Assessment of AI? (Full Version)

All Forums >> [New Releases from Matrix Games] >> War in the Pacific: Admiral's Edition



Message

janh -> Assessment of AI? (10/18/2009 5:03:17 PM)

Hi all,

I have been eying AE in hope of a very powerful AI engine. The detail and realism of the game is fantastic, but since I simply couldn't ever commit to a PBEM game, 'cause there would be too many interruptions, I of course need a similarly challenging AI. So far I have seen that AI is overly aggressive, at least as a Japanese opponent, and looses to much force piecemeal initially. Also I got the impression that AI does often not compose fleets to well. Unfortunately most of the AI AARs went silent within the past month, so I put out more direct questions:

Right on the point, how good is the AI? Does it equally well for Japanese and Allied? How far is it able to deliver you something that could be close to something historical? I.e. if you as a human played the Japanese advance exactly the way they did, would you get something like a battle of the Solomons, Guadalcanal, or something similar to historic battles? Or would -- no matter -- what the AI have depleted it's pools by mid-42 so much that you just have perform a clean-up ops?
Is AI capable of long-term planning? Is it capable of reaction to unusual threats, i.e. say the US human player pulling CV's and a landing force into Sumatra in late 1942? Would the KB, for instance, act flexible enough to counter such events, or could you simple use the US carriers to clean out Japanese surface forces in Malaysia without any KB intervention and than turn the carriers back to CENTPAC?

Any answers would be appreciated, but of course those based less on opinion yet more on experience or even facts would be valued most...






Chickenboy -> RE: Assessment of AI? (10/18/2009 5:14:31 PM)

Hi janh,

I'm interested in responding to your questions, but not sure how. My opinions based on the facts of my experiences will be subjective. Not sure if you are looking for these.



pad152 -> RE: Assessment of AI? (10/18/2009 5:58:47 PM)

How is the AI? I think it's still too early to tell, I've played the first 3 months as Allies, then as Japan, the AI does do a better job then WITP as Japan early on. Japanese subs now hunt off the west coast of the USA and can show up anywhere. You really to need to have a screening force when supplying/resupply forward bases. The AI as allies, still seems to send ships to strange places, Sea of Japan, right pass Japanese controlled bases in the resource area.

I think the real question is how is the AI in 1943/1944? I've never seen an WITP game vs. the AI go beyond early 1943. The AI is expected to improve in the next patch so ony time will tell.






fcam1387 -> RE: Assessment of AI? (10/18/2009 7:33:30 PM)

As a Japanese player, I am interested to see, generally speaking, how the Allied AI plays the mid to late game. Any input from developers or players would be helpful.



Canoerebel -> RE: Assessment of AI? (10/18/2009 8:29:18 PM)

I have come to the conclusion that a game of this complexity is worthy of PBEM only. That is the highest compliment I can pay this game. It is a spectacular game. Do whatever it takes to find a human opponent.



El Savior -> RE: Assessment of AI? (10/19/2009 7:03:03 AM)

Before I started over GC it was sep 1942. I played allied vs. AI Japan.

I think AI acts too aggressively. You will be surprised how fast it makes it moves. Example assault to Noumea will appear early months of 1942 - in my case even before AI tried to take Guadalcanal or Port Moresby. These all-out attacks will deplete AI forces and it usually makes several attempts to take same heavily defended places. Overall AI can make some smart moves, but it has it limits.

In my game Guadalcanal and other battles happened earlier than IRL. By september 1942 I still held Port Moresby, but AI managed to land a division size force there. It was a challenging game, but I have made several big strategic errors early game (assault attempts and lost way too many ships in process).

I say AI is good and it can surprise you some time. But it acts too aggressively early war and this could lead to too easy mid and end game. At least by September 1942 I had a loads of fun against AI and it wasn't too easy.



jomni -> RE: Assessment of AI? (10/19/2009 10:51:29 AM)

Havent you heard?  The AI cheats its way to Victory.
So no matter how bull-headed his tactics are, he can still manage to give the Human player a beating.

Not historical and quite unfair... but it makes for a fun and entertaining game.

