Far too easy?

[Soylent Green]

Everyone keeps saying they should make a better AI, but just exactly how is that done?

AI reactions have to be coded, dont they? Or am I mistaken in that? Dont they have to basically be a series of "if - then" or whatever? If that's the case, how much "predictive" "well, if the player does this, check that that this this those these, them, that over there, etc etc etc" can a set of coders be expected to produce without making them so variable that the AI is frozen in a constant variable check?

I honestly think some folk just will not be happy unless they can play full campaigns against other humans. Then they will complain because the other humans arent smart enough, tough enough, whatever. Then if they start getting beat, it will be the humans cheating.

I know that wouldnt be the case for all of the multiplayers but we'd hear that just about as often as we hear "they should make the stooopid AI smarter".

Some day it might be fun to go back through the forums and see how many of the people whining about pointless battles now were whining about "RTW: Siege War" back then or bitching about the "simple risk map". I remember long and endless discussions on the various forums where the loudest of the players hated the RTW AI because it was all about sieging cities and you never met an unexpected fight in the field. Now that seems to be reversed, agian, again.

There's always chess, I guess, for those that just can wrap their heads around the fact that there are limits to games designs in a digital format.

So basically you're saying there's never been and never could be, a challenging AI, ever. I disagree.

[Carl]

So basically you're saying there's never been and never could be, a challenging AI, ever. I disagree.

I'm sorry to say it, but he's right.

Their is actually a mod for DoW that give the games AI a boost, the problem however that this mod team are begining to encounter is that they are limited in how much better they can make the AI because of the performance hit. They've had a few releases now that they've had to hotfix because of lag issues even on high end DUAL CORE systems.

The TW Games are FAR nmore feature packed than DoW and thus doubtless require a very complex AI coding just to get the current level of difficulty. You COULD code a truly challanging AI, but without a system with several multi-core processers in it you could never run it as the performance hit on even a high end home PC would leave the game unplayable.

Why do you think AI desginers have relied for so long on giving the PC more money/recources/whatever and better units when compared to the player in the same situation. It's because it cuts down the AI processer lag.

[katank]

Also, the rules for TW series are many and it's extremely complex. The AI community still hasn't solved Go yet. Its rules are significantly simpler. The question is a matter of computational power.

This is not to mention all the devious ways that humans can come up with exploits and so on to best the AI. Certain gambits will almost certainly throw off the AI as the game tree will be too big/complex to search.

[Barry Fitzgerald]

I bet you if CA had spent as much time on the AI or gameplay..this forum would be singing praises day in and day out..

Sadly they didnt. Lack of cpu power isnt the issue..if MTW 1 could have decent AI...why not this?

I wasn't planning on having turkey this xmas, but ahem well I have a half turkey instead! lol

On a serious point....I think the TW community as a whole is somewhat let down with this latest version....much as I want to love the game...I cannot...though it does have some great moments...sometimes...

A flawed gem...polish it CA..redeem yourselves...

[Reapz]

Solyent

Just for you vets though - remember the days of Shogun, where you'd destroyed all but the last faction, and then found they controlled half the map? Or the days of MTW, where you had one more territory to go, and then your entire empire when into civil war, and you smashed the computer screen, poured petrol on your case, and sent the whole lot burning to oblivion?

I remember well that phenomenon in Shogun. The genesis of it was that the AI could build without any economic constraints so there was no point in trying to destroy their infrastructure to impoverish them. I remember the first time I used the :matteosartori command to see what kind of army the last AI faction was fielding and seeing five to ten full stacks in most of their provinces. It was insane. It roused a lot of debate at the time. Although it was "hard" to beat AI it wasn't playing the same game the human player was. Personally I didn't find that much fun. That type of AI advantage persisted until RTW. I enjoy the fact that in M2TW the AI doesn't get that kind of crazy economic edge on vh and just field many more units than me. Other than those few factions that actually came in hordes it isn't exactly historically accurate.

So while I would like to see the AI harder to beat I definitely don't want to see it get to build five times as many troops as me. And I don't hold up the earlier TW series games' AI as something to aspire towards. I think that the style of the M2 AI is a step forwards, if not the strength of it. I would be all for a smarter AI that had the same economic rules broadly speaking as human players so that blockades, razing cities etc. actually hurt them. I think there is a lot of scope to make the AI stronger in other ways like unit selection, picking battles more sensibly etc.

