Tobold's Blog
Wednesday, April 08, 2009
 
Empire: Total War

Once upon a time, it must be around 20 years ago now, I played my only ever World War II board war game, I think it was one of the Axis & Allies games. I had met this American guy at university, and he had spent hours explaining the rules to me and setting up this huge number of playing pieces on the large board. But I hadn't quite understood the concept of a game as historical simulation, and did something completely unhistorical: Playing the Allies I spotted something I perceived as a weak spot in the defences of the Axis, and attempted a landing in Bremen, Germany, many years before D-Day. It turned out I had understood or heard only half the rules, and the odds were stacked heavily against me, but a lucky roll by me and an unlucky roll for a counterattack later, I had established a bridge head in Germany. We never finished that game, but my opponent told me that I had basically won the game, as the troops he would need to contain or expulse me in Germany would be missing from the finely balanced situation at the eastern front or elsewhere.

I had to think of that memory of the past when I started my first "grand campaign" in Empire: Total War (ETW) as the British, and instead of following the game's obvious intention of making this a game of empire building in the colonies, simply took Paris in 1702, and completely unbalanced the game. With France out, taking the colonies will be so much easier, and having his starting provinces hugely increases my resources. I guess historical games still don't deal very well with people who refuse to follow history.

Empire: Total War manages at the same time to be a great game, and a huge disappointment. ETW has tons of interesting new features compared to previous Total War games, fantastic graphics, and great battles. It also repeatedly crashes to desktop, not only on my Vista-64 machine, but on many player's computers, even after following advice like running in Windows XP SP2 compatibility mode. It is loaded with various annoying bugs. And it's AI is dumb as a rock.

I have some hope that the bugs and crashes will be fixed with further patches, the first patch already came out just after release. But the bad AI is bothering me far more, especially on the campaign map. After reading up on it, it turns out that it is technically impossible for me to lose the game as Britain: 5 of my 6 starting provinces are on islands, and the AI is too stupid to load an army on a ship and invade islands. And while I simply bundle up my troops into few large armies that are able to take out defended capitals, my enemies are wasting theirs only ineffectively burning my farms with small armies, which are then easily picked off. In the battles the AI isn't as stupid, at least not in the larger battles. I did have a very small battle, with me just having one troop of very bad militia infantry, and the enemy had much better cavalry, which I won easily by putting my infantry into a house, and shooting at the cavalry which was unable to attack me in there and continued to circle the house until they were dead.

I could turn up the difficulty level, but that wouldn't make the AI act any more clever, it would just cheat and give them more troops. So my story returns to that time 20 years ago, when me and my friends were playing a computer game on the Amiga, also called Empire: Why is it that in those 20 years the graphics even of strategy games have become better by such a huge margin, but the AI today is still exactly as dumb as 20 years ago? Why does my computer have a $200 graphics card, but no AI card at all? I'm not asking for human-like intelligence, all I want is a strategy game in which the computer is playing reasonably competent.
Comments:
This is a great point and i think one that is connected again to a fundamental part of today's society. People don't like challenge but they do like eye-candy. Or to simplify it, people are shallow.

More people will buy a game because it looks pretty than because it's going to offer them an interesting challenge. How often do you see games advertised with reference to the AI, as opposed to games advertised with references to the graphics?

Many of those who do say they want a challenge, actually don't. Like realism in games, they want the illusion of challenge rather than the real deal. Which is fine. (Slightly off topic, it's much harder to code AI that gives the illusion of challenge than one that just provides straight up miserably hard gameplay.)

Game Studio directors know what sells today, eye-candy. So that's what they focus on. The limited market for people who actually want any sort of challenge in their games are expected to be satisfied with PvP type gameplay.

Incidentally, one of the most exciting developments for me is the AI behind 'Left 4 Dead'. AI basically designed to set the pace and possibly the challenge of gameplay based on the players themselves. Imagine an instance in WoW that tailors itself to the group. A boss that comes looking for the group when they don't expect it, but when the game has analyzed that they are ready to take him on?

I really think stuff like that needs a bigger spotlight, but it won't.

Today's new breed of 'casual gamers' demand light entertainment with pretty pictures.
 
AI is a game in itself and is still bound to memory. (IBM Big Blue is just a large memory-heap so it can defeat Kasparov in faster time). The biggest problem, which has not yet been solved today, is that a human witha blink of an eye can see a "gap" or a "solution" to a problem, while AI still has to calculate the possible solution. ofcourse AI can have a map of shortcuts, but it still has to go through all the possibillities.
If science can overcome that problem, than AI can be as strong as a human. Now AI still is a scripted machine made by a couple of human beings, so there will be gaps everywhere.
 
http://www.findmysoft.com/news/Artificial-Intelligence-Robot-Scientist-Adam/
or, for the German speaking community:
http://www.faz.net/s/Rub163D8A6908014952B0FB3DB178F372D4/Doc~EBEA95A2001B04D4992E45D5F390AE487~ATpl~Ecommon~Scontent.html
are interesting links. AI is coming along. Just not in computer games.
What I did during my diploma thesis repeatedly remembered me of the Startrek computers with the nice voice. They may not speak today; but otherwise they are quite capable - even if not genuinely intelligent.
 
Maybe we can soon get to the "Chess AI" model where we brute force every possible case and choose the optimal one. But it's probably still too big a computational task. And I doubt most commerical computer games are understood well enough to distill them into such calculations anyway.
 
