. Making a sensible AI is very difficult, but they have a lot of tools to make it better than it is right now! Hardware is a real limitation though.

Theyre not going to change the formula, or update Warhammer 3's ai. Its so successful its a cashcow on its own, not even including the rest of the total war series. The only game companies that care about quality are indie deva, the rest experiment with dlc, pre purchase dlc, day 1 dlc, microtransations, always online drm, mobile spin offs- all things CA has engaged in. Products are made of a minimum acceptable to market quality, and nothing more

It is in Rome 1, Med 2, Shogun 2 and 3k that the AI actually poses a threat. _ In Rome 1, it's almost impossible to beat Rome with autoresolve only. It's doable when I tried as Carthaginians and Gauls. As Greeks, my population would be bled dry and as Seleucid, I can barely fare against an omni-directional assault. _ In Medieval 2, I onced tried to charge my heavy cav general into a Spear Militia. Normally, a knight can charge straight and break infantry like tanks. My general got a few kills before being unhorsed. _ Shogun 2 really depicts the true nature of being a dictatorial personnel and vassals. They're very backstabbing. _ In 3K, I played as Huang Shao and got ganged up by Yuan Shao, Liu Bei, Han Fu, Tao Qian, Kong Rong and Liu Dai. The latter 3 was just opportunistic but that's about it. Yuan Shao can just throw stacks of captains and spear units from thin air like Zerg and Liu Bei's starting stack is just decimating. And because I'm playing Yellow Turban, the morale of just tier 1&2 unit is even worse than their's. So I was forced to commence the first 10 autoresolve battles to gain breathing room.

I cannot wait for chatgpt level AI in Strategy games. Imagine trying to negotiate with chatgpt roleplaying as napoleon and calling him a frog-eater, resulting in a never ending war.

I wonder why they didn't apply the knowledge used for the xenomorph AI in Alien Isolation to the TW titles. I know the gameplay is very different, but surely they could've at least made something work. To this day I still hear wonders about the xenomorph AI that it's easy to forget it was also made by CA.

The main thing TW needs is competition. RTS is almost entirely dead and the kind of RTS TW has been doing for decades is essentially unique. I think it does speak to how difficult what they've done is that nobody is trying to take advantage of their increasing lack of engine innovation.

You give the example of shogun 2 and three kingdoms as having better AI and i think that is to do with the setting as almost ALL the factions in those games have the same units. I cannot be a coincidence. Why the variety of units would make that difference is what CA should be discussing with the community.

Medieval 2, and all its mods was broken for me when I realised you can take a full stack city with one spear unit and two scout cavalry. Lay seige, they will sally out next turn, as they have overwhelming odds. Run away to the far corner, they will chase you. Leave the spear unit to die, and whilst the enemy catches them, run your cavalry to the unprotected city. The gates are unlocked if the enemy are all outside, which they will be, - and you start the capture countdown. Leave one cavalry to block the gates if you need to. Works every time.

To be honest, one of the things that I noticed immediately when playing WH3 shortly after it came out as was how part of its campaign AI clearly felt like it had been ported over from 3K rather than WH2 in terms of how it reacted to sieges and minor settlements. Where it would relentlessly target the minor ones but would almost completely ignore the major ones until it had a massively overwhelming force to bring.

The problem lies with the algorithm of the AI, not so much the engine itself. They need skilled programmers to actually create a better AI. I never understood why they never continued building from Shogun 2 AI. While of course not perfect (and maybe average at best), but one of the best of the Warscape engine games. My feeling is that they just do not want to invest the resources (as usual) and just copy/paste games over and over again (Give me the money!). Or they just have incompatible programmers (uhm starting from R2TW). Or maybe it's just both.

Aside from a few improvements in diplomacy, I don't really feel the campaign AI has improved much at all, I think the automatic garrisons are about the only thing covering for how stupid they can be at times, even then they don't protect their territory, their armies run away, annoys the hell out of me when you literally laying siege to their capital and they still refuse to engage your forces most of the time.

We had atmost 40 unit to commend in Rome2. We still have atmost 40 unit to commend in WH3, after all these years.

Spot on... tbh im mostly sick of fantasy games. I can deal with poor ai if the game im playing is historical. Id take an Empire 2 over ANY fantasy game, ever.

The Terminator calling for improved Artificial Intelligence. Checks out.

Game engine could also use a unit cap update. I want historically sized battles from antiquity.