[Malachus]

I have been playing TW games since MTW, and that was after I tried STW at a friend's house. And yes, I will agree that M2 is too easy at times, especially when you're rushing the AI with armies, forcing them to churn out what's available to them ATM. I was playing as Spain and within 10-15 turns had already driven the Moors out of the Iberian peninsula.

But I felt the same way with RTW. Playing as the Julii on VH/VH, I would face armies from Gaul that outnumbered me 4:1, and I still crushed them. Playing as the (supposedly difficult) Seleucids, I just bribed all my enemies and conquered the entire western part of the map. That game was definitely as easy as M2.

And MTW AI wasn't that much of a challenge, either. Sure, it was definitely cool (and challenging) that there were re-emerging factions and civil wars, but even those could be easily averted by a clever player. And battles were never difficult unless the enemy general had tons of stars (Jedi Byzantine generals, anyone?) giving all his troops a large valor boost.

So, IMO, total war games have always been plagued by an easy AI.

In M2, it's important to note that cities and castles don't produce the same units, and that once you build a stone wall in a city, it can't be converted to a castle ever again. From what I see, the AI usually has tons of cities and relatively few castles, so the majority of the troops they can send at you once they've built that stone wall is militia. Whether or not the AI converts cities to castles or vice versa is another question.

Anyway, there are lots of mods in production that are aimed at improving the AI and realism of the game. If you've played Europa Barbarorum for RTW, you would realize that so long as the modders are dedicated and working hard, there is still much hope left for this game.

[katank]

AI having too many cities? Really? I find that they usually don't have enough cities. A huge proportion of rebel settlements are castles and the AI doesn't convert them, leaving them without sufficient cash to build units out of the castles they have already.

[Julius_Nepos]

I personally find the 1.1 AI to be anything but "too easy." As always I'm handicapped by being a superior field commander but not all that good at waging a campaign on the map itself. I've found playing on hard/hard that the AI is so utterly aggressive that the only real way to survive is to blitz blitz and blitz everyone so they don't have a chance to overwhelm my positions. I don't really like that style of gameplay but I'm finding I have no choice but to engage in it.

The fact that the AI is suicidal, will attack at every opportunity and will do nothing to preserve itself or "win" the campaign in its own right just makes things harder. In my latest failure of a Hungarian campaign I survived by blitzing the Byzantines early but at the end I was under siege by 5 elite armies at every corner of my domains with no hope of holding the line. Sure it's nice to know Milan, Denmark, Turkey, Russia and Venice were able to discover my weaknesses and attack appropriately (in fact I didn't even share a border with Denmark!) But in the end it was just another loss for me based upon my desire to play a laid back campaign of internal development and the AI's insane lust for battles.

In many ways my style of gameplay is just no longer in line with what Total War is all about. I'm seriously considering going to Civ IV, or Europa Universalis or something that would focus more on things other than 2-4 battles a turn. I have very mixed feelings about the AI but "far too easy" isn't a sentiment I share. Yes I can crush AI armies on the battlefield even with the odds clearly against me, but I have yet to figure out how to blitz all opponents, field 5 full stacks (all upgraded) while upgrading my cities and defending all settlements with 20 units of chivalric knights while still pulling down 50,000 florins a turn. It seems to me that some people can do this, but I have yet to determine the secret.

In any event, the game is plenty hard for me, I just wish it was hard in a good way. Cheers!

[Spark]

I'm sorry to say it, but he's right.

Their is actually a mod for DoW that give the games AI a boost, the problem however that this mod team are begining to encounter is that they are limited in how much better they can make the AI because of the performance hit. They've had a few releases now that they've had to hotfix because of lag issues even on high end DUAL CORE systems.

The TW Games are FAR nmore feature packed than DoW and thus doubtless require a very complex AI coding just to get the current level of difficulty. You COULD code a truly challanging AI, but without a system with several multi-core processers in it you could never run it as the performance hit on even a high end home PC would leave the game unplayable.

Why do you think AI desginers have relied for so long on giving the PC more money/recources/whatever and better units when compared to the player in the same situation. It's because it cuts down the AI processer lag.

