Passive AI

Fury Games has now signed with Matrix Games, and we are working together on the next Strategic Command. Will use the Slitherine PBEM++ server for asynchronous multi-player.

Moderators: MOD_Strategic_Command_3, Fury Software

Post Reply
Kharkov
Posts: 29
Joined: Fri Jun 06, 2003 11:27 pm
Location: Birmingham

Passive AI

Post by Kharkov »

Hi

Just watched through the Youtube video series and why the game looks amazing the one thing thats worrying me is the lack of aggression in the AI. It seemed really passive putting little to no response against the Axis forces. The French units didnt put up much of a resistance to the German invasion, little movement, no attacks, British Engineers suiciding into the North African Italian units....whats going on?
User avatar
BillRunacre
Posts: 4815
Joined: Mon Jul 22, 2013 2:57 pm
Contact:

RE: Passive AI

Post by BillRunacre »

Hi

This has also been covered in this thread, with some answers from Hubert there:
tm.asp?m=4162927

Bill
Follow us on Twitter: https://twitter.com/FurySoftware

We're also on Facebook! https://www.facebook.com/FurySoftware/
User avatar
Hubert Cater
Posts: 5112
Joined: Mon Jul 22, 2013 11:42 am
Contact:

RE: Passive AI

Post by Hubert Cater »

ORIGINAL: Kharkov

Hi

Just watched through the Youtube video series and why the game looks amazing the one thing thats worrying me is the lack of aggression in the AI. It seemed really passive putting little to no response against the Axis forces. The French units didnt put up much of a resistance to the German invasion, little movement, no attacks, British Engineers suiciding into the North African Italian units....whats going on?

Hi Kharkov,

The British Engineers is an odd one as normally they should be building Fortifications in North Africa so this is a possible bug. If you could point me in the direction of the video that shows this behavior, and the timestamp, I'd be happy to take a closer look.

As Bill mentioned above, the thread he linked to does explain a bit of the early war Allied AI and why it does it what it does, but I will add here more specifically that the Poles and the French are at a bit of a disadvantage (in the early war) and the AI understands this and shapes its behavior with this in mind.

For example, here are a few screenshots of some of the Polish combat options after the initial Axis turn of the game. Note, these examples are all extracted from an AI versus AI test so it also gives you an idea of what the Axis and Allied AI achieve on the battlefield after "X" number of turns.

Situation in Poland after initial Axis AI assault:

Image


Various Polish Combat Options:

Image

Image

Image

As you can see, none of them are favorable to the Polish forces and would only result in losses to them. The general AI behavior in this case, as it also understands the various combat predictions and potential losses, would be to withdraw to friendly resources or other more defensible positions. Reason being here is that this type of behavior has a better chance of increasing AI supply which improves readiness and morale and can give it the opportunity to withdraw and then reinforce if necessary if there is some respite for any weaker units in subsequent turns. However, in this theater we nudge the AI to collapse around Warsaw instead of just any friendly resource.

The summary here for this specific Allied AI example, is that despite the initial appearance of passivity, I would argue this is the correct play for the AI to make, and one that most human players would also make, as any over aggressiveness here by the AI only makes the Axis job easier as it drives to Warsaw.


Situation in Low Countries after initial Axis AI assault:

Image

The Axis AI is coded to make extensive use of rail and operational movement as well as forced march and will typically shift from the victory over Poland towards the Low Countries as quickly as possible. As seen in the above screenshot, which is from early March, it was lined up and ready to assault by late February and had attacked under clear sky conditions (it will wait for clear skies before declaring war) but there was snow. As a result its initial drive was a bit less than ideal as Brussels, The Hague, Antwerp and Liege are still in Allied hands. But this will vary from game to game depending on weather conditions and the timing of the initial Axis AI assault.

What you'll also notice is the positioning of the Allied forces in France and how the Allied AI has lined itself up along the border and along optimal defensive positions behind rivers and so on.

However, as you can see below, the Allied AI is still at a disadvantage, due to less experience, lower readiness and morale and lower quality HQ support, that most if not all of its immediate potential attacks against the Axis forces will cause more damage to themselves than to the Axis side.