The funny thing is in Warhammer 3 the modder Hecleas fixed a lot of the AI until the 15.0 patch which broke everything for it to the point that Hecleas says he can't work on it till CA fixes it.

Calling CA out on their shit is the one thing we can do, I approve this message

The reality is the complexity vs. profit ratio might be unbalanced for games like Total war and football simulations where incredibly complex dynamics are needing to be simulated at increasingly realistic levels while maintaining profits. Total war was so attractive because it combined being a city planner, with being a despot, with being a general, with playing an RPG, often over an entire continent, with battle maps representing actual campaign terrain realistic to real planet earth nature. My mind was blown the first time I play medieval and realized the mountain in the background on campaign showed in the battle, or the building I just built showed up in the city I was fighting in. The fact I left my general in the settlement and forgot about him made him lazy and his unit moved slower in combat. Mind blown. As time has gone on tho, expectations of graphical fidelity, battle simulation, AI complexity have also increased. The cost of redoing an engine that meets the expectations of the gamers is just so cost prohibitive in 2023. There just isn't a genre that even attempts to meet this same demand of systems all needing to use resources, all needing balanced, all needing realism, accuracy, and graphical fidelity. Total war players are spoiled by prior games, all-time great games and they now demand the same wow, marvel, amazement that they experienced years ago in games that, if released today, would be hated and maligned. Of course they could spend 5 years on a new engine, new modding tools, and charge the same price but 2 million are still going to buy crappy pharoah, and 2 million will buy the next crap one, and the one after and the one after and the 5 releases in 5 years will make more money than the masterpiece released after 5 years. The player base is too loyal but also too limited to justify an overhaul.

In the end it was irrelevant because almost no one bought Pharaoh. But yeah, the battle ai was always predictable. Without the ability to pull rabbits out of it's hat tactically, the battles turn into a chore.

one of the problems with Total War: Warhammer is the insane variation between units, all of the stats and abilities and items and spells. That is such a complex interplay that no wonder the AI, weak as it is, cannot cope.

With CA it's clear their AI changes are one step forward and two backwards. I think they need a new engine and to go back to the drawing board. Them investing in hyena's instead of balls to the wall on a new engine was a mistake.

One thing I don't understand that CA have never fixed is the hide in woods mechanic. The ai can see you moving a detachment of cavalry into the woods. They become hidden and they simply cease to exist in the AI's calculations. What do you mean there are cavalry over there, don't be ridiculous. Seconds later a full rear charge, battle of the pelenor fields moment happens and the AI knows it fudged up. All it has to do is keep a piece of data in memory called last known position and either send a unit to investigate or keep some defensive units ready to defend that direction.

If CA could use machine learning with its AI that might be a game-changer. Imagine if they could just get the AI to play against itself and learn the successful strategies, and maybe then they could be given realistic, human proclivities. Some might be loyal, some insular, some megalomaniacal. They have the money, they should do it. Hell they could even implement AI text to generate interesting diplomacy, instead of a simple set of options. You could "talk" with the AI like you could in a MP game.

I will say the 3K CaoCao non battle campaign AI is very good. Whoever made that should be proud. Just take CaoCao and apply it across the board - battles are a joke at this point. Heavy spear / full siege lineup and horses round the Back on archers and then into back line. Rinse and repeat.

I distinctly remember playing a game of Lizardmen in Warhammer 3, having so much fun with my armies of dinosaurs. Then I go against some vampire counts, loading into the battle, then hit with a line of zombies/skeletons that span the entire map, and that also refuse to move forward. So I have to march up there with my dinosaurs and kill them. Now I'd win the fight, but god it was so boring I just quit half way through. This happens often with other factions, such as Skaven. Where instead of creating intricate or interesting formations, they just create giant lines of enemies that don't engage. Even the Dwarves, who are great at defence and SHOULDN'T move forward, don't turtle. Instead they also create a giant line and shoot into their teammates with their artillery, if they even have it. They should definitely work on making the AI interesting and innovative, instead of making giant lines of enemies so every battle just plays out the same. Edit: The most fun I had in a battle was when I was getting ambushed by the Vampire Counts as a Dwarf faction. Zombies and Skeletons coming out of the trees to attack my army as it rapidly tried to reform. Winning the battle by having some units stall, as my artillery set up on low ground to fire on the steep-ish hill the undead were pouring down from as my units fell back to defensive positions. It was the most fun I had in a Total War battle ever. Even if the AI was the same, it was a rare moment where the AI did something that the brainless hordes of undead would actually do. It felt very in character and I had a lot of fun doing it.

