more detail in unit creation and in army, navy, and air organization. There are a
number of lesser detailed WW2 simulations and I see no reason for HOI to follow
a simpler is better path. I would like to see some optional variations in leaders
with some having "timid" traits (Lucas, Fredenhall, Monty) who won't attack until
the odds are massive in their favor. More correct unit details (get rid of the US
naval aircraft as interceptors! LOL!).
Need to pick up GalCiv again sometime...Good Game, I had the most epic home made race of Space Dwarves!
Wminus what you imply of 'more complexity' => 'the less the AI can handle', is just as false as the implications that you say others make. Having been a map scripter for several RTS games the deal with making a 'good AI' has bugger all to do if the AI can use features or not, but rather the ability for a programmer to program some AI behaviour.
For instance in the Sudden Strike series of games it was comparatively difficult in the older games to program a map AI that would actually be smart about how it played, because the system of programming was a spreadsheet cell array and the logic commands went as far as 'Is Unit [1] in cell [A1]?' Yes? 'Now move Unit [2] to cell [A1]'. I over simplify, but that was essentially the case.
Compared to a game like C&C Generals, which gave programming modules for counters, dynamic areas, unit status etc. etc. and you now have all the tools a programmer needs to program some smart behaviour into your units. Indeed if you chuck the tools out to the community then there is a plethora of programmers who will create the 'mod updates' to this for free.
The point; it is the AI framework that delivers good AI behaviour, not game features.
With HoI what you've likely got is a poor AI framework that is trying to do 9 out of 10 things from some kind of hard coded behaviour set, with the game designers themselves only having access to a few variables that they can alter without having to go back into the very game engine.
What is needed in terms of the AI, is to build it out of AI modules. An AI module is essentially a block of code designed to do a small task, but do it for a range of scenarios, and the key point is that it doesn't have a hierchary, it has a call statement. The difference is rather than the AI being programmed from;
"If this...do such and such"
It does;
"Do such and such....working?....No....Do this or that"
That way, the AI quickly doesn't repeat mistakes over and over again because somebody only coded a few If statements. The great thing about such a system is the fact it has a call statement, and this is the test for if something is working or not. In essence, the AI only fires when those "unit status's" start flagging up red.
In essence for a game like HoI you would have an AI that tries to match the stance it has been given, but if 7 out of 10 attacks fail for arguments sake it fires a construction module that looks at the current enemy forces and makes an estimate of what its forces need and puts it in the queue, this in turn fires the 'let's be patient' commander module, which takes the AI down an offensive stance to only conduct 3 in 10 attacks that it was before. The units requested get built and get sent to the front line. If the AI now starts winning 7 out 10 attacks it ramps up the number of attacks it makes.
In general thus creating a feedback loop based from a sequence of modules.
The downside with such AI design is that the AI will become 'too meta' in that you pitch one AI like this against another and you'll quickly find they devolve into the 'best strategies and builds', and very quickly stagnate in a common theme. For a game like HoI this can be very bad because there are alot of AIs competing with each other, and quickly they will fall into that trap.
However one can code in AI bias to prevent this occurring to a huge degree, but slowly this leads to a hierarchical AI tree with some of the AIs behaviour 'hard coded' so to speak.
Unfortunately HoI is not a game for a dynamic AI like this, because it aims to reproduce historic events. You can see it clearly with all the events and decisions in the game to force the AI into taking specific behavioural paths. This is intentional because a fully dynamic German AI is going to look at the Soviet Union and go "Sh*te! I'm not invading Stalin’s backyard without nukes and America on my side" and thus it will sit there, do nothing, and people will complain that AI Germany doesn't attack. Not because its AI is crap, but because its AI has decided the best decision is not to take the historical route...
This is why asking for a 'master AI', yet at the same time wanting a predefined historical grand strategy are two mutually incompatible ideas. Indeed, when has anyone played a Civilisation game where history was anything like our own...? No, its because the AI there is dynamic and 'free to choose its own path' and does so. But if Belgium was all of a sudden annexing Holland and then France a la civilisation style in HoI, you'd quickly think WTF?
Which is why, I rest my case Wminus, that your asking for something that is not in the sprit of Hearts of Iron, and outside its scope of concept. Stop whining.
Cybver: indeed, that it is a brilliant rule.