Image

Image

Image

Despite this, and at this point of the Axis invasion, I lift the restrictions on UK/French AI positioning (other than the Maginot and Paris and a few others) and allow the UK/French AI to decide for itself if it should go on the OFFENSIVE or DEFENSIVE on a per unit basis. So what you'll typically see is the UK/French AI conduct some attacks, and/or withdraw, and you'll even see some British and French units rush in to fill in the gaps in Belgium if applicable.

This can be seen a little bit in the next image, i.e. the following Allied turn, where both sides have pushed forward. However, in some games the UK/French AI may only fall back and this is again because it is set to behave dynamically at this point and if the initial Axis assault is more successful and menacing, the Allied AI will respond accordingly.


April 3rd, 1940 - Weather improves, Axis AI continues to push forward

Image


Situation April 25th, 1940 - Axis AI has driven through the Ardennes, Allied positions threatened

Image


Situation May 17th, 1940 - BEF begins withdrawal, Brussels surrounded and the French may have overexposed themselves

Image


Situation June 8th, 1940 - BEF completes evacuation, French armor arrives, French situation precarious

Image


Situation June 30th, 1940 - French Army collapses around Paris via partial evacuation of Maginot

Image


Situation July 14th, 1940 - French last ditch defensive effort move their armor into Paris, HQs retreat

Image


Situation July 21th, 1940 - France has surrendered

Image


In a final summary, and despite the appearance of AI passivity, the reality is that for the Allies this part of the war is very one sided. This will hold true for either the Allied AI or an Allied human player. However, and on the flipside, it will still be a tough slog for the Axis, AI or otherwise to defeat the Polish, French and Low Countries and near the historical timeline.

Hopefully this helps paint a better picture of both the Allied and Axis AI for this part of the war, and the fact that the AI is no slouch when it has the advantage. In this case the Axis AI was able to beat back the Allied AI near the key historical dates and in this particular test did quite well as shown in the loss table below. 35 total Allied losses to 0 Axis.

Image
n0kn0k
Posts: 489
Joined: Fri Jun 22, 2007 10:59 pm

RE: Passive AI

Post by n0kn0k »

Nice read. [:)]
Biker1984
Posts: 27
Joined: Wed Oct 12, 2016 4:08 pm

RE: Passive AI

Post by Biker1984 »

ORIGINAL: n0kn0k

Nice read. [:)]
[&o] Amen to that.
User avatar
Christolos
Posts: 957
Joined: Wed Apr 23, 2014 10:45 pm
Location: Montreal, Canada

RE: Passive AI

Post by Christolos »

ORIGINAL: Hubert Cater
ORIGINAL: Kharkov

Hi

Just watched through the Youtube video series and why the game looks amazing the one thing thats worrying me is the lack of aggression in the AI. It seemed really passive putting little to no response against the Axis forces. The French units didnt put up much of a resistance to the German invasion, little movement, no attacks, British Engineers suiciding into the North African Italian units....whats going on?

Hi Kharkov,

The British Engineers is an odd one as normally they should be building Fortifications in North Africa so this is a possible bug. If you could point me in the direction of the video that shows this behavior, and the timestamp, I'd be happy to take a closer look.

Hi Hubert and Kharkov,

The video in question can be seen here: https://www.youtube.com/watch?v=Cz6WrPi ... vJ&index=8 with the British engineer unit attacking around 10:32, which leaves itself open to being destroyed around 23:05

Kharkov, I hope you don't mind that I answered instead of you...?

Thanks,

C
“Excellence is never an accident. It is always the result of high intention, sincere effort, and intelligent execution; it represents the wise choice of many alternatives - choice, not chance, determines your destiny.”

