The more I play Shogun 2, the more I’m starting to think Total War games have a fundamental mechanical flaw. Yes, yes, we’ve all seen the botched attempts of the Rome-Medieval-Empire era – schizophrenic AI combined with a more fluid, realistic campaign map which in practice led to some infuriating gameplay tics – but with each of those games, the problem appeared to be specific to the game in question. Usually it was the CA’s crappy AI coding. In Empire’s case it was also a case of sprawling scope and the series first flirtation with the gunpowder era producing – aha – a serious misfire. But Shogun 2 is the first game to make me think that Total War games, in their current form, simply will not work.
How can this be possible? At first glance, Shogun 2 is, by the CA’s standards, nothing short of a triumph. Not only have they moved back to the easier-to-implement melee-based combat of the original games, but being set in Japan means that the grand campaign is smaller in scope. This is A Good Thing; I have trouble believing anybody enjoyed starting up Empire for the first time to find themselves presented with a six-province empire embedded in a 120-odd province map (or three). That’s too big for even seasoned veterans of the series to swallow, and so cutting it down not only makes the game more manageable but it also gives it an increased amount of focus, making the game personal again. Every army is valuable. Every general has to be used. There’s no stack of forgotten family members sitting in the capital in Shogun. As a grand strategy experience, Shogun regains much of the gravitas that Empire and Medieval 2 sorely lacked.
The return of melee-dominated combat is also a welcome shift back to the series’ roots. The cavalry-spears-swords-archers dynamic works well and is easy to understand – easy enough that the CA’s battle AI coders have finally pulled their heads out of their asses and produced something worthy of the name. In pitched field battles, the battle AI isn’t just competent, it’s downright threatening. It doesn’t hurt that the campaign AI fields balanced stacks of troops this time around (although it gets more than a little bit of help with this – more on that later), but the AI grasps the concepts of skirmishing, scouting and not sitting on a ridge being shot to pieces by a horde of archers. It’s still pants-on-head stupid during siege battles, unfortunately, but I’ll take whatever I can get.
The economy has also been rejiggered a lot. In previous games it was something that took care of itself – I just built the odd farm and road network and left the game to get on with it, since that approach brought in enough money that I was never short. In Shogun, though, not only is money a lot harder to come by (unless you’re the Shimazu and happen to live next to all the trade nodes) but unit upkeep costs have been raised, with the result that it is now impossible to field more than two doomstacks of proper Samurai. On the face of it, this is a change that I endorse; previous Total War games got to the ludicrous point of having battles which involved five or six stacks at a time, and that was just on one portion of the front. Armies are no longer disposable entities, which again lends the campaign experience a lot of gravitas since having your prized shock army get mauled in a cataclysmic battle actually hurts. Similarly, routing the enemy’s army should be a significant victory, since they shouldn’t have two more waiting just over the horizon. But this is the point where we come to Shogun’s fundamental flaw:
The campaign AI isn’t playing by the same rules as you.
Now, coding good AI isn’t a matter of programming something that’ll act like a human. Coding good AI is all about programming something which will appear to act like a human. Half-Life is the famous example of this, where the rudimentary Marine AI was cleverly disguised with a range of audio cues; having the Marines occasionally shout out that they were flanking the player created the perception in the player’s mind that that was what they were doing, when they were in fact moving around fairly randomly. There’s nothing wrong with coding AI which is fundamentally inhuman, which gets all sorts of advantages compared to the player in order to present a stronger challenge, as long as these advantages are low-key enough to be hidden. Sadly, the advantages given to the Shogun AI – the advantages that were necessary in order to make it a credible threat – are game-breaking.
Let’s go through them in no particular order:
1) The AI isn’t restricted by its economy. Basic example: higher level castles require food. Single province AI clans will happily run a tier four castle with a tier two farm. This means they should be dealing with food riots every single turn. They don’t. This adds insult to injury when the player takes the castle and gets plunged into food shortages that the AI didn’t have to deal with.
2) The AI isn’t restricted by its buildings. Oh, it’ll cheat those into existence as well, but I’ve seen more than one occurrence of an AI clan fielding units it didn’t have the capacity to build.