I never implied such a thing, read my post again. I specifically said that most new complex features would require much AI programming, which is hard to argue against.Wminus what you imply of 'more complexity' => 'the less the AI can handle', is just as false as the implications that you say others make. Having been a map scripter for several RTS games the deal with making a 'good AI' has bugger all to do if the AI can use features or not, but rather the ability for a programmer to program some AI behaviour.
As for your other comments: I don't know about module AIs, actually I know little programming (Computer Science isn't something I'm planning on studying.. So sry if I misunderstand you), but it's rather obvious that the vast majority of the HoI AI must be of the scripted type. Just like the CIV5 AI (scripted AIs already depend allot on randomness, btw). I mean, I don't know about games like Sudden Strike (I never heard about this1 either), but if the HoI3 AI went with this "try and fail" approach, it'd just fail 99,9% of the time.
It's fair that people don't want the game to be made simpler and like the complexity hoi3 gives.
But surely we can all agree that spamming Carriers isn't the most fun thing?
I agree with the OP in that there needs to be more incentive to build a wider variety of units.
It always feels weird to me playing the UK that I forgo Strategic bombers for Tacs or something else, there's a reason why all these different combat units existed in the first place.
Right now I feel I have no need to build Battleships or Battlecruisers at all, its just Carriers Carriers and more Carriers.
As you can tell I'm pretty biased. The naval war is totally unfun for me atm. :/
Gensui Yamamoto, I'm not sure why it has to be all one way or another. The German AI, for example, should have certain parameters on a general course to follow in order to get the historical events going, but there's no reason it should not have a more adaptable and dynamic AI for actually running the wars it starts or deciding what it builds and what it researches.
Exactly. Besides, without the WWII the game is pretty pointless, so it's a non-issue.Originally Posted by Flayer92
Also, who said that the AI has to follow the same path every time? Maybe once in a while the Germans should try the Soviets First strategy or maybe they should try to delay the war until 1941/1942 and create a big navy. For those who want to play against the same AI every time a switchable "Historical AI behaviour" option could be available.
Perhaps you have edited some of your posts to read better. However you do very much imply 'more complexity' => 'the less the AI can handle' => 'less strategy' in inverted commas with statements like these;
...It goes like this: more complexity, with an AI able to handle it => more strategy...
...I'm merely saying that reducing complexity can result in a more strategy-demanding game...
...The AI could design decent ships, but it won't be able to be anywhere near the capabilities of a human. This would give the human player just another advantage, just like most other complexity additions...
My points were;
1. That complex features, do not imply that you need more programming or needing a 'better AI'. The key thing is the structure that surrounds the AI code that becomes important. When you think of the Civ games for instance, they are/were very complex games with a good enough AI to give you a challenge. Complex and good AI, because the game was designed with an AI focus. Then you can compare it to something like the Vanilla Command & Conquer series with a simple game, with effectively a rubbish AI to go suit. This is because the game designers while they implimented a great AI structure didn't take it the full hog. Indeed taking that line is what the Supreme Commander people did, and expanded on the same kind of structure with a modular more dynamic AI. Anyhow I digress.
2. Asking for a scripted, but dynamic AI. It's a complete contridiction of purpose.
Indeed you can't 'mix the two' ideas because then you end up with an AI that does silly or stupid things. Which might be an AI Germany that having 'won' its 'scripted wars' then goes on to declare wars on all the other nations of the world just because its dynamic AI knows it can win them all. Then you have an issue with having to beef up the dynamic AI to do one set of things, and beat it down again to do another. At some point in the process stuff gets left behind leading to clunky behaviour. Indeed, that is exactly what HoI3 is like. Because the game designers have some AI code that matchs with the game engine, and then they go and lump on a load of 'events' to match history.
The result is an AI that tries to do its 'scripted features' but then gets totally screwed over by the fact that its being prevented from doing so by some internal dynamic code, or vice versa.
A case in point would be naval invaisons. Here we have the 'scripted code' that pushes two nations to war, but then the code doesn't work for the AI to work out when/where to do a naval invasion. Or it does a naval invaison, like the UK, but at a time/without the resources to sustain it because it wasn't a scripted event!