I'm playing Japanese and the AI Allied Player sent a carrier force to sink a lot of my AK's invading DEI. In March 1942.  Very smart move in my opinion.



vlcz -> RE: Assessment of AI? (10/19/2009 11:21:45 AM)

We must face that in the present technical moment an AI never can be even nearer a medium player in so a detailed and high scope game.

Even so I think is a most important feature in a game as this, real life time constraint makes solo playability a must. to people , as me , who likes take long afternoons enjoying a great game.

The current AI is quite good, for a game of this scale probably near state-of-the-art.
its [recognised] cheats are quite comprensible and understandable (althought a little annoying from time to time)), it could (and should) surely be modified for better, - but we all must remember it is NOT a human.

By personal experience and after reading much of forums AAR and opinions is quite clear that Japanese AI have one major problem escorting and protecting his transport assets, we all are liking its agresivity and this for sure overstretches his resources but repeatedly launching unescorted invasion against known defended positions should be halted by devs.



Mike Scholl -> RE: Assessment of AI? (10/19/2009 12:37:22 PM)


quote:

ORIGINAL: janh

Hi all,

I have been eying AE in hope of a very powerful AI engine. The detail and realism of the game is fantastic, but since I simply couldn't ever commit to a PBEM game, 'cause there would be too many interruptions, I of course need a similarly challenging AI. So far I have seen that AI is overly aggressive, at least as a Japanese opponent, and looses to much force piecemeal initially. Also I got the impression that AI does often not compose fleets to well. Unfortunately most of the AI AARs went silent within the past month, so I put out more direct questions:

Right on the point, how good is the AI? Does it equally well for Japanese and Allied? How far is it able to deliver you something that could be close to something historical? I.e. if you as a human played the Japanese advance exactly the way they did, would you get something like a battle of the Solomons, Guadalcanal, or something similar to historic battles? Or would -- no matter -- what the AI have depleted it's pools by mid-42 so much that you just have perform a clean-up ops?
Is AI capable of long-term planning? Is it capable of reaction to unusual threats, i.e. say the US human player pulling CV's and a landing force into Sumatra in late 1942? Would the KB, for instance, act flexible enough to counter such events, or could you simple use the US carriers to clean out Japanese surface forces in Malaysia without any KB intervention and than turn the carriers back to CENTPAC?

Any answers would be appreciated, but of course those based less on opinion yet more on experience or even facts would be valued most...




I am quite sure that if you speak to the folks on the AE team that have been trying to improve the AI system for the last year, they will be more than happy to show you how to work the "scripting process" for it. Then you can sit down and create the perfect AI for for yourself and everyone else. [8|][:D][8D]



treespider -> RE: Assessment of AI? (10/19/2009 1:10:28 PM)


quote:

ORIGINAL: Mike Scholl


quote:

ORIGINAL: janh

Hi all,

I have been eying AE in hope of a very powerful AI engine. The detail and realism of the game is fantastic, but since I simply couldn't ever commit to a PBEM game, 'cause there would be too many interruptions, I of course need a similarly challenging AI. So far I have seen that AI is overly aggressive, at least as a Japanese opponent, and looses to much force piecemeal initially. Also I got the impression that AI does often not compose fleets to well. Unfortunately most of the AI AARs went silent within the past month, so I put out more direct questions:

Right on the point, how good is the AI? Does it equally well for Japanese and Allied? How far is it able to deliver you something that could be close to something historical? I.e. if you as a human played the Japanese advance exactly the way they did, would you get something like a battle of the Solomons, Guadalcanal, or something similar to historic battles? Or would -- no matter -- what the AI have depleted it's pools by mid-42 so much that you just have perform a clean-up ops?
Is AI capable of long-term planning? Is it capable of reaction to unusual threats, i.e. say the US human player pulling CV's and a landing force into Sumatra in late 1942? Would the KB, for instance, act flexible enough to counter such events, or could you simple use the US carriers to clean out Japanese surface forces in Malaysia without any KB intervention and than turn the carriers back to CENTPAC?

Any answers would be appreciated, but of course those based less on opinion yet more on experience or even facts would be valued most...




I am quite sure that if you speak to the folks on the AE team that have been trying to improve the AI system for the last year, they will be more than happy to show you how to work the "scripting process" for it. Then you can sit down and create the perfect AI for for yourself and everyone else. [8|][:D][8D]



Mike is quite right ...Andy would more than welcome the idea...and that is why the scripts are in the Scenario Editor available to one and all to monkey around with.

