Page 1 of 1

Is the AI adaptive?

Posted: Thu Aug 29, 2024 7:14 pm
by ProblemChild5
I understand that the game has scripted events within the campaigns but is the AI dynamic in anyway? Meaning, does it shift its strategic and tactical objectives according to the situation? Is it able to coherently dispatch, defend, attack, hold, sense player movements?

I am mainly asking because I don't play historical campaigns just to play them, I actually like to do what Ifs and having an AI that adapts and responds to your movements and actions is extremely compelling. Especially in a game like this, where there are 100s of different scenarios that just beg you to experiment, but without the ability to have an AI that can dynamically adjust its tactics etc, is a big dissapointment.
A basic example would be Directive 21. The 1941 - 1945 Eastern front campaign, I can't imagine having an AI that plays the same way every single time.

So. Could someone tell me wether it is adaptive or if its linear?

Re: Is the AI adaptive?

Posted: Thu Aug 29, 2024 8:24 pm
by ncc1701e
AI is from 1998 meaning it is scripted by the scenario designer.

This is not HAL 9000.

Re: Is the AI adaptive?

Posted: Thu Aug 29, 2024 9:12 pm
by ProblemChild5
ncc1701e wrote: Thu Aug 29, 2024 8:24 pm AI is from 1998 meaning it is scripted by the scenario designer.

This is not HAL 9000.
Thanks for answering. Could you elaborate more? I know that the AI won't be master mind and I wasn't expecting that. Believe it or not I've sunk over 300 hours in this game and I love it. Only thing that got me questioning is when I played Trotsky's War. After long weeks of Soviet advances into Germany I finally got the initiative and started encircling Soviet armies, army after army and it was getting ridiculous. The AI wasn't really fighting back it wasn't planning anything.

So I just wanted to know more about it.

Re: Is the AI adaptive?

Posted: Thu Aug 29, 2024 9:41 pm
by Curtis Lemay
There are five objective tracks. So...a designer could encode some flexibility if he had the willpower. Few do, though.

Re: Is the AI adaptive?

Posted: Fri Aug 30, 2024 12:41 am
by sPzAbt653
Snips from past posts:

AI / PO Explanation:

For TOAW, Single Player Mode has two levels. There is Level 1, an AI, it is a written code program that is in place for every scenario and is not accessible to scenario designers. Level 2 is separate, it is the Programmed Opponent [PO] that must be programmed by the scenario designer for every scenario in order to function. Level 1 won't really do anything unless given direction by Level 2. The amount of programming for Level 2 by scenario designers varies. It can range from just enough to allow the game to function [which for most players would not be considered a challenge], to full blown 'my programmed opponent will crush you'. Level 1 has built in Variability, therefore the PO may not do the same thing each replay and may not even do exactly what Level 2 tells it to do.

CIV estimates that 95% of all games are played solo, so these are valid concerns. Not all might be interested, but the Level 2 that I talk about above is fully accessible by all players in the Editor, and therefore there is the option to set up your own PO in a scenario that doesn't have one, or to modify one that is in place.

There are Objectives [five tracks available], Strategic Bias, Formation Orders and Loss Tolerance's. Each can be set in order to adjust the PO. Scenario Designers can make it simple or can go for complexity. For example, D21 went for complexity and there are an estimated 80,000 Objectives for the Soviet side, plus events to change the Strategic Bias [Aggressive to Neutral stuff], plus Formation Orders [which have to be there anyway but in some cases they switch by event], plus Loss Tolerance's, which are the same as Formation Orders in that they have to be there but can switch by event. There is no 'book' on how to achieve desired results, so once you complete all the input you have to playtest it repeatedly to see what happens and then make adjustments. In a small scenario it can be quick and fun. In larger scenarios it is tedious, and boring, and probably never finished because there are so many variables. This is why some scenarios have no PO or PO for one side only [but most do have a PO].

I consider that the PO for TOAW does not 'think' and 'adapt', but rather it 'reacts'.

Re: Is the AI adaptive?

Posted: Fri Aug 30, 2024 10:14 pm
by ProblemChild5
sPzAbt653 wrote: Fri Aug 30, 2024 12:41 am Snips from past posts:

AI / PO Explanation:

For TOAW, Single Player Mode has two levels. There is Level 1, an AI, it is a written code program that is in place for every scenario and is not accessible to scenario designers. Level 2 is separate, it is the Programmed Opponent [PO] that must be programmed by the scenario designer for every scenario in order to function. Level 1 won't really do anything unless given direction by Level 2. The amount of programming for Level 2 by scenario designers varies. It can range from just enough to allow the game to function [which for most players would not be considered a challenge], to full blown 'my programmed opponent will crush you'. Level 1 has built in Variability, therefore the PO may not do the same thing each replay and may not even do exactly what Level 2 tells it to do.

CIV estimates that 95% of all games are played solo, so these are valid concerns. Not all might be interested, but the Level 2 that I talk about above is fully accessible by all players in the Editor, and therefore there is the option to set up your own PO in a scenario that doesn't have one, or to modify one that is in place.

There are Objectives [five tracks available], Strategic Bias, Formation Orders and Loss Tolerance's. Each can be set in order to adjust the PO. Scenario Designers can make it simple or can go for complexity. For example, D21 went for complexity and there are an estimated 80,000 Objectives for the Soviet side, plus events to change the Strategic Bias [Aggressive to Neutral stuff], plus Formation Orders [which have to be there anyway but in some cases they switch by event], plus Loss Tolerance's, which are the same as Formation Orders in that they have to be there but can switch by event. There is no 'book' on how to achieve desired results, so once you complete all the input you have to playtest it repeatedly to see what happens and then make adjustments. In a small scenario it can be quick and fun. In larger scenarios it is tedious, and boring, and probably never finished because there are so many variables. This is why some scenarios have no PO or PO for one side only [but most do have a PO].