HoI3 is a great case study for why mixing these two branches of AI design doesn't work, and why it is a clunky buggy game. Some programmers are trying to script behaviour in, while others are trying to code behaviour in, and it doesn't gel well.
3. Finally, if you want a streamlined, balenced, combat orientated unit roster. Then go play an RTS; historically Battleships and Battlecruisers were on the way out in favour of the Carrier, therefore carriers should be better, and a gamey player will play with that in mind. But somebody who wants to do a bit of a historical sim, is going to be preaty cheesed off by their carriers not being as effective as they should be, or tanks rolling over everything left, right and center, as some examples.
Many people don't play HoI because it is a 'balenced game' they play it because is a grand stratergy set in the era of WWII, and they expect it to be a fairly accurate simulation of WWII. In general, they are not looking for a sandbox game loosly based around WWII like half of all the other games out there, they are looking for something that tries at its heart, to be the most accurate representation it can be.
We don't want 'WWII Chess', we don't want 'Supreme commander without the sci-fi' what we want is the weapons, nations and leaders of WWII at our disposal, fighting WWII, with us allong for the ride proding the input here and there.
EDIT 4. Final point I might make, is that potential strategies are always expanded by additional content, not less. That is because stratergy doesn't revolve around making moves, and working out the best moves. But its about putting a combination of moves together. Games like Chess have far less strategy than say Go, because even though both have simple rules and 'low complexity' with Go the board complexity increaases as the game is placed compared to decreasing with chess as pieces are captured. You can then through poker into the works as an incredibly complex game (indeed infinitly complex due to the 'infinity of people to play it with'), with just as much strategy as both of Go and Chess put together.
http://en.wikipedia.org/wiki/Strategy_%28game_theory%29
Last edited by Gensui Yamamoto; 05-05-2012 at 20:22.
I disagree. Things happened IRL for a reason - most admirals still believed that BBs were the kings of the sea and the game should reflect that. They didn't want to make life harder for themselves, they genuinely thought that BBs were better and many experiences from the previous years supported such claims. If the player has no reason to build BBs, then why should he do it? Why should the AI do it? There should be mechanics in place in order to encourage/force the player to build BBs. Better early doctrines for BBs? No CV-oriented ministers? Build limits for CVs because of the opposition within the Navy? The possibilities are numerous - there are many ways of beating hindsight. If somebody says: "you should RP your behaviour in order to make the game challenging", then for me it's an indicator of bad game balance.3. Finally, if you want a streamlined, balenced, combat orientated unit roster. Then go play an RTS; historically Battleships and Battlecruisers were on the way out in favour of the Carrier, therefore carriers should be better, and a gamey player will play with that in mind. But somebody who wants to do a bit of a historical sim, is going to be preaty cheesed off by their carriers not being as effective as they should be, or tanks rolling over everything left, right and center, as some examples.
But Cybvep, you know as well as me WWII was fought with unbalenced armies, with unbalenced weapons.
Therefore calling for 'unit balence' is also a contradictory notion to the spirit of the game.
Indeed, there are many way to script the game in such a way to reflect the generalmanship of the times, and thus constrian the player and there is room for improvement for HoI on that score (I might suggest 'advisors' who if you don't do what they say, slowly cause party support and dissent to rise) to mirror the common thought of the times.
This would be the preferable opition, but its not really 'balence' in the terms i was talking about in the post before; unit balence. This is 'tech tree' balence.
I didn't say that BBs should be as good as CVs. There are thousands of ways of achieving balance and probably even more ways of screwing it. If countries are building 251501515 CVs and everybody suddenly thinks that tanks should be used in spearheads in 1936, then the game cannot seriously be called a "simulation" or even a realistic wargame. If I have to force myself to do the bad decisions, then it's a bad game. Maybe I should hit the "Surrender" button as well?
1. You're overly abstracting the point here. The Hearts of Iron 3 AI-controlled nations do not perform as well as they ought to, and whether it's the AI itself or the structure around it, something needs to change to make it do that.
2. Those examples don't show that you can't mix dynamics with scripted, they just show that the dynamic parts are poorly designed. The British AI can't manage its resources correctly, neither can the German. This doesn't show they should be more scripted at all. The German AI should be capable of figuring out how to do an amphibious invasion when it has all the elements it needs. Likewise the British AI, having estimates of the German army's strength should be capable of figuring out when it needs to hold back. The AI is not smart enough. The AI will keep pouring troops into an amphibious landing well past its ability to support them, because it can't figure out the right balance between how much strength it needs and the limits of its ability to supply. The Italian AI in North Africa will always prioritize Tunisia over Libya because (I suspect) Tunisia is closer to the enemy capital. The AI needs to have a better grasp of how to wage the wars it is in.