3) The AI isn’t restricted by its bank balance. It’ll spam those expensive high-level castles and buildings with wild abandon, as well as fielding multiple stacks of expensive professional samurai when it blatantly doesn’t have the economic capacity to support them.
4) The AI isn’t restricted by the basic game mechanic of having to physically build units. As in, it will happily spawn half-stacks out of nowhere under cover of the fog-of-war.
5) The AI isn’t playing to win. As bad as points 1 through 4 are, they are nowhere near as bad as this one. Civilization IV had AI that on higher difficulty levels cheated like there was no tomorrow, but it was at least playing with the same basic goals as you – - so if you turned wars off, it’d build the spaceship and win the game in 1800AD or something. This meant that the AI civilizations in the game were actual competitors to the player civilization. By comparison, the other clans in Shogun 2 are nothing more than speedbumps; obstacles placed in the player’s path to make it difficult (and at times downright infuriating) to make any headway.
Taken together, these flaws in the campaign AI take all the restored gravitas I described in the first half of this review and chuck it in the bin. When the Asai declare war on me, they’re not doing it because they see an opportunity to increase their power base to the point where they can claim the Shogunate. They’re doing it purely to make my life more difficult. I’m not one of numerous clans vying for dominance, I am instead a person thrown in a pit with a bunch of unfeeling machines whose only goal in the game is to pound on the squishy fleshling, and who get a hell of a lot of help in order to do so. This makes playing Shogun 2 on difficulty levels above Normal an exercise in pure masochism.
The problem is exacerbated by the Realm Divide mechanic. This attempts to replicate the cool idea of the civil war in Rome, with the critical difference being that in Rome you had time to build up a fairly large power base. In Shogun, not only do you have fewer armies to defend yourself thanks to the altered economy, but the Realm Divide event fires when you’ve taken less than a quarter of the provinces in the game and makes everyone else want to kill you, no matter what they thought of you before. This is just… well, from the very beginning of my Oda campaign I have never been at war with less than two clans at a time. Very often it was four, and at one point it went up to five before I wiped one of them off the map. I had been clawing and scraping my way to the top of the local pile of clans ever since the game started, I was just finally getting on top of things and looking forward to actually maybe spending some of my koku harvest on buildings to rejuvenate my war-ravaged empire rather than endless ranks of ashigaru, and then Realm Divide fired. It was at this point that I gave up.
Now, you might think it’d be a good idea to make some friends before the Realm Divide. It is possible to make allies who won’t immediately betray you when it fires (and of course vassals don’t have a choice). Unfortunately, because of the way diplomacy works in Shogun, the only way to make people like you is by giving them huge wodges of cash. And even if you do this, Realm Divide is a constant negative modifier that increases every turn. Eventually it’ll outweigh all your bribery money, and you’ll be at war with everyone.
This is awful. Awful, awful, awful. Every other clan in the game – every single one – is a clan that doesn’t care about winning, that can replace its troops and buildings at a whim, and that will declare war on you sooner or later. As a final incidental point, all that stuff mentioned above makes it impossible to sit back and build your economy. When somebody declares war on you it’s a fair bet you’ll be at war with them, on and off, for the rest of the game – and even if not they’ll turn on you once Realm Divide comes up. This makes total annihilation of the clan in question the only valid response to a war declaration. Except, taking another clan’s territory will add a negative modifier to your diplomacy rating with other surrounding clans. This negative modifier will tip one of your neighbours into war, and then another, and so on, ad infinitum.
All this is a very lengthy way of saying that, in adding all the things that were apparently necessary to create a campaign AI that could present an actual credible threat to the player, the Creative Assembly broke their game to the point where it simply doesn’t work on Hard and above. This is why I say that Total War, as the concept currently stands, may be fundamentally broken; if this is what it takes to make the game challenging, if the only way to make it hard was to add layer after layer of shit to artificially boost the difficulty level, then that’s a challenge I don’t really want any part of. If an AI that can function without heavily weighting the game against the player is truly beyond the CA’s abilities then they need to adjust the basic game mechanics until such an AI drops back into the realm of the feasible. And that’s the end of it, really.