Congrats AI (Full Version)

All Forums >> [New Releases from Matrix Games] >> Gary Grigsby's War in the East Series



Message


squatter -> Congrats AI (12/20/2010 10:28:52 PM)


As you may have noticed, I'm not much of a positive poster (I'm generally not motivated to post unless annoyed by something). When I'm happy I just play.

Just to buck the trend, I've been playing Stalingrad as Axis vs challenging AI, and I have to say that the AI is good. Now, of course, this is relative to the fact that AIs in all games are horrible (dont want to get too positive here), but the AI here is counterattacking where I am week, laying down a strong carpet where I am strong, breaking out of encirclements wherever it can, and in one case, making a counter enveloping move of its own.

Now I know it is assisted by full recon (I laugh when it makes air recon flights, as if trying to kid me that it doesnt know what's out there), but I would say at this early stage that the (Soviet) AI is as good as you could hope for in a strategy game of this scale.

Nuff said.




Timmeh -> RE: Congrats AI (12/20/2010 10:35:40 PM)

Kudos,

I on the other hand tend to be more forgiving and hope to turn a blind eye for the sake of immersion. Both eyes still wide open here, head still spinning around in circles.




jomni -> RE: Congrats AI (12/20/2010 10:35:58 PM)

Well some don't like the AI despite being good because it doesn't follow history.




janh -> RE: Congrats AI (12/20/2010 10:53:23 PM)

Think carefully about that argument....: If you had an AI that was following the doctrine of that time exactly, and, hence, making exactly the same mistakes that the Soviets or Germans made at that time, you would only need to avoid a few mistakes as a well-versed player in history of warfare and would be able to exploit these mistakes and win the game with extreme ease.  I think for most players, a game AI written "today" also needs to take into account that learning curve.  From what I can gather here at the forums (is there a demo, or will there be one???), the AI here does a pretty good job at this.

I believe myself to be a player that can stick to doctrine and intentionally repeat historical mistakes for the sake of getting a more appropriate overall experience, but also I fail often to do things that were historically, but today seem unacceptably illogical (political maybe at that time) to me.  Ideally, I would wish games to have an additional "difficulty" setting that lets AI pick its tactic, i.e. "historical/advanced hindsight doctrine", but such implementation is still beyond the level of most current AIs.  Maybe as much as graphics have evolved over the past 15 years, and are rapidly converging on an almost steady-state of "realistic movie image", I would expect that AI development will be one of the next "distinguishing factors" of better games and might make rapid progress in the next couple of years.




Schorsch -> RE: Congrats AI (12/21/2010 1:01:15 AM)

ThereŽs one thing iŽd like to say @ players looking for sthng going Historical:
GO, WATCH A DOCUMENTARY FILM [:D]




Rasputitsa -> RE: Congrats AI (12/21/2010 7:58:59 AM)


quote:

ORIGINAL: Schorsch
ThereŽs one thing iŽd like to say @ players looking for sthng going Historical:
GO, WATCH A DOCUMENTARY FILM [:D]


There's one thing I'd like to say @ players looking for a non-historical balanced game:

GO PLAY CHESS

This game is called 'War in the East', you would expect it to represent war as it was in the East. [:)]




Sheytan -> RE: Congrats AI (12/21/2010 8:19:00 AM)

A good AI is one that is dynamic and appears to think outside the box. Phew phew to gamers that complain about that. The fact of the matter is this simple, history is remade the first turn you play in any game. People that dont understand that most basic reality are doomed by limited vision.




karlxii -> RE: Congrats AI (12/21/2010 8:28:57 AM)

+1




Sheytan -> RE: Congrats AI (12/21/2010 8:31:09 AM)


quote:

ORIGINAL: karlxii

+1

[;)]




Chris H -> RE: Congrats AI (12/21/2010 8:38:53 AM)


quote:

ORIGINAL: Rasputitsa


quote:

ORIGINAL: Schorsch
ThereŽs one thing iŽd like to say @ players looking for sthng going Historical:
GO, WATCH A DOCUMENTARY FILM [:D]


There's one thing I'd like to say @ players looking for a non-historical balanced game:

GO PLAY CHESS

This game is called 'War in the East', you would expect it to represent war as it was in the East. [:)]


You can satisfy someone the people some of the time, all of the people none of the time.

Personally I get very tired of the old arguments re, historical or not. I don't care, what I do care about is a competant AI.

As far as this game goes, note the word game as opposed to simulatation, the AI does appear to be verging on competant.