3. The carrier situation in the current game is not representative of how they were used historically. In World War II, there were plenty of battleships that could match carriers for speed, and in fact most carriers that were sunk were lost to submarines.
Last edited by Flayer92; 05-05-2012 at 22:04.
I really think that next game must be designed for multithreading only. The only way we are going to get good speed and a better AI and more modablity is to have it run on faster multi-core processors. I know that this limits the number of potential customers, but the customers need to realize, the more things not hardcoded, the more need for multithreading to game. Like I said before, the game mechanics has room for improvement, but the AI and the ability to do things like get your allies to help you do Operation Overlord even if it is not in Normandy. If allies really can't work together, then the whole premise to this game doesn't work. Hard coding of goals in the AI hurts the game. Its too predictable that way. Wouldn't it be neat if the AI as Germany decided to attack france first then Poland/USSR every once in a while, may a 50% chance to take Norway. 30 for Sweden etc. it needs to be a little less predictable. Does it matter the order it takes countries out in the beginning, probably not.
Ammerica and All the Troubles II AAR a Dies Irae: Stars and Stripes and FtML Comparison AAR
http://forum.paradoxplaza.com/forum/...1#post13667536
haha so many TLDR posts.
1. AI needs most improvement, low budget game means it will be a slow process release to release but time will only make things better until ai out performs even the best humans.
2. tech should be more realistic (but equally or more complex), ie set percent of research intelligence to researching new small arms weapon (with certain efficient sweet spot), variable/kinda random time to finish research instead of exact to the day (possibly setable as deadline) and different historical guns have different strengths instead of all 1940 level small arms of all factions being same strength. Pretty much think of how technologies were developed IRL and make that in the game for each faction depending on government type. a lot of event based technology gains that are historically accurate.
3. button to assign all leaders to army with best generals in highest positions and then go down to the worst being in charge of smallest divisions, makes reorganizing leaders before war much much easier.
generally i don't like a system that auto puts all best generals in best divisions down to
the worst in the worst. seldom happened that way. generals were only relieved after
they screwed up in many cases. A good division with a mediocre commander can do
well and often did. more fun to let the dice roll where they may in that respect. The
AI has enough going against it as it is.
I think the question of more or less complexity has been answered by people in the thread - let those who want to do it but let everyone else turn on a historic default that runs it for them. If they want the historic espionage or diplomatic results, fine. Or you could choose a group of units and have them automatically organized into an army.
Here are a few suggestions for changes:
1) Each division has "frontal area" that they are responsible instead of provinces. This area would have a minimum and maximum size. A group of divisions could be chosen and then two points and they would be automatically evenly distribute along that front. Any 'extras' would automatically be made reserve units also evenly distributed. This would allow crowding the eastern front or setting 'pickets' in the soviet border with china for instance. As divisions withdrew or were wiped out the other division would have to adjust thier front or a hole would form (mode set by user).
2) Related to this any study of history knwos that certain locations are more critical than others, particular for transportation or things such as crossing rivers. In the game now it does not appear to make much difference - anywhere long the river is the same. This may strictly be a scale issue but it would be nice if significant geographic locations, like the Fulda gap, were addressed in the game.
3) As far a weapon deveopment goes, I would prefer to pick a specific configuration, like a 75mm gunned medium tank, and set that as a development goal and let my researches go. I hate having to step through the historic sdequence when I know what I really want. I also find the date-dependent development frustrating. I have read that if Germany concentrated on jet engines when they were first developed in the 1930's they could have had wings of jet fighters by ay1941. Now I know that leaders did not always get what they wanted, but some random misses could make the game interesting also.
4) The automatic assignment of resources during war can be very handy - I liked the feature though at times was frustrating.
5) I am sure people have already mentioned the completely unreleastic system of influence in HoI3. As Germany who had only declared war on Poland I had 7 historically neutral countries declare war on me. Not that it was inconvenient, just a huge surprise.