I consider that the PO for TOAW does not 'think' and 'adapt', but rather it 'reacts'.
Thank you so much for this detailed explanation.

Re: Is the AI adaptive?

Posted: Sat Aug 31, 2024 12:10 pm
by ProblemChild5
sPzAbt653 wrote: Fri Aug 30, 2024 12:41 am Snips from past posts:

AI / PO Explanation:

For TOAW, Single Player Mode has two levels. There is Level 1, an AI, it is a written code program that is in place for every scenario and is not accessible to scenario designers. Level 2 is separate, it is the Programmed Opponent [PO] that must be programmed by the scenario designer for every scenario in order to function. Level 1 won't really do anything unless given direction by Level 2. The amount of programming for Level 2 by scenario designers varies. It can range from just enough to allow the game to function [which for most players would not be considered a challenge], to full blown 'my programmed opponent will crush you'. Level 1 has built in Variability, therefore the PO may not do the same thing each replay and may not even do exactly what Level 2 tells it to do.

CIV estimates that 95% of all games are played solo, so these are valid concerns. Not all might be interested, but the Level 2 that I talk about above is fully accessible by all players in the Editor, and therefore there is the option to set up your own PO in a scenario that doesn't have one, or to modify one that is in place.

There are Objectives [five tracks available], Strategic Bias, Formation Orders and Loss Tolerance's. Each can be set in order to adjust the PO. Scenario Designers can make it simple or can go for complexity. For example, D21 went for complexity and there are an estimated 80,000 Objectives for the Soviet side, plus events to change the Strategic Bias [Aggressive to Neutral stuff], plus Formation Orders [which have to be there anyway but in some cases they switch by event], plus Loss Tolerance's, which are the same as Formation Orders in that they have to be there but can switch by event. There is no 'book' on how to achieve desired results, so once you complete all the input you have to playtest it repeatedly to see what happens and then make adjustments. In a small scenario it can be quick and fun. In larger scenarios it is tedious, and boring, and probably never finished because there are so many variables. This is why some scenarios have no PO or PO for one side only [but most do have a PO].

I consider that the PO for TOAW does not 'think' and 'adapt', but rather it 'reacts'.
I wanted to also ask. Since you mentioned that the AI is reactive. If I were to do an unhistorical approach to D21 or for example Ukraine on fire 6.0 and did the exact opposite that was told level 2 PO. How would that turn out? Because the way I understand it is that I have to basically play a certain way to expect a certain reaction?

So If I were in D21 for example or any other scenario, and I moved all troops from Army Group South to Army Group North (I know this is ridiculous) and told Army group Centre to defend Army group South's previous positions, and I decided to use all the troops towards Leningrad, the AI isn't to think react, "Oh he committed 2 army groups to Leningrad, I have to counter this" ?, It needs to be specifically programmed and triggered? Surely there's gotta be some Independence from the AI?

The reason I am asking all this is because I myself am interested in designing a scenario at some point and having to hand hold the AI and telling it to do specific things here and there just seems to make it feel too linear and static. It would be cool to be able to rewrite history and perhaps see hypothetical situations unfold based on each side's actions.

Re: Is the AI adaptive?

Posted: Sat Aug 31, 2024 8:08 pm
by sPzAbt653
Well, you have to program the opponent, so there is no way around holding its' hand. Therefore, when you design your scenario, you will program the PO as you see fit. Then, when others play your scenario and do different things, you will see the results and most likely will adjust the PO. This is the way it usually goes.

Whatever you do, it will always be a PO. If the PO's human opponent does ridiculous things, you may expect ridiculous results, right? But in one case that you put forth, if the Axis directs two Army Groups on Leningrad in 1941, could any human opponent prevent its' loss? I think not, therefore do not expect a PO to do any better. That is actually an extreme and unfair example though. The Axis had too great a superiority in June 1941. Reverse the situation and ask if the Soviets in June 1944 concentrated all on a drive to Berlin could any human or PO prevent its' loss?

Re: Is the AI adaptive?

Posted: Sat Aug 31, 2024 8:19 pm
by sPzAbt653
... perhaps see hypothetical situations unfold based on each side's actions.
It's all very possible. For example, if you want the PO to defend against a possible concentration of force on Leningrad, then give every Soviet Formation one objective at Leningrad. Or half of them, or 2/3 of them, whatever you want. This will ensure a heavy reaction by the PO in the case of Leningrad being enemy occupied. Of course, other areas of the front may suffer.

Re: Is the AI adaptive?

Posted: Sun Sep 01, 2024 6:30 am
by cathar1244
sPzAbt653 wrote: Sat Aug 31, 2024 8:19 pm
... perhaps see hypothetical situations unfold based on each side's actions.
It's all very possible. For example, if you want the PO to defend against a possible concentration of force on Leningrad, then give every Soviet Formation one objective at Leningrad. Or half of them, or 2/3 of them, whatever you want. This will ensure a heavy reaction by the PO in the case of Leningrad being enemy occupied. Of course, other areas of the front may suffer.
And, in line with your previous comments, such an event would be a valid trigger for shifting objective tracks, "change of priorities".

Cheers

Re: Is the AI adaptive?

Posted: Sun Sep 01, 2024 8:01 pm
by ProblemChild5
Thank you all for the explanations :).