Nice timing with the recent Warhammer 3 update making the AI noticeably worse lmao You're right. It's one of the most core features of a game like this, and their refusal to touch it is superbly frustrating. Recent CA behavior is a grim portent of the franchise's future.

I'm really hoping for Empire 2. They should create a new engine and ensure it's moddable. The primary focus should be on improving AI in both the campaign and battles. It should be adaptable, opportunistic, and extremely challenging on the highest difficulty settings. The game should encompass the entire Americas, Europe up to India, and extend into a portion of Russia and Asia. It should feature realistic bayonet charges and well-organized line infantry formations. City sieges should have unique maps, and forts should offer a variety of layouts. Diplomacy should be as advanced as in Three Kingdoms, with the same kind of flavour found in the older Rome I. Include features like commander speeches, city visits during peacetime, and various agents like diplomats, merchants, and spies. This would truly be the ultimate Total War game, fulfilling the dreams of many fans.

I think the reason why the battle AI in Shogun and 3K are so good, or atleast seem so good is because the sieges were simple and the battles were simple, if you know what I'm getting at. And for the campaing I think it was the faction personalities that made them work, like you know what Takeda is gonna do, fight Uesugi and Oda gonna blob, like in 3K you know Yuan Shao is gonna do unless the player (you) pick a different path for the story, deny a marriage or change how the event goes on. Idk, the new campaing AI just seems to be confused on wtf to do with all the new systems and multilayer siege maps and stuff, keep it simple and yeah, let the player have more options.

Great video - thanks for covering this and I agree with your points. All these mods look great, but poor AI ultimately kills my urge to go back and play most titles

I hate when difficulty differences are boosting the AI and bending the rules. That doesn't make it strategically more rewarding, just forces you to use more "meta" builds and cheeses. you aren't being challenged by an opponent who makes better moves based on the circumstances, just by non-sensical rules. in fact, you are punished for thinking strategically, because strategies that could actually be clever ways to outmaneuver your opponent are useless because the enemy is able to unrealistically churn out 3 new armies (which they won't use in an innovative way) even when they are wrecked.

Preach bro. I loved the series up until Medieval II but once you get fairly decent battle AI has got nothing to offer thus single player mode gets repetitive and pointless.

Play many RTS/Strategy games, AI is always the issues in most of them, even for Paradox games. Getting a good adaptive AI will always be a great challenge for any RTS/Strategy devs

when I play as Rome in Rome 2, I liberated the 3 cities on Sicily. Carthage just decides to sail around my 3 military allies and go straight to the mainland Italy. I just wanted 1 campaign where I didn't have a 150 year war against Carthage

Melee combat has gone backwards in the series. Rome and Medieval 2 being the pinicle. Newer games are blob attacks blob, with attack and defend animations for individual combat. The units don't get forced back and thier lines don't bow. Go back to Rome and ME2, the weaker unit will get forced backwards. If they're in a thin line the line could break in 2. Bridge battles could be won by sending a strong unit to litterally push the opposing unit off the bridge. In Rome 2 and beyond, they'd stand in lines while individuals would step out and have individual combat. The Campaign AI might not have changed, but when melee combat is significantly worse surely that has to be the biggest issue. People are calling for the release of ME3 next, it would be an absolute travesty if they introduced a melee combat prominant series with the current combat mechanics.

Long-time AI modder here: I'm a bit late, but the CAI (Campaign AI) is fully moddable and has been for 15 years. With even a basic understanding of how the personality profiles and the TMS (Task Management System) works, the CAI can be made to play competently, up to and surpassing the point where just zeroing out the AI economic and battle buffs is still incredibly unfair for players (i.e. the AI must be given the same penalties as the player to level the playing field). The problem on the CAI side of things is that the systems involved *look* incredibly complex, even though they are only incredibly tedious to work with due to how segmented they are. The *real* problem from a modding perspective is that the BAI (Battle AI) is all but untouchable. What can be optimized or otherwise altered involves eating up the most commonly used tables, presenting a further compatibility problem. What's worse, those changes ultimately do little about the overall ineffectiveness of the BAI in a real environment. When one considers that the BAI has been noticeably lobotomized half a dozen times in as many years (sometimes severely, as is the case in both WH3-era lobotomies), it's little wonder that the main modding feature I - and others - scream for is honest-to-God under-the-hood access to the BAI.

fall of the samurai naval AI was aweful, all they did was hiding in the maps corner....

