AI Behaviour

[mcaryf]

Apart from build priority and R&D is there any other AI behaviour that can be modded e.g. if I want to tell the Axis to prioritise attack through Grozny or to do Sealion etc?

Mike

[JanSorensen]

I asked basically the same question a couple of months ago - the answer was negative.

[pzgndr]

Looking through ai.txt and the ai build controls, I note there are no airborne builds for WA or Russia. Certainly they should have some after 1943? I'm wondering what else might be missing that could be added. Or might there be a good reason why they weren't included? Just curious if this ai.txt file can be improved for more historical results, even if other ai bahavior cannot be modded.

[Lebatron]

I had concidered tinkering with the AI build control. But I realized in the end it will not be worth the effort. Tweeking the builds would have little impact on gameplay since the AI is not good enough to satisfy a smart player for very long. The way I see it, the AI is good for one thing only, to be a willing participant while you learn the ropes. Once you understand the strategy of this game you will begin to critique all the AI's moves and say to yourself "boy that was stupid of it." Once your at that point its time to play against people via the PBEM option. If you don't give that a try you will tire of this game fast and you will have missed out on enjoying one of the best 'beer and pretzels'(not to hard) games ever made.

[pzgndr]

I know pbem is available, but that is not what this thread topic is about. Given the limitations of the ai behavior, there ARE some adjustments to the ai.txt file that could be made. Even with pbem games, you may want the ai to control some powers or to control research/production for yourself. So the question remains. A straightforward answer would be nice. I'm sure somebody out there has ideas for improvements. Else, I'll continue to fiddle with it myself as time permits.

[Lebatron]

I tried to be strait. What I'm saying is is that it won't matter much what you do to change the AI build control since it can't use its forces in any smart manner.

You could add airborne builds for the WA or Russia if you wished. But I'll say again that it will not change gameplay enough to make it worthwhile to force them to do so. In fact I would be willing to bet that if you forced the WA and Russians to build an airborne or two they would just turn out to be a waste of resources. I suspect that 2by3 left them out of the build cue because they new their AI was not strong enough to use them properly. So in the end adding something like airborne units into the mix would just hurt rather than help the AI's war effort. You don't want that do you?

That being said there can be ways to make Russia build better. For one you could program it to make more artillery. Thats a favorite strategy of Russian players. In this way you may get it to mimic that build strat but I see a problem with that. The difference between a human and AI making good use of this build strat is huge. Once again I will point out that the AI would not know how to make good use of them artillery, and also mismanage their protection. It would be far to easy for a German player to cherry pick all of Russias artillery with airpower. So in the end you probably shouldn't mod Russia to make it build more artillery as that would probably end up hurting it more than help.

So you see, changing this or that will do next to nothing in my opinoin to make the AI more challenging. If you want more challenge, us the option 2by3 gave us to make the AI more difficult to beat. It won't make the AI play smarter it just gives it combat bonuses and free supply.

When you tire of seeing the AI making bad moves its time to see what another human can do against you. Once you try real opponents you will quickly lose interest in playing the AI.

[Drax Kramer]

ORIGINAL: Lebatron

I had concidered tinkering with the AI build control. But I realized in the end it will not be worth the effort. Tweeking the builds would have little impact on gameplay since the AI is not good enough to satisfy a smart player for very long. The way I see it, the AI is good for one thing only, to be a willing participant while you learn the ropes. Once you understand the strategy of this game you will begin to critique all the AI's moves and say to yourself "boy that was stupid of it." Once your at that point its time to play against people via the PBEM option. If you don't give that a try you will tire of this game fast and you will have missed out on enjoying one of the best 'beer and pretzels'(not to hard) games ever made.

Which is not good in my opinion. Norm Koger's "Operational Art of War" is still installed on my disk and I consider its AI a worthwile opponent. I dare to say that TOAW seems as less "beer and pretzel" game than GGWAW, so one would, perhaps naively, expect that a newer and simpler game was going to have at least as good AI as the game from the previous century.

Instead, we got a grand strategy I go you go game that cannot run with less than 512 MB of RAM. If perhaps less programming time was spent on graphics and more on AI, majority of players that play against the AI would have enjoyed the game more.


Drax

[Joel Billings]

We spent hours and hours, months and months on the AI. Sometimes games with simpler rules but with many variations can be much more difficult to program a great AI than more complicated games where the choices made by the computer are less clearly poor choices. Also, in WaW there are choices made in strategies, some depending on difficulty level. Sometimes a player won't experience a particular AI strategy since they always play at a lower difficulty level or just get unlucky and the AI decides to do similar strategies game after game. Even so, I admit there are only a handful of strategies being used by each country instead of the many more used by humans. Doing more is just not possible given the amount of programming time we can devote to the AI of any game.

In most games, especially real time games, the "cheating" by the AI is hidden or built into the very design of the game. One of my favorite computer games of all time, Sid Meier's Gettysburg, is a great example of this. At lower play levels the human player is given subtle but very important benefits. Once you get to the higher levels, the computer is given those benefits (like unit morale that degrades slower with damage than the human player units). I don't remember the benefits being described anywhere, they just were built in. There was no way to play a game with the human and AI players having an equal playing field. We took the approach of making the "cheats" very visible to the players, and giving them complete control over each of these. Although as pointed out by many players, human play is where WaW really shines, we think that the AI is quite competitive with other game AI's, and that if you use the play levels, you can make the game very competitive for yourself. This is especially true if you choose to play only one of the 5 player slots and let the AI take the rest. We're sorry if you're disappointed in the AI. We worked hard on trying to put in the best AI we could and provide players with the options to give themselves an ever more challenging game. Compared to the AI in Axis & Allies where the computer US would build tons of infantry with no transports and no hope of getting the infantry to the front, WaW is light years ahead.

