HoI 3 AI vs. E:TW AI

[unmerged(135578)]

OK, seeing two major games with goal oriented AI fail, I must ask some questions. Is this approach realistic enough or would it require too much work to make it reality? Did they just came too early, when this approach is not yet developed enough, so what we see are just "infant deseases" (as we call it)? Are there even any games with GOAP which managed to behave in some challenging way? I never understood AI programming much, so maybe my questions are stupid for the rest of you, but I am sure I am not the only one who askes those questions...

GOAP is a mature concept and there is a half a dozen games employing it. Several games have used it with superb results - especially in first-person shooters (F.E.A.R being most famous).

In my mind GOAP's biggest promise is that it allows creation of a library of actions which some or all agents can use. This makes code a lot simpler (and reusable) and it allows doctrinal uniqueness amongst agents (for example some agents may have a host of possible actions some agent will never use). I believe that these were very good reasons to employ it with E:TW. The next E:TW comparable game is the War Leaders: Clash of Nations. We shall see if they have managed to get better results.

E:TW very probably suffers from bad implementation on AI part so true value of its new engine will probably be seen with N:TW. They have now roughly a year to fix troubles in game engine to see how well new Battle AI performs.

 

[TheLoneGunman]

The is not one paradox game that they havent fixed or made better after the release date, i trusted they we do it with HOI3 too, i still play EU3 IN AN, CK DV, and Ricky, even after all these years. But the total war series has some good games to essecially the MTW 2 expansion pack, but when i learned that they were coming out with ETW i had my misgivings, not because the company was bad, but because that era is so diffacult to recreate. The strategic thinking and practice of that era cannot be capture accuratly with the total war engine, and it is a shame to say that for a napoleonic era game, L'Emperuer by KOEI for the orginal NES is still a better game, it may not have the graphics, but it is far more enjoyable. In two years time L'Emperuer will still be better than ETW, but HOI3 will be a classic.



p.s. if we want to compare these games on these forums. We can cause we are part of the paradox community too

Well in all fairness EU:Rome is still in desperate need of at least one more patch. However, Johan has indicated they may make one for it if they have the time available, so in the meantime we'll hold our breath.

 

[telesien]

GOAP is a mature concept and there is a half a dozen games employing it. Several games have used it with superb results - especially in first-person shooters (F.E.A.R being most famous).

In my mind GOAP's biggest promise is that it allows creation of a library of actions which some or all agents can use. This makes code a lot simpler (and reusable) and it allows doctrinal uniqueness amongst agents (for example some agents may have a host of possible actions some agent will never use). I believe that these were very good reasons to employ it with E:TW. The next E:TW comparable game is the War Leaders: Clash of Nations. We shall see if they have managed to get better results.

E:TW very probably suffers from bad implementation on AI part so true value of its new engine will probably be seen with N:TW. They have now roughly a year to fix troubles in game engine to see how well new Battle AI performs.

That's good to hear. But as I already said, I know almost nothing about AI programming, so I have no idea how different is the AI concept in shooters and strategies. Do they have basicaly the same level of complecity or is one of them more complex?

 

[unmerged(135578)]

That's good to hear. But as I already said, I know almost nothing about AI programming, so I have no idea how different is the AI concept in shooters and strategies. Do they have basicaly the same level of complecity or is one of them more complex?

Shooters have generally very well defined behavior (they try to kill you) and single person's activities in combat are fairly simple to model. Currently the difficulty lies in making a mob function as a infantry squad. I am not sure any group has yet managed to do that well (this is the holy grail of FPS - to make an infantry squad that will make you pee your pants as it uses fire and movement to roll over your character while moving from cover to cover).

Strategy games have huge array of possibilities and thus number of possible tasks will go from dozen or so basic national level priorities to thousands of tactical level tasks (for an idea look at "Universal Joint Task List Manual" of US military what modern military game tasks should encompass). However, the tasks are themselves fairly straightforwards and many are often long-term. I believe that planning what tasks should be used (and what subtasks each of them should require to succeed) is the holy grail here.