Rasputitsa -> RE: Congrats AI (12/21/2010 8:40:47 AM)


quote:

ORIGINAL: janh

Think carefully about that argument....: If you had an AI that was following the doctrine of that time exactly, and, hence, making exactly the same mistakes that the Soviets or Germans made at that time, you would only need to avoid a few mistakes as a well-versed player in history of warfare and would be able to exploit these mistakes and win the game with extreme ease.  I think for most players, a game AI written "today" also needs to take into account that learning curve.  From what I can gather here at the forums (is there a demo, or will there be one???), the AI here does a pretty good job at this.

I believe myself to be a player that can stick to doctrine and intentionally repeat historical mistakes for the sake of getting a more appropriate overall experience, but also I fail often to do things that were historically, but today seem unacceptably illogical (political maybe at that time) to me.  Ideally, I would wish games to have an additional "difficulty" setting that lets AI pick its tactic, i.e. "historical/advanced hindsight doctrine", but such implementation is still beyond the level of most current AIs.  Maybe as much as graphics have evolved over the past 15 years, and are rapidly converging on an almost steady-state of "realistic movie image", I would expect that AI development will be one of the next "distinguishing factors" of better games and might make rapid progress in the next couple of years.



There is a difference between an AI programmed to have the capabilities and behaviour of historical forces and an AI scripted to follow the exact actions of particular events. I don't think that anyone here wants to exactly reproduce history in their games, mostly they will want to try different strategies.

The game will have a mixture of both of the above, but what mix suits each player. Here it gets complicated - want to play revised German forces against historical Russians, standard Germans against Russians with revised armoured doctrine, what-if Hitler/Stalin had done this or that, and so on, and so on. There is are a huge variation of possible combinations, it not surprising that the developers would find this very difficult, if not impossible. The historical leaders designed their forces, equipment and doctrine without the benefit of the hindsight we enjoy and not to suit our replaying of the events.

Perhaps when large scale modding gets under way, scenarios and GC variations will appear that will go a long way to meeting the needs of most players, it's early days yet and there is a long road still to travel. I am looking forward to what follows, this is just the start. [:)]




Rasputitsa -> RE: Congrats AI (12/21/2010 8:56:09 AM)


quote:

ORIGINAL: Chris H
You can satisfy someone the people some of the time, all of the people none of the time.

Personally I get very tired of the old arguments re, historical or not. I don't care, what I do care about is a competant AI.

As far as this game goes, note the word game as opposed to simulatation, the AI does appear to be verging on competant.


I sympathise with the sentiment, but the argument will not go away, one person's competent AI is someone else's disaster. I agree that the AI looks very good and alternative modded scenarios should give enough flavours to satisfy most tastes. [:)]




Sheytan -> RE: Congrats AI (12/21/2010 8:59:04 AM)

Well said. 1+

quote:

ORIGINAL: Rasputitsa


quote:

ORIGINAL: janh

Think carefully about that argument....: If you had an AI that was following the doctrine of that time exactly, and, hence, making exactly the same mistakes that the Soviets or Germans made at that time, you would only need to avoid a few mistakes as a well-versed player in history of warfare and would be able to exploit these mistakes and win the game with extreme ease.  I think for most players, a game AI written "today" also needs to take into account that learning curve.  From what I can gather here at the forums (is there a demo, or will there be one???), the AI here does a pretty good job at this.

I believe myself to be a player that can stick to doctrine and intentionally repeat historical mistakes for the sake of getting a more appropriate overall experience, but also I fail often to do things that were historically, but today seem unacceptably illogical (political maybe at that time) to me.  Ideally, I would wish games to have an additional "difficulty" setting that lets AI pick its tactic, i.e. "historical/advanced hindsight doctrine", but such implementation is still beyond the level of most current AIs.  Maybe as much as graphics have evolved over the past 15 years, and are rapidly converging on an almost steady-state of "realistic movie image", I would expect that AI development will be one of the next "distinguishing factors" of better games and might make rapid progress in the next couple of years.



There is a difference between an AI programmed to have the capabilities and behaviour of historical forces and an AI scripted to follow the exact actions of particular events. I don't think that anyone here wants to exactly reproduce history in their games, mostly they will want to try different strategies.

The game will have a mixture of both of the above, but what mix suits each player. Here it gets complicated - want to play revised German forces against historical Russians, standard Germans against Russians with revised armoured doctrine, what-if Hitler/Stalin had done this or that, and so on, and so on. There is are a huge variation of possible combinations, it not surprising that the developers would find this very difficult, if not impossible. The historical leaders designed their forces, equipment and doctrine without the benefit of the hindsight we enjoy and not to suit our replaying of the events.