[JanSorensen]

WaW is like chess in the sense that one wrong move can mean defeat. Yet, it still have so many possible moves that its hard to find the right one. I dont think that combo makes it easy to make a really good AI.

Sure, the AI could have been better as it certainly makes its share of stupid moves - but if you play on "impossible" its certainly no mean feat defeating the AI (in particular if you play the Axis). If you then play just one power it becomes even harder as Joel also mentioned.

So, have you defeated the AI playing on impossible?

[pzgndr]

we think that the AI is quite competitive with other game AI's, and that if you use the play levels, you can make the game very competitive for yourself

Joel, thanks for your comments. All AIs are limited to some degree. I've actually been impressed by some of the things GGWAW does, like kicking me out of Italy a couple of times, so I agree that AI is competitive. My remarks were focused more on what can be done with mods to enhance solo games, given the current AI, rather than criticize it. [:)]

Reading through harrybanana's Allied Strategy Guide, there are a lot of good ideas that could be implemented in the ai.txt file regarding build and research priorities. Something similar for the Axis would be good. Maybe the AI won't care worth a hoot, but so what? If it doesn't matter, as Lebatron asserts, then it doesn't matter. At least we could get something perhaps a little more challenging, or more historically correct (like building airborne, etc.), particularly if we play solo and use the AI to handle research and production. I'm surprised there aren't some mods that already try to do this.

Another option for more challenging solo play is to create different mods for Axis and Allies. Beyond the difficulty levels which only adjust supply, transports and attack help, different mods could provide additional handicaps in the form of more units, research advances or adjustments to various regions. I wonder if placing a resource in Libya would help the Axis AI defend it longer rather than abandon it at the first sign of stalemate in Egypt. There are things that could be done. I did this with Strategic Command, with limited success. Others could try with WAW.

I do play WAW occassionally, but I do not have the time to add this to my list of projects. Strategic Command 2 and Anglo-German War are keeping me busy these days. [;)]

[JanSorensen]

You can definitely custom design one or more scenarios with loop sided advantages for the AI. I know of atleast one person that was putting quiet alot of work into just that but I dont recall seeing him around lately, alas. It likely wont make the AI player any smarter but it should make it tougher without resorting to the difficulty levels inside the game.

Heck, I might do just that once I finish debugging the damn bug thats been avoiding my detecting for far too long.

[Drax Kramer]

Even so, I admit there are only a handful of strategies being used by each country instead of the many more used by humans.

The lack of strategies doesn't really bother me. In fact, as a student of WW2, I am convinced there weren't that many strategies available in reality. What bothers me is the lack of strategy. For example, I can't grasp what German AI is attempting to achieve except for trying to be saved by the bell. It always goes for Leningrad, it always neglects Atlantic Wall and it always shuts down the North African front. And yes, it never achieves anything against Soviet AI.

Allies AI rarely tries to eliminate Italy despite suffering losses from Italian Navy. Needless to say, Italy is usually left empty by German AI once Allies land in France, but Allied AI leaves Italy alone.

This is especially true if you choose to play only one of the 5 player slots and let the AI take the rest.

My entire WAw experience is consisted of playing that way. That's how I was able to see how bad German AI is versus Soviet one, or how timid Allied AI is in Pacific. The fact that I had to invade Japan with Soviet Pacific Fleet carrying veterans from the battle of Berlin is a witness to Allied AI versus Japanese AI.

We're sorry if you're disappointed in the AI. We worked hard on trying to put in the best AI we could and provide players with the options to give themselves an ever more challenging game. Compared to the AI in Axis & Allies where the computer US would build tons of infantry with no transports and no hope of getting the infantry to the front, WaW is light years ahead.

The AI in Axis&Allies was a disaster even at the time it was released. It was definitely worse than one in TOAW. But exceling A&A was hardly a difficult goal to achieve. In my games with and against AI, I rarely did anything ahistorical, heck I never ever invaded Siberia with my Japanese, but for no awail. Germans always send a pair of tanks next to Moscow only to be obliterated. Germany never sends AA guns to Atlantic Wall to help protect all those militia it sends every turn to make up for those bombed out and so on.



Drax

[WanderingHead]

ORIGINAL: Joel Billings
We took the approach of making the "cheats" very visible to the players, and giving them complete control over each of these.

This is certainly one aspect that I really appreciate.

It does seem that there are some certain distinct activities of the AI that could be improved. It's not so much strategic as operational. It is the continuous placement of exposed underpowered units, which allows them to get slaughtered piece by piece.

I just finished a game as Axis with AI on challenging. The Japs took Hawaii in the opening move against USA, and every turn thereafter the WA put a many units as possible in the Hawaii sea zone. The problem was, it was always less than the IJN. So every turn I raked in 1-2 carrier groups plus 1-2 HFs. If the WA just waited a single turn to consolidate the forces, it would be far more challenging.

The same applies to the infamous transports. Even after you take Britain and they have nowhere to transport anything, they flood into the Atlantic to be slaughtered.

In short, it seems as though the consideration of the reprisal of the next turn is what is missing.

At any rate, I've no doubt that it is extremely tough to get a great AI going. I'd rather play people anyway, and I think that I prefer much of the effort to be spent on those things that improve such play. Like bug fixes, balance improvements, and new games (when's that civil war game expected anyway?!). We can't have it all.

[JanSorensen]

Wanderinghead,

I quite concur. In particular the placement of WA transports is puzzling.

[Joel Billings]

Gary spent many hours working on transport placement. He finally gave up when further changes weren't making things better.

[WanderingHead]

ORIGINAL: Joel Billings
Gary spent many hours working on transport placement. He finally gave up when further changes weren't making things better.

it seems like "if not frozen (by transported stuff) and not necessary for resources, then go to port" would be an improvement.