I have only played Warhammer 1 and I find it to be a very well made game. In the campaign it often uses "march" to get out of trouble and "ambush" to appear or even disappear before your eyes. Chaos armies do this and also they gang up together and are hard to defeat. Yesterday I was Bretonnia next to a Norsca city and an Empire army invaded the city so I played the battle as reinforcements to the computer haha (I was the "reinforcements have arrived" army haha). After about 200 hours in the game, this entertained me quite a bit. I know the campaign can be boring, but generally not too much, it's Ok and I have played 4 so far (Empire, Bretonnia, Orcs (failed), Bretonnia).

I downloaded the Empire 2 mod thanks to your video and I gotta say, a minor complaint I have is sometimes the absolute sheer amount of unit cards on my main capital. Is there a way to cycle through these cards beside your mouse ?

Great vid Termi! Unfortunately AI is a huge problem in many genres, shooters being another evident example of non-responsive or super-powered AI... TW does need huge improvement in this sector. Hope CA takes the iniative in upcoming games. Keep it up boss.

Yesterday I booted up the new patch for Total Warhammer, and even tho the AI is still not great, I saw for the first time ever a unit of enemy infantry maneuver around my flank and hit the rear. Some blobbing still happens for sure, but it's a step in the right direction. I'm intrigued to test out different difficulty settings now.

Shogun 2 is the only modern Total War game where I play battles. They work and are enjoyable. All the rest since Empire suck and I opt always to autoresolve them. It honestly feels to me that Shogun 2 had whole different team making it. They made the melee work perfectly almost with the engine known for only working fine in ranged combat. They mostly nailed the naval combat as well in FOTS. The campaign AI is challenging as well and doesn't starve or lose units to attrition too much. It knows when to naval invade and when not to. Oh and the sieges are enjoyable in Shogun 2 too. No unit blobbing and single gate fights. It's all around attacks. Ofc you can exploit the AI with ranged units in sieges, but why do that and then complain?

"since it's difficult to make a good product, we'd rather cover a pile of turd with cake dressing !" is the META of gaming And then high quaility games makes buzz and oversell like crazy (Elden Ring, BG3 to name a few)

Unit collision should affect how armies fight in formation. Cant believe they haven't figured it out yet

Truth has been told. I even prefer brain dead AI than anti-biased one. I played almost all Total Wars, thousands of campaigns, but only to the point when literally every faction around starts to hate and attack me. Yea, usually am kinda great at that point, but whole freaking world starts to fight me, former ally, enemy, the trade partner on the other side of map, doesn't matter. At that point game turns into battle simulator, diplomacy doesn't matter anymore, cuz AI will never ever take peace proposal or any other deal. Only way to play from that point is to turn everything into war and unwanted conquer of world, because enemies will fight to the last city and never stop spawning armies. It quickly gets boring or tiring with all this harrasment, no fun left. The only exception is WH3, when playing good faction, even without end game scenario, alliances lasts to the end, but with more evil faction its the same story.

It's interesting how much the devs can do with the setting to make up for the lack of depth of the AI. For example, Empire was a cakewalk, Medieval I was almost painfully easy once you got rolling. Conversely, vanilla Rome I was a chore, Rome II's civil wars got pretty excessive, and the tight, packed maps of the Shogun games really proved a struggle for me. I try to use proper tactics as much as possible, but I generally struggle to win when badly outnumbered. And in Shogun, the resource cheating by the AI combined with the few "open" provinces at the start makes it a challenge. I only played Warhammer I, but I found that very challenging as well - again, not due to the AI, but due to the setting. For me the peak of TW is still the Europa Barbarorum mod for Rome I. It fixed CA's crude cursory historical references and names, added provinces to make a much larger map, populated it with tons of factions while still leaving large chunks of the map open at the start, added and rebalanced some units.

Yep, easily the worst part of Total War games is the AI. After a hundred hours in a title, I can't help but KNOW what the AI will do in almost any given circumstance; turning up the difficulty doesn't give me a more satisfying challenge, it makes me more aware of the AI's failings as they do the same thing with buffed troops. Any actual "improvements" are things like dodging artillery shots instantly without fail, which again snaps me out of the immersive strategy, back into 'how do I cheese my way to victory here'. It's a real shame too, because it often hamstrings any mods developed on top of that AI; campaigns in amazing mods like Divide Et Impera, or SFO Grimhammer get stale because even with new mechanics, units, depth, I'm still working towards having a better army that I still feel like autoresolving most of the battles I fight with.