To return to HoI series of games I'd be very curious to learn more of how the various AI's work in those games but so far Paradox has been playing her cards very close to her chest.

 

[telesien]

To return to HoI series of games I'd be very curious to learn more of how the various AI's work in those games but so far Paradox has been playing her cards very close to her chest.

Thanks for great info. I just didn't understand the last part. I thought that "opened" AI in modable lua files and the resulting slowness of the game is one of the hot issues here. Shouldn't it mean that you can view how AI works?

 

[unmerged(149383)]

If it bothers you that much, PM a mod about it.

They're not Gods you know, I don't think they read every single thread on the forum from top to bottom.

I wont bother them with PMs as it doesnt bother me that this thread is up here, just that others seems to be closed much faster. I would anyway say that this one has served is "purpose". And as people have stated, does bad AIs in other games make the one on HOI better? They will both get fixed, and for that I am glad.

 

[unmerged(135578)]

Thanks for great info. I just didn't understand the last part. I thought that "opened" AI in modable lua files and the resulting slowness of the game is one of the hot issues here. Shouldn't it mean that you can view how AI works?

Yes, looking at lua scripts allows you to look at several AI functions and potentially tweak AI functionality in the future. However, I do not have the Hoi3 yet as I am waiting for most bugs to clear out (and then buying it to my greatest love - Me).

Lua scripts are generally much much slower than doing the same thing through C++ but I believe that as better option in the long run as it allows modders to tweak AI on their own. I do believe that problems with speed are related primary to memory leaks and secondly to the frequency lua scripts are called.

I believe that most problems so far are related to:
1) memory leaks slowing and ultimately stopping the game.
2) weather code having erroneous calculation within C++.
3) supply line routing algorithm not functioning properly over other nation's territory
4) trade deal calculations being sent originally once an hour (I'd think once a day could be a good compromise - perhaps similar exercise could be done with many other functions?)
5) building unnecessary transports (very probably an error in lua file but I cannot remember where the post pointing it was)

Once these are fixed the attention probably turns towards tweaking various modifiers to find the "right" play balance within diplomacy, unit statistics and so on. This is also a good news to me as I am avidly waiting for Victoria 2 and Hoi3 engine is used to it - so any fixes made to Hoi3 model are steps towards the ultimate prize: Victoria 2.

 

[unmerged(44659)]

If you asked me before hoi3 to compare CA and PI AI i would be laughing. Now its stuck in my troat. This is by far the worst "big" release from PI. We are comparing the dirty CA and PI, then they have sunken low. Shame on you Paradox.

 

[unmerged(26764)]

There is not one paradox game that they havent fixed or made better after the release date, i trust that they will do it with HOI3 too, i still play EU3 IN AN, CK DV, and Ricky, even after all these years.

I don't know if I agree with that.

EU:Rome was not really "fixed" until the expansion. People who only bought the original game never got a fully working game. For example, I'm pretty sure the relgion system never fully worked for vanilla buyers. They certainly never got a compelling game (even if everything "worked").

Second, it took so long for the patches and expansions on that game that most people lost interest before they came. I'm simply not okay with it taking a year to make a game I bought compelling enough to play. And the fact is paradox did not really "jump to" to fix the game, knowing how people felt. They took their sweet time on fixing it, spending resources first on new games and expansions for other titles.

So, in sum, it took customers over a year to get the game they thought they bought, and then it cost them money. That's the most recent history of paradox. It really bothered me, they never appologized for it, and its what I'm afraid is going to happen here.

What was the worst part for me was the fact that fixing the game was never really a priorty for them -- they got around to patches whenever they had a hole in their existing development schedule instead of making the time to fix what their customers deserved fixed.

 

[Rockstone]

Empire: Total War. A game which apparently everyone hates but me. :rofl:

Nope... you're not alone!

 

[unmerged(60148)]

For example, I'm pretty sure the relgion system never fully worked for vanilla buyers.

It never worked with the expansion either, it was just changed from one broken state to another.