Hope some of this helps
That just makes it easier to organize instead of unassigning all and then manually having to reassign every single one. The best at the top is just a simple rule to get the leadership close to where they need to be, you could also make them in better roles depending on how many soldiers are under that command so that your best general is in charge of the top of your biggest front. This would make reorganisation at the top easier since I doubt that many people custom choose the level 2 generals they want in one of the hundred divisions they have, it's kinda redundant for non specialist divisions.
Cannot possibly be too complex, MOAR COMPLEXITY.
1- resource management: individual resource production, intermediate products, finished products, consumer goods, food, etc... Transportation network and stockpiles. This also means extraction facilities, and electrical production.
2- Improved military hardware simulation: first build the weapons/vehicles, then use the pool of what you have to train the unit. For instance: Armored brigade needs 200 medium tanks, 50 light tanks, 50 trucks, 3000 small arms. Doesn't matter where you get them (build them, buy them, lend-lease, capture stuff, whatever), once you have it you can train ye little brigade and it also takes time. Of course the brigade stats would inherit the hardware's.
3- SCIENCE! research teams need a comeback, the US managed to get da Bomb first because they had the Eisteins, Oppenheimer and others. No brain, no gain. Also unit designs so you can specialize (medium tanks being good at something vs other, for instance the Sherman had good mobility and was cheap, while pz-4 was good all-around but more expensive.
4- FINANCE! Currency exchange, foreign currency reserves, monetary policies, gold reserves, taxes, etc...
5- Revamped air power: the system of moving the entire air wing as a single body is retarded. Aircrafts should never be controlled directly but should generate air missions based on player parameter. for instance, tell the unit to support friendly troops in this and that province and the airwing will generate some missions consisting of a fraction of its force to do just that.
6- Ground combat: re-org timer gone, replaced by the good old organization loss. The top 50% org should regenerate much faster than the bottom 50% so that a badly defeated unit takes wayyy longer to re-org than a unit which retreated in an orderly fashion. Artillery units to bombard ennemy in adjacent provinces without actually engaging. Otherwise the current system is interesting.
7- Intel: COLOSSUS, ULTRA and codebreakers won the war, it should be represented more accurately. Break the ennemy's codes to see their positions and movements until they make something better. Create spy network, manage them, find the enemy's network and flip them to feed false intel. Make cute little unit markers appear on the ennemy's map while there's nothing over there.
8- Real nukes please.
Ugh, try to not make huge replies like these 1s.
1) No, I implied no such thing, nor did I edit my posts. I simply said that making a game more complex will require a more complex AI to handle it, in most cases.
2) Haha, no Civilization 3 or 4 weren't complex. That's the entire beauty of it - their gameplay was extremely simple, but could support very advanced strategies. Sure, there were a huge amount of different technologies and units, but these mostly did the same things over and over instead of adding much new. The combat, research and production systems were pure simplicity.
3) Ugh, apologies, I think you misunderstood what I meant with "scripted". I meant that for example the USSR should be "scripted" to not build bombers when its opponent has many fighters, not that USSR should always invade Finland in 1939. Regardless, coding or scripting is used by pretty much every single AI in the videgame industry. This "module AI" you were talking about simply would never work for HoI3. The AI must be properly scripted, and for that to happen the game must be simple enough.
4) Obviously chess is an extremely simple game compared to every modern strategy videogame, but it doesn't matter. Chess is more than complex enough for humanity's finite intelligence, and thus there is plenty of challenge. Thus any more complexity thus in no way provides addition challenge, and thus simplifying HoI3 wouldn't dumb it down.
5) You completely misunderstood my OP about the units. I'm in no way saying that every unit should have an equal value-for-money ratio in every circumstance, I'm saying that every unit should be balanced so that it would be useful at least in some circumstances . This obviously doesn't mean that building battleships for a high-seas fleet should be just as smart as building carriers, but it does mean that building battleships instead of carriers should be smart in some circumstances. Simply put, if built under right circumstances and used right, any unit should be able to achieve a high ratio of value-for-money.
Not quoting an entire post helps with brevity yourself Wminus. Don't call the kettle black.
