You are misreading my post, I am not critisiing the Victoria fans for their feedback. I am simply observing that you guys, unsurpsingly some would say, keep asking more. The lesson we have learnt is that sometimes we simply can't give you more.
-
We have updated our Community Code of Conduct. Please read through the new rules for the forum that are an integral part of Paradox Interactive’s User Agreement.
You are misreading my post, I am not critisiing the Victoria fans for their feedback. I am simply observing that you guys, unsurpsingly some would say, keep asking more. The lesson we have learnt is that sometimes we simply can't give you more.
I'm still waiting for a game to be released with challenging, intuitive AI. Not sliders which impose restrictions to give it an 'artificial' difficulty, but actual AI that can plan and react and surprise the player
All of these games like EU3, HOI2&3, all of the Total War series, Sins of a Solar Empire etc, they are all fantastic fun to play, but every single one of them falls short in the AI department. The AI in these games is not actually AI at all but just a bunch of very very basic logic routines, and apart from the multitude of obvious bugs involved in the programming of those routines they are, more than anything else, predictable
When someone does eventually come up with real, challenging, interesting AI, it won't matter how shiny the graphics are, or how realistic the historical background, it will be an instant huge massive success that will take these types of games to a whole new level - and make the people who designed that AI system a LOT of money in the process
So my personal plea to developers like Paradox Interactive is - Please, before thinking of another gaming concept, or polishing an existing franchise title, use your obvious talents and come up with the holy grail of strategy gaming - A Real AI
Thanks
All of these games like EU3, HOI2&3, all of the Total War series, Sins of a Solar Empire etc, they are all fantastic fun to play, but every single one of them falls short in the AI department. The AI in these games is not actually AI at all but just a bunch of very very basic logic routines, and apart from the multitude of obvious bugs involved in the programming of those routines they are, more than anything else, predictable
When someone does eventually come up with real, challenging, interesting AI, it won't matter how shiny the graphics are, or how realistic the historical background, it will be an instant huge massive success that will take these types of games to a whole new level - and make the people who designed that AI system a LOT of money in the process
So my personal plea to developers like Paradox Interactive is - Please, before thinking of another gaming concept, or polishing an existing franchise title, use your obvious talents and come up with the holy grail of strategy gaming - A Real AI
Thanks
Fair enough. I think a lot of the HoI3 criticism is valid, I think the game ideas were just too ambitious initially, I'd bet the game will work as designed after another patch, however that's far from ideal.
HttT has a few bugs, but no one's making a big fuss over it because none of it is game-breaking. I realise thats a lot easier to do with an XP than a new title though.
I think compared to most games, PIs AIs are very good (HoI3 being an exception), especially given how complex the games are (and the fact that there is really no end goal).
I'm no techno expert, but I wonder will an AI ever be able to play as complex a game as EU3 as well as a human.
You are not asking for much there.
There is a very good reason why there isn't a 'real AI' it is because it is so bloody difficult to do. AI is something we do constantly refine though to make it better. Doomdark is a dedicated AI programmer for the 1.04 patch, with Johan on game play fixes and another on optimisation. Even then there will be always room for improvements. Some of this is simple observation of how the AI behaves. Take the HttT AI, it is now more likely ot go after major threats if they are exhausted in a long war. Why did we not think of this before you may ask? It simply took many hours of observing and refining the AI behavour to add this nifty little feature which players like. The same is true for the HoI3 AI. There will be various AI improvements, some small which you may not ever notice, but all together make the AI a more challenging oponent. The AI is a process of evolution rather than revolution.
I understand completely. My first thought on seeing the reactions to HOI3 was, "Well, I'll be damned. Paradox was over-ambitious and wasn't able to do what they intended so quickly." But the same things you added that now are now "broken" are things people wanted: Realistic supply system, more invasions, more micromanagement (which is more for the AI to have to do properly if a player choses to let the AI do it). Everyone knows AI is never good enough for strategy game nerds. Hell, I'm glad it isn't I would feel stupid.
IMO we need to (as customers) have realistic expectations of Victoria 2. The more complicated you make something, the harder it is to find problems and the more problems you will have. Apply your knowledge of complex, unreliable, hard to fix and produce yet potentially kickass German heavy tanks to the damn game you want to buy
.