Don't know if you've ever played Civ 3, but the campaign (or strategy map) AI in that game makes a joke out of the M2TW AI. I'd believe it wasn't possible if I hadn't seen it somewhere else, but it is clearly possible to make AI much more challenging than the current one. So complaints about the shitty AI are very valid.

[Soylent Green]

Don't know if you've ever played Civ 3, but the campaign (or strategy map) AI in that game makes a joke out of the M2TW AI. I'd believe it wasn't possible if I hadn't seen it somewhere else, but it is clearly possible to make AI much more challenging than the current one. So complaints about the shitty AI are very valid.

Agreed. Although part of that was down to the AI beginning with additional workers and settlers. And getting certain bonuses along the way. But i don't see why this couldn't be done with MTW2, it's certainly more valid than the player having to edit the game to achieve the same.

And with AI scripting, did anyone ever script missions for TA? I don't have a clue about programming, but i scripted complex unit behaviour for some missions, making them particularly challenging. It didn't affect system performance in the slightest. Now granted, TA was set piece maps and a different kettle of fish, but the principle is there.

Would you rather have 'huge' unit size or a challenging campaign? If performance is the issue, i know that aesthetic elements of the game could be glossed over to make room for a more powerful AI. There's so many claims that blitzing is 'taking advantage' of the AI's inadequacies, well, not only do we have very little choice, but being at war with multiple factions simultaneously provides the great challenge this game has to offer.

Total War people. Not total faff around making concessions to the AI.

[Kobal2fr]

@SoylentGreen : there's an old French saying that translates roughly as : "You can't please everybody and your father". Meaning, whatever you do, someone is bound to not like it. Case in point : earlier TotalWar titles gave the AI free cash, every turn. Hence no matter what you did, the AI would field stack after stack after stack, even when you just destroyed their whole army the turn before. Very, very hard to beat without blitzing.

We, the players, complained that it was unfair, and that it tremendously reduced the point of economic/agent/non military warfare against the AI, since what worked against you didn't work against them. CA changed this and made the AI not cheat anymore, merely banding up and being much more aggressive at higher dif levels. Then you complain the game is too easy if you blitz ;)

And scripting is all very well, but it's static. I remember being scared right out of my shorts when playing the haunted house level in Vampire : Bloodlines for the first time. I kid you not, I had a bruised knee for a month afterwards, I hit it against my desk so hard. The second time around, I still jumped at times, but the scare was gone. Now, I routinely race through the level without being scared in the least. That's because everything in there is scripted, appears and pops up at the same precise moment every time. I certainly wouldn't want that in TW. Replay value zero, when part of TW's appeal is to play a faction, then another, and have a different campaign each time.

[Spino]

AI issues aside I believe the biggest factor behind the game's lack of challenge is the much lauded provincial 'castle or city' system. Regarding the strategic game in previous TW games most of us complained about the AI's inability to build balanced armies, consolidate its stacks and not engage in suicidal multi-front wars. In MTW2 those issues have been addressed to a certain degree but now that the castle/city feature has been implemented it's as if we're back to square one.

Over the course of a few campaigns in versions 1.0 and 1.1 it's been my experience that army stacks produced by city provinces are pretty much a pushover in battle. I'm currently playing a campaign as the Danes and now that the early game 'rebel province rush' has subsided I'm finally going toe to toe with AI faction armies. Scotland has been giving England a hard time and now has eyes on the European mainland. In the past five turns they've sent no less than three full stacks to besiege my forces in Bruges. I am currently holding Bruges with a full, well balanced stack consisting of a couple of Royal Bodyguard units, several Mercenary Spearmen & Mercenary X-Bow units, Dismounted Huscarls, Norse Raiders and a few heavy and light cavalry units. Each time I leave Bruges or sally forth during a siege to engage these stacks I brutally massacre them. Why? Passive AI bug? No. Clever tactical deployment on my part? No. I'm mauling these stacks because every single one of them consists mainly of Town Militia, Crossbow militia and perhaps one or two cavalry units, for the most part crap units. Admittedly the last Scot stack I faced sported some Militia Pikemen in full plate but the lack of decent support units (i.e. axe, sword or heavy cav) meant they were simply dead men walking. Were it not for the castle/city build I would probably be having a much harder time of it. Sure, city provinces make a faction wealthier but at what cost? If an AI faction only has city provinces in the area you know you're going to have a much easier time of it. Take out an AI faction's last castle province and it's only a matter of time before it's knocked out of the game.