-Aristotle-
User avatar
Hubert Cater
Posts: 5112
Joined: Mon Jul 22, 2013 11:42 am
Contact:

RE: Passive AI

Post by Hubert Cater »

Thanks CC1, much appreciated and it looks like the Engineer was either trying to get to the area in front of El Alamein where it normally builds fortifications, or it thought it had a good chance of inflicting some losses. It looks like it was a 3:1 set of losses but this could also have been a 2:2 and then with the +/- 1 modifier could have been a factor. Not sure completely as I didn't run a more thorough test but either way I've made a change so that the Engineer will not attempt to fortify those frontline positions if the Axis are that close to El Alamein already and rather it will fortify its secondary lines between El Alamein and Alexandria instead and that should resolve the issue under similar circumstances.

Hubert

User avatar
IainMcNeil
Posts: 2814
Joined: Tue Oct 26, 2004 10:01 am
Location: London
Contact:

RE: Passive AI

Post by IainMcNeil »

The AI in SC is one of its biggest selling points for me. You need to look at the big picture to understand how good it is, and also understand sometimes the best move is to not move. Something many AI's often fail to understand.
Iain McNeil
Director
Matrix Games
Biker1984
Posts: 27
Joined: Wed Oct 12, 2016 4:08 pm

RE: Passive AI

Post by Biker1984 »

A smart choise if I may add, as a lot of new games focus more about the graphics and putting in more content, which might be good for the eye, but will become dull in the end if teh AI is not being able to work.
That's also why I'm very interested in this game, as this is IMO what makes a SP game replayable and keeps getting more important as content is added designing the AI is more difficult then it ever was.
The UK AI has already conviced me with the feedback of Hubert, now I'm waiting the first moves against the Soviet and how the US AI will behave in future video posts. [:)]
User avatar
Christolos
Posts: 957
Joined: Wed Apr 23, 2014 10:45 pm
Location: Montreal, Canada

RE: Passive AI

Post by Christolos »

ORIGINAL: Biker1984

The UK AI has already conviced me with the feedback of Hubert, now I'm waiting the first moves against the Soviet and how the US AI will behave in future video posts. [:)]

I agree. I just watched the Italian navy against the AI UK navy in the Med in the latest Paradogs gamer videos (Parts 10 to 12) and found the back and forth slug fest riveting!!!

C
“Excellence is never an accident. It is always the result of high intention, sincere effort, and intelligent execution; it represents the wise choice of many alternatives - choice, not chance, determines your destiny.”

-Aristotle-
User avatar
kirk23
Posts: 2890
Joined: Fri Oct 15, 2010 1:19 pm
Location: Fife Scotland
Contact:

RE: Passive AI

Post by kirk23 »

ORIGINAL: CC1

ORIGINAL: Biker1984

The UK AI has already conviced me with the feedback of Hubert, now I'm waiting the first moves against the Soviet and how the US AI will behave in future video posts. [:)]

I agree. I just watched the Italian navy against the AI UK navy in the Med in the latest Paradogs gamer videos (Parts 10 to 12) and found the back and forth slug fest riveting!!!

C

Yes the naval slug fest was interesting,until you consider that the British naval AI should never have attacked in the first placed,because it cost itself its own 2 battleships,in the next turn,I must say that was really a clever move NOT!
Make it so!
User avatar
Christolos
Posts: 957
Joined: Wed Apr 23, 2014 10:45 pm
Location: Montreal, Canada

RE: Passive AI

Post by Christolos »

Hi kirk23,

I read your postings in the "Naval Game" thread ( tm.asp?m=4163954 ) and you certainly bring up some interesting points in terms of balancing the naval game.

I don't think, however, an AI has to necessarily be perfect 100% of the time to be realistic when one is trying to model a multitude of factors that come into play such as the fog of war (and misinformation) from the AI's perspective, training and equipment, war doctrines, and the need to balance offensive versus defensive play, not to mention the fact that mistakes and/or poor judgement can sometimes factor in as well. And let's not forget the element of surprise and Lady luck too. These are not easy things to model and so I think that some level of abstraction, which takes a wide set of variables and circumstances into consideration, and for the sake of playability, also needs to be considered. Was it a clever move (or not) for the Hood and Prince of Wales to attack the Bismarck and Prinz Eugen in the Denmark straights? Was it a clever move for the Japanese (or not) to attack Midway? We all know the outcome of these famous battles and so it is easy to say now that these were not clever moves at all. Could the UK AI have performed better in the slugfest we are discussing? Yes, and this also would include playing better by not attacking, but this is easy to say now that we have seen the outcome.