I'm guessing that less historical play makes the AI harder to program because it can have more and diverser goals.
King beat me to it, but seriously, don't you think it is incredibly difficult (bordering on the impossible) to break down into logical algorithms the complexities of human decision making (assuming it is intelligent in the sense of 'AI'). Not to mention our capacity for adapting to different situations. So yes, I think you may be asking for way more than you realize you are asking.
I think PI is doing a great job. That being said, once we, as humans, have inferred the underlying algorithm for ANY AI behavior we, since we can adapt, can usually beat it since we can then deduce how to exploit it.
Chess is different because there are no hidden moves, and given enough CPU and RAM, a computer can predict every single remaining move and pick the best. That's not possible in a game like HoI since there are hidden moves. Even so, Kasparov managed to beat Deep Blue by making moves that were actually sub-optimal and was able to 'surprise' it. OK, just a bit OT but I felt it was relevant.
Yes and no, it some ways it acutally makes it easier because you do not have to have the AI constantly diverge from its best plan to go do something historical but, in the current game context, stupid. However in HoI3 we added steering scripts and refined them and will continue to do so.
Guys, you wanna see a screwed up game, look up the Modern Warfare 2 forums (I know, I know, an FPS lover, he should be stone!
SERIOUSLY broke game that requires some people to open ports on their router to be able to play MP, and then asshats upload trojans to them thru the game! What game should require you to open ports????
A cheating control system that not only does NOT stop cheaters, it stops those legitimately trying to make the game playable thru mods! And you apparently (I have never gotten on to MP, because I refuse to open ports willy-nilly) can't find an MP game that doesn't have at least one hacker, running around aimbotting or wallhacking.
Not one post EVER from any devs or support! The PC forum moderators, who BTW play the PS3 version, NOT the PC version, have basically given up on the PC'ers, and dont even post, moderate, etc...
Infinity Ward's way of releasing game update information....TWITTER!!!!
Soooo, count your blessings. King is posting on this forum on his day off. Find another company that does that.
I am a fps clanguy myself and the MW2 forum is nothing compared with the poisonous storms of honest rage which did ravage the fora for Wolfenstein and Operation Flashpoint: Dragon Rising. But they all 3 companies got in it common, they did not and will not fix the main issues with their games (mainly broken PvP aspects in 3 game series renowned for PvP).
So how is this on topic? PI will at least eventually make the game a state of the art game, no matter what bugs you had at release. That is not the case in the majority of releases I have spent money on recently. Ok one can say, that is no excuse for bugs, but this comparison sets things into perspective.
Is any one aspect of AI behaviour harder to program than other, King? I imagine diplomacy might be the hardest
I am not a programmer, but as a tester I feedback to the programmers. With any aspect of AI behavour it is articulating why it should do something. If we take HttT, it is all very well saying the AI should seek to check blobbing countries but what does that mean? It becomes the country you rate as a threat and then check its war exhaustion and things like that. So it moves from a general statement of intent into a clear rule that an AI can follow.
How far in advance can an AI plan? Can it think far ahead, i.e. could Sweden at the start of a 1399 EU3 game foresee that it's not in its interests to let Muscovy get too big?
Also, can the AI think 'on behalf' of other countries? Can Prussia in Vicky realise that keeping AH reasonably strong could eventually be to its advantage, even if at that stage it isnt?
Also, can the AI think 'on behalf' of other countries? Can Prussia in Vicky realise that keeping AH reasonably strong could eventually be to its advantage, even if at that stage it isnt?
There is a long term and a short term component to the AI behavour. The works between the two goals. The debug information in EU3 is very interesting in that regard, you can see a list of provinces the AI wants to conquer, although could conquer might more accurate, and a value. This then feeds into the AIs decision making logic on who to ally with and who to decalre war on. The province might not be the most valuable but the defender is one of the weaker ones.
The AI can also read decision triggers, the Castile AI weights the provinces it needs to from Spain much higher than other provinces because it complete a decision with them.