AI and games did a great video about total war ai. Campaign ai has actually improved a lot. Empire had a big jump but it came with a downside of taking ages to compute the more complex strategy map ai, which only really got fixed in the Warhammer 2 update that decreased end turn loading times. However battle ai has used the same fundamental system since Rome 1. It has gotten stuff added to it ofc, but tis still using basic decision trees and they honestly need a scrapping and redone ai system instead of continuing to add on top of the 2 decade old codebase.

So true! Good example for AI problems, correct me if I'm wrong but u can bait any Ai army to u if u have at least 1 Cata, Trebuchets, Canon... unit even if they have the better battle ground in all total war games Good video + can someone tell me which mod this is in 0:52 ?

One total war i played where the AI made semi- decent moves was napoleon. The AI would usually try to encircle me and would react to my attempts to flank their units. They didnt really pull off any amazing strategies but they were not terrible.

I think Total War AI peaked in Shogun 2 and has been downhill ever since. That goes for basically every other feature too. It all peaked in Shogun 2. I don't think any Total War game since Shogun 2 has had enjoyable sieges. You either have to bumrush the fort/city or cheese the AI by sitting out of range and just bombarding them with some sort of massed artillery, then walk in when you've killed 90% of their troops. Every game since Shogun 2 has had infuriating and totally un-fun campaign map AI, where enemy factions always stay 1 movement point ahead of your own armies and do nothing but hit your undefended cities from the sea, then retreat. This is even more infuriating when factions that shouldn't even be able to have a navy - LANDLOCKED BARBARIANS! - just walk their army into the sea and spawn an instant magic navy, then sail like they were born to the water, covering half the map in one single turn. Every game since Shogun 2 has had boring generals, boring tech trees and boring agents.

Lets not forget to mention that unit commands have been limited to right clicking and that is it. Warhammer has no unit abilities and it makes lackluster

Ai is the most import thing in total war games imo. The ai feels so dumb that doesnt challenge me at all to play this game

hate it when ai just straight-up cheat

I wish they would implement contemporary machine learning models to their in-game AI. They could train the models with all the multiplayer games going or leave the AI to learn by itself like it has learned GO and Chess. Sure, TW games are more mechanically complex than either of the classic board games, but I think this is the way forward for computer strategy games because having people manually program increasingly complex "if-else" statements is simply unfeasible. Let the computer do the heavy work and just polish the rough edges, add difficulty curves, etc. The tech is probably out there, we just need someone brave enough to implement it.

Alright, never seen your stuff before but the title caught me. Let's see what this channel is.

Totally true. Why aren't there huge differences in diverse generals of opponents? Some aggressive, some defence, some very sneaky springing traps? Simple, logical and raises the need to study what general stands opposite you.

It's clunky. The bit that really grinds my metal about the clunkiness is the soldiers all performing the exact same action at the exact same time, and it's such an immersion extraction. I hate to say that the film "Troy" [not the total war game] did something well. However, there is a scene where the Mycaenians charge straight into the Trojan lines. When you see it, you can see the absolute chaos lines clashing into each other. Compare that to total war combat with the current engine. The units are just big blocks, melding into one another with the exact same actions and movement. Rant over

A bit late to the debate but I actually thought about your sentence at 8:39 "a balanced AI that can adapt to a variety of situations". I'm only a TWW3 player but I feel that if the AI is still bad after so many iterations, why not take the opposite approach: make an AI that uses a variety of situations where the player should adapt. As an example: I don't need all the dwarves lords to have the exact same army. Create a database with 20-30 viable army comps and let the AI pick from it at random. Then during fighting instead of the established long line of infantry, followed by archers and artillery, create another database with various formations. You just need to open youtube to find tons of formations which are good to some extent (chess board, 3 squares, etc). It doesn't have to be particularly efficient, it needs to bring variety. Variety from the races played isn't enough to keep me entertained after 100 turns. I'm perfectly fine with battles that are more like solving a puzzle than fighting another player, I can just launch an online campaign if I wanted that.

The brutal truth of Total War is that its gone in the wrong direction focused on the wrong features and made it simpler so broden the audience. Now its gone the same route as COD, same shit different skin.

one of the things that pisses me off when u need to click 20000 times for the units to move away because one enemy unit model has touched my unit. While the ai just ram across my units and ignore them.