All you need to get started is, in order of importance:

1. A bottle of Scotch
2. Some time
3. The ability to think in 4 dimensions and not go insane...

Just kidding about #1 & #3.[:D]



janh -> RE: Assessment of AI? (10/19/2009 4:01:27 PM)

Thanks guys, so far.
Mike, Treespider, I am fully aware of the science behind AI.  There is plenty of publications on that available about different ways to "mimic" an intelligence, and even learning ones.  I am a scientist and fill hundreds of nodes on large clusters with quantum chemical computations every day, so I know how much effort goes into programming and scripting.

That was unfortunately not my question, as you may realize by now.   That is really nasty thing mentioned by some people called "real life", that would not allow me to do any sophisticated programming on a game.  Simply not enough time available.  But I am highly interested in playing a good strategy game like this if it delivers a good experience as is.

Treespider, I take the Scotch.  With that one should grasp quickly that dimension is little more than another index on a basis.




Graymane -> RE: Assessment of AI? (10/19/2009 6:02:34 PM)

The AI will give you an entertaining game if you are just starting out. In fact, it will probably kick your butt. It is mainly scripted though, so it isn't really doing long-term planning. But it will punish you for making tactical errors. A couple of times I have had thinly escorted convoys attacked in port while loading or unloading. I've learned to always have naval search patrols and combat surface groups in my forward ports. Does it cheat to know I'm unprotected? Probably, but then again, I don't really care since it is hitting me where it hurts the most and making me think about my operations more. If you do play the AI, just pretend/understand that it has open access to your map and knows where everything is and all of your planning targets and then plan from there.



Rob Brennan UK -> RE: Assessment of AI? (10/19/2009 8:42:03 PM)


quote:

ORIGINAL: janh

Thanks guys, so far.
Mike, Treespider, I am fully aware of the science behind AI.  There is plenty of publications on that available about different ways to "mimic" an intelligence, and even learning ones.  I am a scientist and fill hundreds of nodes on large clusters with quantum chemical computations every day, so I know how much effort goes into programming and scripting.

That was unfortunately not my question, as you may realize by now.   That is really nasty thing mentioned by some people called "real life", that would not allow me to do any sophisticated programming on a game.  Simply not enough time available.  But I am highly interested in playing a good strategy game like this if it delivers a good experience as is.

Treespider, I take the Scotch.  With that one should grasp quickly that dimension is little more than another index on a basis.





Will you enjoy a game vs the AI - Yes
Will the AI suprise you from time to time - Yes
Will you spend enough time playing to cover the $/hour investment - Big Yes

enought said ..

Go get it my man , and enjoy.



vils -> RE: Assessment of AI? (10/19/2009 9:33:31 PM)

The AI is above par as for games in general id say, but once you learn the game you won't have problems anymore, that's why you should launch the game immediately to get the most out of the major campaign. The AI can't think of flanks nor longterm planning but it's still better than useless :-)

Most definately the AI would be harder to beat the shorter the campaign is, because once time starts to addup the more the AI stupidfy (as in any other game i think)






AW1Steve -> RE: Assessment of AI? (10/19/2009 9:50:32 PM)

It's a training device, and compared to other AI's it's brilliant. Yes it cheats. So what? It's a bunch of ones and zero's for crying out loud, it needs every advamtage it can get! Yes it's very agressive. And that's good, because it's unexpected. It forces you to play conservatively, and cautiously. Or it will whip your fanny. For those of you folks who scream that "It's too tough" , ask yourselves, aren't you really just upset that you got mauled by a brainless entity? Isn't it a little like being beaten up like a girl?

Seriously, if the AI completely fries your brain, try it on an easier setting. Then build up. Don't start out on 11. Face it, in this game you are a recruit. Everyone who has had military service knows that drills and training should always be tougher than real . All the AI is for is to 1) train you up for a PBEM or 2) take the place of a human being because for some reason none is available or convienient.

The more the AI kicks my butt, the happier (but more frustrated) I am. Then I'm ready to take on a real player!!!!!  "Oh Bobo"...."Care for a PBEM?" [:D][:D][:D]



Page: [1]

Valid CSS!




Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI
0.640625