Perhaps when large scale modding gets under way, scenarios and GC variations will appear that will go a long way to meeting the needs of most players, it's early days yet and there is a long road still to travel. I am looking forward to what follows, this is just the start. [:)]






hmatilai -> RE: Congrats AI (12/21/2010 9:40:49 AM)

Yeah, I think it's good. In my game as Soviets against Axis AI in grand campaign the turn before blizzard hit, Axis AI stopped attacking completely, withdrew to shorter frontlines and in front of Moscow prepared for defense in-depth, having 4-5 hexes full of units to prevent me from trying encirclements.

My original plan was to try encircle most of AGN between Pskov, Peipus, Luga and Lake Ilmen by capturing Pskov and cutting the rairoad north at Narva (AGN was stopped at Luga).

In the center I was hoping to cut off both panzer groups, as they had created nice bulges near Vyasma and Roslavl, but AI withdrew from these positions.

Also in the south there was a huge bulge around Vinnitsa, where one panzer groups was about to be encircled, but it withdrew from this position as well. In the south I was able to stop AGS rather easily well before Kiev, since the first turn encirclement failed for AI. Also Romanian troops are lined up before Dnestr.

I think I can only hurt Germans badly, but I can't destroy many divisions, as my tank brigades are still crap if I try to move them as a stack of just armor, so I don't have enough MPs with cavalry and infantry to try to make huge encirclements.




Sheytan -> RE: Congrats AI (12/21/2010 9:53:56 AM)

Outstanding, I havnt gotten as far as you yet. The Axis AI is really pushing me in front of Smolesk and driving towards Leningrad now. I am weak in these areas as I have concentrated on building defenses in depth in front of Moscow early on. In the south and south center ive really built a massive layered defense intended to soak up the armored spearheads and slow them towards Kiev. outside of the opening turn encircilments by the Axis AI ive managed to avoid suffering any major pockets. Its been fun so far.

Edited to add the weakness in the center was intended, I am trying to see if the AI outruns its supply lines advancing towards Moscow, if it does I have quite a bit there all set to refit ready to counter. Heres to hoping im not the sucker[:D]

quote:

ORIGINAL: hmatilai

Yeah, I think it's good. In my game as Soviets against Axis AI in grand campaign the turn before blizzard hit, Axis AI stopped attacking completely, withdrew to shorter frontlines and in front of Moscow prepared for defense in-depth, having 4-5 hexes full of units to prevent me from trying encirclements.

My original plan was to try encircle most of AGN between Pskov, Peipus, Luga and Lake Ilmen by capturing Pskov and cutting the rairoad north at Narva (AGN was stopped at Luga).

In the center I was hoping to cut off both panzer groups, as they had created nice bulges near Vyasma and Roslavl, but AI withdrew from these positions.

Also in the south there was a huge bulge around Vinnitsa, where one panzer groups was about to be encircled, but it withdrew from this position as well. In the south I was able to stop AGS rather easily well before Kiev, since the first turn encirclement failed for AI. Also Romanian troops are lined up before Dnestr.

I think I can only hurt Germans badly, but I can't destroy many divisions, as my tank brigades are still crap if I try to move them as a stack of just armor, so I don't have enough MPs with cavalry and infantry to try to make huge encirclements.





janh -> RE: Congrats AI (12/21/2010 10:06:29 AM)

quote:

ORIGINAL: Rasputitsa
There is a difference between an AI programmed to have the capabilities and behaviour of historical forces and an AI scripted to follow the exact actions of particular events. I don't think that anyone here wants to exactly reproduce history in their games, mostly they will want to try different strategies.

The game will have a mixture of both of the above, but what mix suits each player. Here it gets complicated - want to play revised German forces against historical Russians, standard Germans against Russians with revised armoured doctrine, what-if Hitler/Stalin had done this or that, and so on, and so on. There is are a huge variation of possible combinations, it not surprising that the developers would find this very difficult, if not impossible. The historical leaders designed their forces, equipment and doctrine without the benefit of the hindsight we enjoy and not to suit our replaying of the events.

Perhaps when large scale modding gets under way, scenarios and GC variations will appear that will go a long way to meeting the needs of most players, it's early days yet and there is a long road still to travel. I am looking forward to what follows, this is just the start. [:)]



Sure enough. Of course any AI in utmost instance ought to be "dynamic" and react in sensible fashion to player moves, whatever principles it is programmed to follow. A purely scripted AI, where units would just follow strictly a timeline and path, would be plainly boring -- just trap up because in the 2nd or 3rd game you'd know where they want to go. I saw to many of those games in the past.