The biggest improvement I saw is in Warhammer 3. I just downloaded Rome and Sparta mod for it.

In my opinion, the issue of CA's complete unwillingness or inability to improve the AI is the number one reason for the features that have been cut from the game. They basically have to dumb down the game because their AI is too stupid to handle the systems they made. Before rome 2, units were not locked to generals, and the game didn't impose hard limits (like rome 2 and attila's imperium) or soft limits (total warhammer supply lines penalty) to how many separate stacks of units could operate at one time. This allowed for far more flexibility (detach cavalry to catch the enemy and have the rest of the army reinforce, customize your own garrison for each town, trade units between armies without two armies needing to directly meet, etc). The number one excuse I see from supporters of this change is that this feature would often lead the AI to create numerous tiny and annoying armies instead of consolidating their forces to provide a big battle. But instead of actually fixing the AI and letting players have more tactical choices, CA just chose to remove the feature so the poor little AI wouldn't get confused. Unlimited building slots? AI can't manage their towns, so all towns now have arbitrary building slot limitations. Naval battles? AI can't handle it, so why bother with it? Siege battles? AI can't handle it, let's just give every unit magical ladders they pull out of their ass and limit the siege to a single sided wall segment (to be fair this is more TWH1 and 2 specific, but the sieges in 3 were also considered godawful) And some of these decisions lead to a cascade of failures. Units are now locked to generals -> you can't use cavalry or faster units to catch fleeing enemies unless you have 0 slow units in your army -> enemies are leading players through a benny-hill chase sequence -> agents that can slow enemies are now mandatory and you managed to recreate the initial problem of annoying weak enemies that run from the players and harass them. Except it's even worse because now the player is locked into garbage settlement defenders given by the buildings or spending money and limited buildings slots on a garrison building, so while the AI army won't fight the player army, they can easily take settlements... I'm not going to pretend like improving the AI isn't a monumentally challenging task, but at a certain point it feels like it'd be far better to actually deal with the fundamental underlying issue of the AI rather than taking a hacksaw to your feature list until your game can be played by the braindead AI.

I’m glad you brought this topic up. It’s really bad and it’s their biggest problem going forward

No cap 💯 The battles in Troy total war feel so shallow for some reason. The troop movement, the neat rectangle formations. It all bothers me. The way your unit would move in Medieval 2, while clunky at times, the unit would move in a more organic way. They wouldnt all move in a perfect rectangle. And the weird cone firing ranges for missiles units is so dumb. They could do what bannerlord does where they will attack behind them, just like missile cav already does it Total War. Sorry, I'm just passionately upset with CA 😂

If we were to switch to multiplayer to play a challenging enough campaign, I'd prefer a mode with an already built up map say tier 3 settlements, lords ranked up a bit, an economy that can afford a few armies. Each faction already vassalized or confederated with many of the sub factions. If I start a multiplayer campaign I won't play with the same person long enough for it to get interesting. And with them upping the graphics the loading screen takes longer and lag gets worse, making multiplayer less attractive to play.

TBH totalwar needs to give some cuts to Darth mod and afew other modders which make their game playable

I wonder if armchair session logs can teach the Ai how to play. a record of all the movement of each faction and a score system is needed to develop a good Ai. it is more or less a chess game where pawns are either sacrificed or eliminating the other pieces on the board.