I can't see why you are complaining about units then, because to be frank all the units in HoI3 do have their place for being useful. Any player worth his salt will know when to build the more specialised units. Of course the weapons of WWII were 'unbalanced' and that reflects in the concept that some of the units are in general, far superior in the general sense compared to others. All that your argument appears to be based on is; 'A gamey player will only build X, Y and Z, when there is also A,B,C in the game; y u no luv A,B,C Paradox?'.
I make a joking comment there, but the jist of it is like the punch line; if units A,B,C were not historically 'that useful' then they shouldn’t be 'that useful' in HoI either. Since units like AT, Engineers, Battlecruisers, Cav, Militia etc. do have their place for certain nations and strategies. Then they are; "...useful at least in some circumstances." which is what you claim you want... ¬.¬
This 'lack of depth' in knowledge with respect to HoI3 appears also to reflect in you knowledge of AI design. I know this will sound pretentious and patronising to you, but its the impression you make. An AI script, as opposed to AI coding are two very different things well understood in terms of getting computer to 'play the game', and were defined in previous posts by me. So why you claim to have 'changed terminology' and also not defined what you are changing it to, as for why I didn't understand something, is beyond me. It just strikes as being argumentative for no apparent reason or trying to cover up an unthoughtful post.
AI script is precisely telling the AI how to behave i.e. 'If [this] do [this]'
AI coding is uses call statements i.e. 'Find [state], [state is] do [act on state*]'
The former gives no flexibility in roll, the latter allows you to develop a set of behaviours for an infinite set of states which are defineable outside of the games intrinsic parameters.
This latter form of coding in used in nearly all modern games with a complex AI, indeed even HoI uses it from what I know from my modding. The difficulties arise when trying to mix the two together, because on the one hand your trying to look at absolutes i.e. invade Poland 1939, while on the other your trying to get the AI to balance itself i.e. build list, without knowing that the 'invade Poland' command is coming, because its not in the intrinsic dynamic AI hence not part of the [state is] part of the AI design. Hence the build plan is not tailored for that war unless it was scripted to be.
The Civ series of games has all its AI coded in the latter format, hence why it acts like it does. When the call statements are easily modifiable or added to, then you have a great AI structure, and a great game. If however you have to code every single behaviour from singular IF commands the AI will play the game in jumps in a clunky manner in a yes/no factor.
The games 'complexity' means nothing in all of this since a call statement can use as much or as little information as you want from the game, and define variables in ratios as opposed to absolutes.
*This is where logical statements act
-----------------------------------------
Right then, take tic-tak-toe on a 3x3 grid. I think off-hand there are 10 different combinations of moves that can be made, and 5 strategies for player 1, two strategies for player 2.Chess is more than complex enough for humanity's finite intelligence, and thus there is plenty of challenge. Thus any more complexity thus in no way provides addition challenge, and thus simplifying HoI3 wouldn't dumb it down.
Expand tic-tak-toe to a 4x4 grid with as your game. The number of different combinations of moves jumps up into the hundreds of thousands, and the number of strategies is similarly large.
We can take this analogy further with games like Connect 4, we introduce a larger grid, and the rule that moves can only be made cumulatively from one side and you get a vast myriad of strategies beyond that of either of the other two preceding games.
Additional complexity, provides additional challenge.
Watch some videos on game theory on youtube, there is a great series for a beginner there, it may help you understand where I am coming from on this point and why additional complexity provides additional challenge. Note: I do not relate this to AI behaviour.
I do understand that you are making specific comment to the challenge and the human mind; that we can't really explore all the possible strategies out there, so a suitably large enough number (again known to be about 60 between all players) should suffice for creating a good enough game. True. Hearts of Iron already surpasses that number, and I think it is a suitably complex enough game as is. But I wouldn't like to see it 'dumbed down' because if I feel restricted compared to an earlier incarnation, even if it did have more strategy, I'm still going to feel a short changed. After all, with a 'simulation' you want the nitty gritty details, not an abstraction. If I wanted an abstraction of WWII then I might play Axis & Allies board game.
Since HoI has goal of being a simulation as good as it can be, rather than some abstraction of WWII, then I want it to stay with all the nitty gritty, and I'm sure many others like the detail as well, even if it the game is buggy and clunky.
This was mentioned ealier.
I'd love more complexity, but the AI is already incapable. A competant AI is the most important feature I want for HOI4.
Posting Permissions
Reply With Quote