The provincial castle/city system does provide an added challenge to players but it also seriously handicaps the AI's ability to field effective, well balanced armies. Should CA have better programmed the strategic AI so that it's more effective at dealing with the castle/city system? Sure, but an effective AI that can effectively combine units from different castle and city provinces to create balanced stacks in the same fashion that a human player does sounds like a really, really tall order.

I'm inclined to say that modders should, if possible, dispense with the castle/city system altogether and instead leave only the castle tech path in place and add in the necessary City-only buildings for full economic & cultural development.

[Yun Dog]

I must admit I like the castle/city system - I think it would be a pity to have to mod it out - its always been one of the battles to try and code the AI to composed mixed stacks/armies and not just field masses of archers or seige weapons like it used to do in MTW. Pity the AI couldnt be tweaked a bit more to combine armies more effectively.

The CIv3/4 AI - yeh its hard - but why is it hard - it just gives the AI gifts - you have to learn to be an A grade exploiter of the AI weakness to even stay competitive - not my idea of fun - got bored with Civs for this reason - every game was the same - if I wasnt using template X for victory then you were losing - so really playing over prince difficulty wasnt really playing anymore - just exploiting IMO.

For my 5c - Ive had some enjoyable campaigns despite whatever limitations the AI must operate under.

My M/M camp as france - I turtled so well the game ended before I took the necessary provinces - maybe that was the AIs tactic to slow me enough by attacking me that I couldnt take the necessary land. ( hey I choose to see the posistives) - still had some epic battles against the germans and the danes - and good stacks of hard hitting troops - I lost battles and citys - cant complain - kept up relations with the pope - lost a general to inquision - didnt let it continue by increasing piety.

My H/VH short camp as the egyptians was also fun - fighting the crusaders - losing antioch to a combined polish/ventian army of 3000 hard core troops - rebels incited by the moors kicked me out of coroba and marakesh - it took almost every man I had to take one of those citys back and execute everyone in the city.Desperately had to find and wipe out the last Moorish city before I was indundated by the mongols who had arrived on the steps.

The Ai does the best it can - mods can tweak the game eg EB .. to make it a little tighter for the human via cash income and unit strengths etc, in EB the AI is improved this way - fielding better stacks of troops and making some better moves on the campain/empire front - if they can do it for RTW then they can make M2TW a hell of a game

[Zenicetus]

Don't know if you've ever played Civ 3, but the campaign (or strategy map) AI in that game makes a joke out of the M2TW AI. I'd believe it wasn't possible if I hadn't seen it somewhere else, but it is clearly possible to make AI much more challenging than the current one. So complaints about the shitty AI are very valid.

Complaints are valid, but they should take into account the differences between those games. CA has to spend programmer hours and testing on both a strategic map AI, and a realtime tactical AI for the battlefield. Civ has one AI engine for the strategy map, period. That's the whole game.

Games have to sell for roughly the same price, with similar production budgets and lead times. So, is it fair to use products like Civ or GalCiv4 as a comparison, where 100% of the developer's programmer hours and testing can be spent on the strategy map?

That doesn't mean I'm happy with the strategic AI. There's plenty of room for improvement, and given CA's recent history I think we'll see at least a few more patches to improve things. But I'm not expecting it to match or beat a pure, 100% strategy game like Civ4, or GalcCiv2. That's just not realistic (IMO).

[Ar7]

Complaints are valid, but they should take into account the differences between those games. CA has to spend programmer hours and testing on both a strategic map AI, and a realtime tactical AI for the battlefield. Civ has one AI engine for the strategy map, period. That's the whole game.

Games have to sell for roughly the same price, with similar production budgets and lead times. So, is it fair to use products like Civ or GalCiv4 as a comparison, where 100% of the developer's programmer hours and testing can be spent on the strategy map?

Well the tactical AI isn't something to be proud of either. So CA left everything unfinished.