However 20-20 historical hindsight is much harder. the Swedish AI cannot rate the threat of a strong Muscovy in 1399 because it cannot know how strong Muscovy will be. The Prussia AI cannot rate the long term value of a freindship with AH, because it cannot know if France and Russia will ally against it.
Theres more truth in what he said than you people think. It wasn't anywhere near ready for release, then why did they release it even though they MUST have known that it would be this bad and buggy, and they would have to patch it at least 5 times. Scary... Nevertheless i do WANT to like HOI3 its just hard to at the moment, its has so much scope and is far superior to HOI2 but its just not there yet but in time it will be.
Roll on 1.4 patch!
Roll on 1.4 patch!
Last edited:
King: Istead of ranting the game I will try to turn this into a positive thread by humblily suggesting a way to improve the AI. In my opinion the tactical behaviour of the AI is already quite good, of course there is always space for improvements, like programming the AI to use paratroopers for example. In terms of global strategy the AI is quite bad at the moment. For example a hungarian AI does little more than taking a few provinces from the enemy nations, whereas a human player controlling Hungary ends up with a huge Balkan empire stretching all the way to Persia. Improving the strategical aspect of the AI can only be done by using what I call behaviour patterns. Every nation should have several preprogrammed behaviour patterns that trigger given certain circumstances. For example, Germany AI is really bad at invading the UK, but for human players it is quite easy. This is because human players follow certain patterns that the AI ignores. Japan failures in China have been reported many times. This is because Japan AI doesn´t know how to cope with the limited port capacity in China. Then simlply give the AI the pattern to cope with it. If the pattern is to take several ports at the same time, teach the AI to do so, instead of sending everything to the same port to end stuck in a supply trap.
@King OK, thanks for responding. Interesting stuff. I have more questions on this but I'll take them to the Vicky forum as they mostly relate to that game.
First of all, thx King that you use your precious freetime to give some official feedback to customer requests. Actually i was surprised to find it in this thread here - where i more or less only expected some more trolling.
Also actually what you state is the first official word that an 1.04 is planned for and that an (internal) timeline has been set up. I would have been disappointed if i had to purchase an expansion before we get any more fixes. Since we know a little more now, hope returns.
Admittedly i have spent hundreds of hours on testing the game and fun is not yet a term i primarily associate with HoI 3. Most of the time was spent understanding mechanics and then hunting bugs and identify failing mechanics. The only times i felt something like fun was when playing a SCW as Nationalists and Italy where i didn't care for any historical accuracy and let AI do their trips into an alternate universe. Of course the fun was over when looking at map results and what other AI nations had done in the meantime.
Regarding the AI you are actually discussing about - apart from sleepmode for US, (usually destroyed) JAP and some more nations not trying to preserve themselves i am not disappointed. Combat AI does work. I also do not have met anything being 'broken' with supply mechanics. And in my eyes it does as good as one can expect from a computer game. I have never met any unbeatable/superior AI in any game. But in this case - scripts seem incomplete, sometimes seem to shine via absence and do not yet create a consistent world to play in. After three patches, this is not the best of records.
The main problem i see in this field is the resilience of some devs to not create proper scripts. Correct me if I'm wrong - but it seems as PI is working on an entire engine and HoI 3 isn't more than just a module to it. Therefore reasonable scripts are being left out if they could not be used in a complete different game that is intended to use the same basic engine.
If this observation is at least partly correct, then i am disappointed with the policy of PI staff. I'd not say that everything must be scripted but if the product does not work properly without them - and products in the past which made use of scripts did work - then i have doubts that the recent approach goes into the right direction.
Please don't get me wrong. I do understand the importance of streamlining and accept that Pi wished to reach a mass audience for economic success (one of the main driving forces to run a business) but please do not forget what made PI get to where they are today. The outstanding success and fan attraction (i.e. unique selling proposition of PI games) was and is based on
- mechanisms that offer playing options that no other games do have
- games being worked out into deep details to create atmosphere and a believeable environment
- fairly fast responsiveness to customer requests and a feelable urge to get the job done to the best of possibility
- elements to ensure long term motivation and that also even attract new customers years after initial release
and some more secrets in your closets.
I hope for the best still and i still feel divided on my overall reception of the game (i do not want to come to a final conclusion yet) but apparently some work is left.