@TheTerminator I am so glad to see more and more creators starting to come back to covering this crusial subject as it seems like most of the heavy criticisms of the past gone silent as most of the criticizers left the Game franchise behind, with new comers like the WH folk just mainly focused on visuals and more units... Unfortunately, we got to this situation because initially we had expectations of improvement from CA, and by the time we released they aren't committed to improvements (for many Napoelon or Rome 2 were the wake up calls) it was to late cause CA figured that the best way to keep making more money is to ignore the issue and just find new crowds who don't care and pedel them the AI less game. Now, its not that there was no improvements over the years, but for the amount of the new games, for the prices they cost, for the amount of supposed support CA were promising to put into them ... the AI stayed rather unchanged. Even on the campaign side things are look better than they are, CA focuses on finding as many shortcuts as possible instead of making the AI 'understand' systems how to use them and how to counter them, there is infinite examples of those shortcuts that really ruined the players experience especially for the players who play into it. Cheats that negated any thought of using certain features were common and clear to see but what people were less keen eyed to see is the fact that till today in WH series many the AI isn't really playing the factions, at best its using scripts, at worst it doesn't do anything which means that as a player you don't experience the faction you often paid for good cash to experience (as and against). Now credit is where credit is due, CA made a huge improvement with 3K's diplomancy not only at the UI front but also at the actual AI implamented, while on release the AI would often sell its own farms for food, after a decent amount of work I found the 3K's AI to be the most enjoyable to play with as a lot of the options weren't just gimmiks that don't function but actually had the AI weight and react to them as proper negotiations occured for the first time in a TW game. AI finally started to consider gameplay aspects like recent loses which meant beating a faction to a pulp often impacted the results diplomacy beyond just blank 'strength check'. Those improvements didn't come only from good implamentation and execution of the AI but rather from good design which leads me to.... Battle AI is so damanged not only because of the lack of investment which is clear to be the case (the company more than doubled in size since the days of Rome 2[if I recount correctly] having 900~ employees yet the quality of the products at best stayed the same (any improvements, such as graphical one is purely based on the latest iteration[) but more so due to the problem with design, CA isn't just masking their problems for easy sales with graphics they are doing it with selling a slew of content, mechanics, and systems that the AI isn't developed to 'play' with. I have explained it a million times but beyond poor implamentations of systems, like introducing the ability to spawn units anywhere on a map, which is one of the worst features they could have added in my opinion - the mechanic can barely be handled well by a player vs player scenario (good luck keeping your eyes on that artillery piece when you know your opponent have 10 vermentides it can use any time and he just is waiting for you to get distracted for a second - no tactics, just CA provided cheats), then surely an AI will be shit using it to but even in that the AI is really preforms poorly as the AI just spams the units on predictable targets without cosidering to use them as distaction in shape or manner... So the sins of CA are even greater than what many may think. Pointing out a broken feature is easy but how about something which is more core identity to the WH series such as Single Entity Units, those since WH had been a soar point for the AI as CA never considered to think how an AI should act around this feature, when controling those it rushes those incoherently into the enemy front lines to just suicide, when fight against them it just blobs up ... but is it pathfiding ? is it to hard to program AI ? nah... again the same problem it is the design. The AI simpley was never programed to understand what players understand - things like not blobbing around them, mix chaff with strong units, use chaffs to hinder and slow down, use heavy mass to block from escaping, use singe entites with a mix of hit and run and so on... Now that isn't to say that CA didn't improve the AI during WH's time, with the third expansion pack CA introduced an AI that finally can avoid artillary shots or magic on the go! beautiful but not enough, as one of the biggest issues to follow up is predictability, one of the biggest issues that lead to 'exploits' is when a thing is to easily predicatable, a good AI hence needs to be able to be read into (such as reading traits and understanding how the AI might act - a hotheaded AI will be more prone to be provoked, while a stratigic one will try to bait) but not to the point of 100% accuracy in every scenario, something CA despite my best afforts still seems to not understand. But beyond the lack of AI 'understanding' of behaviors, like skirmishing as a defender which still isn't a thing or properly setting up a flank, the problem keeps digging dipper as CA has clearly doesn't other to implament an AI that connects the campaign and the battle. What do I mean? Well just like the player assess an army, the battle AI needs to be 'aware' of enemy army composition as well... hidden or not the AI needs to know a cavlary unit is on the field and to 'expect it', it needs to know that if the dwarven army which is slow as hell doesn't have cavlary it can lose the fight and be able to retreat as it will out run its pursuers and so on... As you understand from what I am writing the AI is tied to many aspects of the game (still, thank god, less than before unfortunately - population was amazing) which means there are many factors at play but the biggest issue of them all is that AI simpley don't put the resources into developing the AI and unfortunately people keep voting with their wallet for CA to not bother aobut this issue or even worse, make up excuses or say it is the best that can be achieved (with zero knowledge of any programming) or simpley accepting the low standards CA dictated :( . I am with you, mods like DEI were god sent but with all the amazing work the people that worked on it did, bringing to Rome 2 crucial features like population systems and supply systems the mod could only do so far as you put it as it could never teach the AI to work with those features and thus the only way to enjoy them is Multiplayer Campaign which I did but its not something I can easily commit to any time. Modders give their best but things like AI and Autoresolve (another one of the most major issues that plauges the series, barely simulating any of the game systems which tilts the results far from what manual battles look like). What is worse is that TW series isn't linear in its improvements - Want great sieges play Attila or ToB, or maybe don't cause Attila is preforming like shit, and both are plagued with AR that doesn't understand how sieges work making them the easiest AR in the sieres (I autoresolved a battle today in Attila with zero tools to break walls, and only 2 siege towers which the AI on manual easily sets them on fire and the AI won it easy with 8 breaches to the settlement... how?!?! there are only 3 gates and 2 towers if you get lucky to park them which was close to improssible with the amount of fire arrows). Want great diplomacy play 3K, it also has decent AI in battles and amazing gaphics, but what it doesn't is balance with the game focusing on a Romace mod that is clearly giving god like characters that reduce armies solo to ashes still ... and did I mention I love good sieges, in 3K they almost look good - they got hte best tower placements in the series variaty of maps and settlement types and sizes and even anti artillary on walls, but what they also have are the infamous ASS LADDERS... cause fuck making walls meaningful. And the issue continues as each TW has a different slice a player might want and this is what CA had been doing for ages.... Players asked for a siege rework for WH, sure thing let us just package it in a 60$ package named Troy. I hope to see more videos , I always said as you said, a good AI is one who challenges not beats to a pulp and does it through 'understanding' the game, with more focus on this I hope we will see change if not in CA - in the fans who might start understand that is up to them to force CA's to make better AI.