I've played one campaign with the Dutch. And found it to be a bit too easy. Conquered Paris and from there on it was "I've won". Could just pick up their ex-colonies in the US. And with Paris in hand I overwhelmed Europe from France to India. If I'll try another campaign, it'll be with the hard setting, normal felt too easy. I did enjoy the game though, making those "what will I invade next?" decisions are fun.

Armies not invading armies is a serious bug too. And yes, I don't see much use in burning a few farms when you can just attack the capital. Blockading trade routes is effective though :)
 
I had the same feeling in Medieval 2: Total War. The campaign was too early clearly decided in my favor. They managed to improve city sieges over time, Rome and Medieval 2 were definitely better than the old Medieval 1 in this regard. They also featured the new 3d unit graphics style for the first time, but the AI got worse (it cannot handle the new features they added). In ETW, units sometimes have problems climbing their own ropes when they try to assault a fort. The general trend was better city sieges, better graphics, and new features that the AI could no longer manage. This is especially the case for ETW, which got some UI and control improvements. Naval battles were a letdown, boring and longish. Just do not try to maneuver too much with many ships, you will cause yourself more harm and micro than necessary.

Not only the strategic AI got worse, the tactical battle AI got worse, too. It basically only boosts units with extra stats by now. In Medieval 1 it had some moments of glory, like trying to prevent flanking attacks and attacks from the rear.

I wonder why the AI is no longer capable of fielding a strong army that actually could cause problems. I rarely face strong enemy armies in either MTW2 or ETW. In MTW "one" I often had to run away from the Saracens in the Holy Land. It had much more interesting and exciting fights, and I did not have to play a mod or expansion when I wanted to start in the early, high or late medieval period.


The phenomenon is not only restricted to the TW series. Sins of a Solar Empire also has a very predictable AI that willingly attacks death traps over and over, while it is scouting undefended systems and does not even bother to attack it with a small force or at all.

What I really like about ETW and Sins of a Solar Empire is the improved UI. The interface of strategy games got greatly improved over time. I still fear that the AI gets worse and worse because we add more and more options and niceties for human players that even much more advanced AI scripts could not manage properly.

I am a bit disappointed that each Total War game gets such high ratings, were it not for the graphics one could say the TW games are actually slowly getting worse than their predecessors.
 
That's not a Total War problem. That's just an AI problem. It doesn't have to do with people not wanting a challenge, its that its apparently really f'ing hard to get a computer to fake it. We're barely to the point where FPS npcs don't just run right at you guns ablazing. I have yet to see a strategy game where I thought the enemy had even a spark of intelligence or did anything brilliant.

And as far as why its hard--- it takes a supercomputer to do halfway well at Go, which is a lot more free form than Chess. The only way to get a computer to behave "intelligently" is to do massive amount of brute force calculations, and the more complicated the task the more calculations. Strategy video games are generally more Go like than Chess like, even though chess is the preferred strategy metaphor in the western world. And that's why AI hasn't improved all that much since the first Civ. You'd need a team of scientists and a frigging supercomputer to get the computer to handle a game like TW well.
 
Speaking of AI:
http://www.jeeney.com/
:)
 
I'm pretty sure I've read that turning up the difficulty makes the AI more formidable, as in it will assault islands and such, not just add more troops. I'm curious if you've read or heard something to the contrary, or if you just made up your mind that upping the difficulty wouldn't make the AI play you harder without "cheating" ?
 
I found this link: http://forums.totalwar.org/vb/showthread.php?p=2201202 saying AI isn't better at very hard, and just cheats. Please link if you have other information.
 
Your anecdotes mainly point out the value of good playtesting, something that doesn't happen enough with most games. Back in 1981, I read a guide to wargame design (just checked Google, you can download it at http://index.rpg.net/display-entry.phtml?mainid=9824) and one major point was that you don't just need playtesters but you need the right kind. Experts in the subject matter will never try crazy things but novices will so you need people that aren't familiar with the genre. That's what you did by seizing Paris. An expert in that era or genre would "know" that you can't take and never try so the designers wouldn't realize they need to guard against it.

Bad AI-wise, I seem to recall that there is a computer game that had really good AI, I think it was Galactic Civilizations. My understanding is that the designers built it from the ground up with a design that lent itself to having a good but easy-to-program AI and I guess it worked. Most games bolt the AI on at the end and don't try too hard to make it good with more difficult settings usually just increasing how much the computer cheats.
 
Ya Galactic Civilizations has a really good AI system. There have been some really interesting stories of what the AI has done in that game on the harder difficulty settings against some skilled players.

This is also why I don't care to play games on hard mode. Increasing the health pool and damage output of a creature doesn't make it harder. If you raise the mobs HP and then your own stats proportionally it doesn't make a different. That's what WoW has done. Make mobs harder and then players stronger, so percentage wise everyone is as equal as they were 2 years ago.
 
GalCiv was actually exactly what I was talking about.

It's not BAD AI by any means, but it also uses the 20/40/60 production bonus on the higher levels of difficulty rather than getting smarter.

I have never really be surprised by an AI doing something really smart. The game basically revolves around getting a strong enough economic base to conquer your neighbors; once you establish sufficient production you be able to mount an effective blitz attack the game is pretty much yours. But that is equally true of Master of Orion back in '94.

But like I said, best AI I've ever seen in the genre. But not exactly smart either.
 
I've always been disappointed with the strategy layer in the Total War games. I find for that period feel, Europa Universalis II delivered a far more satisfying experience. If you could simply add the Total War tactical layer to EU2's strategy layer, that would be one heck of a game.
 
Post a Comment

<< Home
Newer›  ‹Older

  Powered by Blogger   Free Page Rank Tool