I would gladly purchase an expansion and further my support in this small genre towards PI keeping it alive but only if the basic game does work properly before.
So, to conclude - please continue your dedicated work to get the game in 'fighting shape'. I really look forward to see ww2 on this engine which is so promising in many regards.
For now, i wish you merry christmas and some days of relaxing and getting the batteries charged again!
Post Scriptum:
- Before 1.3 i never had any crashes. Now it crashes regularily with vanilla.
- game performance during war in the Chinas and Barbarossa still is hardly bearable on a machine that did cost me 800 bucks and was bought exclusively to meet the high hardware demands of HoI 3.
Also actually what you state is the first official word that an 1.04 is planned for and that an (internal) timeline has been set up. I would have been disappointed if i had to purchase an expansion before we get any more fixes. Since we know a little more now, hope returns.
Admittedly i have spent hundreds of hours on testing the game and fun is not yet a term i primarily associate with HoI 3. Most of the time was spent understanding mechanics and then hunting bugs and identify failing mechanics. The only times i felt something like fun was when playing a SCW as Nationalists and Italy where i didn't care for any historical accuracy and let AI do their trips into an alternate universe. Of course the fun was over when looking at map results and what other AI nations had done in the meantime.
Regarding the AI you are actually discussing about - apart from sleepmode for US, (usually destroyed) JAP and some more nations not trying to preserve themselves i am not disappointed. Combat AI does work. I also do not have met anything being 'broken' with supply mechanics. And in my eyes it does as good as one can expect from a computer game. I have never met any unbeatable/superior AI in any game. But in this case - scripts seem incomplete, sometimes seem to shine via absence and do not yet create a consistent world to play in. After three patches, this is not the best of records.
The main problem i see in this field is the resilience of some devs to not create proper scripts. Correct me if I'm wrong - but it seems as PI is working on an entire engine and HoI 3 isn't more than just a module to it. Therefore reasonable scripts are being left out if they could not be used in a complete different game that is intended to use the same basic engine.
If this observation is at least partly correct, then i am disappointed with the policy of PI staff. I'd not say that everything must be scripted but if the product does not work properly without them - and products in the past which made use of scripts did work - then i have doubts that the recent approach goes into the right direction.
Please don't get me wrong. I do understand the importance of streamlining and accept that Pi wished to reach a mass audience for economic success (one of the main driving forces to run a business) but please do not forget what made PI get to where they are today. The outstanding success and fan attraction (i.e. unique selling proposition of PI games) was and is based on
- mechanisms that offer playing options that no other games do have
- games being worked out into deep details to create atmosphere and a believeable environment
- fairly fast responsiveness to customer requests and a feelable urge to get the job done to the best of possibility
- elements to ensure long term motivation and that also even attract new customers years after initial release
and some more secrets in your closets.
I hope for the best still and i still feel divided on my overall reception of the game (i do not want to come to a final conclusion yet) but apparently some work is left.
I would gladly purchase an expansion and further my support in this small genre towards PI keeping it alive but only if the basic game does work properly before.
So, to conclude - please continue your dedicated work to get the game in 'fighting shape'. I really look forward to see ww2 on this engine which is so promising in many regards.
For now, i wish you merry christmas and some days of relaxing and getting the batteries charged again!
Post Scriptum:
- Before 1.3 i never had any crashes. Now it crashes regularily with vanilla.
- game performance during war in the Chinas and Barbarossa still is hardly bearable on a machine that did cost me 800 bucks and was bought exclusively to meet the high hardware demands of HoI 3.
Strategic AIs are harder than you think but there are a couple of things we are working on. With regards to the UK, we are currenlty looking at not sending the entire UK home forces off to the Chanel Islands in a death glory style defence. If the UK AI can stop doing that then hopefully the UK will be a tougher nut to crack, Doomdark is on this. Regarding Japan in China there is the two part problem that the AI reinforces beachheads without regard to port capcity, Doomdark is on this too, however there is also the interesting little issue that once you sieze a port in China the supply net stops trying to draw supply from Manchuria, I have sent a save game off to Johan and he is looking into this one.