Safe from a handful of cases in 20 years of playing total war games, Shogun 1 and Medieval 1 were the only ones to give a challenge on the strategic map. That's entirely because of the chess like nature of the map in those games.

I am always puzzled why people claim Shogun 2 has good AI. They are either hipper-agressive, or static. That was my first Total War back in 2017, it was for me good enough to try games I was actually interested which are Empire/Napoleon that are infamous for poor AI. Truth is however that those games forced me to actually use proper battle tactics. I had to position artillery properly, I couldn't just circle around with cavalry like in every Shogun 2 battle, I had to watch out what is an actual attack and what is a diversion/raid. I was forced to give up on Finland on Russia which due to Swedish raids. AI actuall fight and captures trade roots. Don't get me wrong, AI in those games was prone to stupid mistakes, but it was far less then in Shogun 2 where AI would simply attack army and either go all in, on the attack or would stay still and be picked from affar. And don't even get me started on how after you grow powerful enough all factions arbitrery delcare war on you. That's not smart, that's artificial diffculty at its worst. Hell in Rome 2/Attila, AI would form grand alliances organically of which you can be part of, which is especially fun in Atilla where you carve Roman Empire and plan how to backstab your partners only to find out they were planing the same thing, but it doesn't matter because Huns come to butcher you all. Even AI will ocationally try to do something new like, forming defensive square, burning buildings, bring in the artillery. Even in Medieval 2 AI will try to surround your army, wait for your borders to be left open to attack, etc. Haven't played anything after Attila, but honestly Shogun 2 is for me a big let down and the fact I am not facinated by Japan and Samurai culture in general didn't help. I've heard that Fall of the Samurai is much better and has more in common with Empire/Napoleon. I may try it, but my hopes aren't very high.

Haven't played since Rome 2, kept hoping CA would spontaneously get better, they never did. Glad for all the money I saved at least!

Used to get so excited about a new TW coming out. Ever since they only started releasing strictly Warhammer games i lost interest. Britannia and Troy don't count because they're very watered down TW games that feel like less of a stand alone title than DEI does for Rome 2. Don't get me wrong, i bought Warhammer, WH2 and then WH3...And each time i did i felt a lil' less good about it. Absolutely no more. The Warhammer games are so ridiculously similar it's a slap in the face to consumers. Where as if someone was asked to list all the differences in Empire Vs Napoleon Vs Rome 2....The list would be a long one. Warhammer has brought CA so much money and the audience is wide enough where they can just "make" a new warhammer game with barely anything new and it will sell like hotcakes. CA used to put effort into their titles, now they're just cookie cutter.

You perfectly put my thoughts into a video. I've quit buying Total War titles and have no intention of coming back until we see some progress. I too have owned every Total War since the beginning and actually refunded WH3 for this exact reason. Maybe Bannerlord will shake things up.

Three Kingdoms Total War has the best diplomacy mechanics, best campaign AI and best Battle AI. Shogun 2 ,Rome 2 and Attila had some decent battle AI as well but still not as advanced as Three Kingdoms.