The sense in which I use "historic" is the choice of moves or actions ("doctrine") with which AI would respond to player moves. An AI would also be limited if it were too closely following the principles of that time, say for this example if German AI here would be too attached to the "Feste Plätze" idea of Hitler, and allow itself to be surrounded by the player -- I suppose a player could easily exploit such things, and be in Berlin in 43 if AI would repeat too many historic mistakes. So having a truly historic AI, even if dynamic, would maybe not lead to be funnest experience for the player?

Therefore I think the AI design they chose here (and elsewhere) makes more sense: it is primary goal is to create a fun game.




Sheytan -> RE: Congrats AI (12/21/2010 10:19:08 AM)

Thats a matter of considerable thought because one presupposes doctrine is a linear extension of what took place. This is also a paradox because by the same token as you say IF the AI reacted predicably according to doctrine one would be able to anticipate said doctrine.

Therein lays the weakness of rigid adherence to such a process. In reality soviet doctrine for example evolved according to the opponent it faced, in other words they learned, and ultimately mastered armored warfare from the masters of this doctrine of the time.

A radical departure would be this effect over the progression of the campaign game, but wait...they(the devs) modeled this with adaptions of higher unit structure in a sense.

Frankly I think WITE is a very very good stab at simulating the War in the East and still not pigeonholing a player with constrants that may appear linear simply because of a historical byline in some book.

One other factor occurred to me regarding this, the greatest advantage the Axis had over the soviets up to 1943 was in command and control, from a material standpoint the soviets had in a number of cases superior equiptment, they simply were incapable of employing it as well as the Axis were. For example the KV1 and T34 were far superior to anything the Germans fielded until the Germans developed the Panther and Tiger at any rate. AND! That opens a entirely new arguement, the production cycles involved in building these and the numbers turned out clearly indicate that soviet production routines are actually MORE efficent, because post Kursk the soviets were turning out even more formidable armored platforms.

Frankly the soviets beat the germans at thier own game [:D]

edits for spelling, its in the wee hours for me. [>:]

quote:

ORIGINAL: janh

quote:

ORIGINAL: Rasputitsa
There is a difference between an AI programmed to have the capabilities and behaviour of historical forces and an AI scripted to follow the exact actions of particular events. I don't think that anyone here wants to exactly reproduce history in their games, mostly they will want to try different strategies.

The game will have a mixture of both of the above, but what mix suits each player. Here it gets complicated - want to play revised German forces against historical Russians, standard Germans against Russians with revised armoured doctrine, what-if Hitler/Stalin had done this or that, and so on, and so on. There is are a huge variation of possible combinations, it not surprising that the developers would find this very difficult, if not impossible. The historical leaders designed their forces, equipment and doctrine without the benefit of the hindsight we enjoy and not to suit our replaying of the events.

Perhaps when large scale modding gets under way, scenarios and GC variations will appear that will go a long way to meeting the needs of most players, it's early days yet and there is a long road still to travel. I am looking forward to what follows, this is just the start. [:)]



Sure enough. Of course AI in utmost instance ought to be "dynamic" and react in sensible fashion to player moves, whatever principles it is programmed to follow. A purely scripted AI, where units would just follow strictly a timeline and path, would be plainly boring -- just trap up because in the 2nd or 3rd game you'd know where they want to go. I saw to many of those games in the past. The sense in which I use "historic" is the choice of moves or actions ("doctrine") with which AI would respond to player moves.

Therefore I think the AI design they chose here (and elsewhere) makes more sense: it is primary goal is to create a fun game.






Rasputitsa -> RE: Congrats AI (12/21/2010 11:15:37 AM)

There is a certain amount of scripting even in real life, as the Germans could attack Leningrad and Moscow in the knowledge that the Russians must defend these cities, there will be 'must-do' actions for the Germans in their turn. [:)]




Grouchy -> RE: Congrats AI (12/21/2010 11:57:03 AM)

I'm pleasantly sursprised with the AI on normal mode (Playing the '41 GC as axis).
It tries to put up defensive lines at the right spots, using natural obstacles to slow down your advance and does quit a nice job trying to not get flanked.

Only minus is that it leaves cities like Smolensk etc when you put the pressure on it....kind of an anticlimax.
What i don't understand is the HQ's and 0-0 units that it puts into the frontline/checkerboard formations a couple of hexes away from my troops.

I read some nice AAR's about her 1941 winter behaviour, already looking out for that period...or maybe not :[:'(]




Page: [1]

Valid CSS!




Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI
2.929688E-02