C
“Excellence is never an accident. It is always the result of high intention, sincere effort, and intelligent execution; it represents the wise choice of many alternatives - choice, not chance, determines your destiny.”

-Aristotle-
Biker1984
Posts: 27
Joined: Wed Oct 12, 2016 4:08 pm

RE: Passive AI

Post by Biker1984 »



[/quote]

Yes the naval slug fest was interesting,until you consider that the British naval AI should never have attacked in the first placed,because it cost itself its own 2 battleships,in the next turn,I must say that was really a clever move NOT!
[/quote]


Kirk23 is correct in the final outcome, however I think for (naval) battles, the AI works as it was designed.
This is the limitation of how far the AI can go and the benefit of a player and it is a great improvement in comperison to earlier/other game AI. If you are going to compare the AI with a player who can evaluate and change his mind when he sees something is getting worse or the desired outcome will be better, this is indeed not the case and practically not possible to design. Even with chess in which the moves are for more static, the AI is a copy of moves by grand masters which and limited to the moves at a chess board (no luck involved).

Back on topic, the situation was that Malta was under attack.
Lets assume that the fog of war was not present and the AI has clearly knows were the enemies ships are due to Malta being surrounded.
Would it be better to have a defensive position and wait until Malta was captured or should it advance and engage the Italian fleet?

1) A human player could evaluate what is more beneficial, he would also know that there is enemy aircraft in Southern Italy supporting the Navy (because they previously attacked Malta). I have doubts if the AI is considering if its should have air supremacy to whether or not to engage. Also, I have doubts if it evaluates that there is a possible counter attack.
If you would portrait this to the land battles a similar scenario would be France. The AI is defending every city and the Uk aircarft is 50/50 strategically bombing/attacking units. In any case only attacking units and destroying them would be cause much more damage then the strategically bombing at the moment (this need to be finetuned IMO).
When the opportunity is there to focus an attack and destroy the units they will do that, however a human player would maintain a stable front and try to attack the weak and/or valuable units (e.g armor) until it is destroyed if it had a chance to do this. The AI will evaluate if all cities are defended and attack the unit(s) in which they have a favored outcome.

Offtopic: Additionally I have not seen the AI prioritize closing supply lines in cases the Germans advance more then they are capable to handle.

2) The prioratization feature. I have seen that the UK had 1-2 CV presents in the Meds, still it prefers attacking ground units in Libia. If the human would play as the UK and he would consider attacking the ships at Malta, he would surely have the CV(s) supporting them in that battle rather then keep attacking the ground units/MPP.
Additionally the AI has I believe a step protocol. For example the player knows in the previous turn if he should move a ground unit to see if there are new opportunities, the AI is calculation all the options and begins with naval movement rather then evaluating every step. I believe Hubert has explained that the AI could take up to 20 minutes with only a minor gain on the overall outcome, should the AI be improved to this kind of level.

3) In the sea there is minimal or no effect of ZOC, which means that attack and retreat is only beneficial with the fog of war. As this is more unpredictable as were ground units will operate, retreating damaged ships are vulnerable when spotted in case of a chase, which is something a human player would consider. I have serious doubts that the AI remembers that a heavy damaged unit was present in the neighborhood and should chase its to destroy it. If it is not seen during the calculation proces, the AI will not consider this as an option.

4) The AI is evaluating every opportunity again instead of looking at future outcomes. Maybe the designers can provide more information about this, but what I have seen is that damaged battleships will attack again if the opportunity is beneficial for them, instead of sending them in for repairs. Maybe the threshold for repairs is 5=< but I have seen a BB of strength 6 attack reduce to a strength 3, which is ofcouse doomed if it is spotted.
That is the reason why I had asked if the UK will send more ships to the med. If there is no threat for Sealion the ships in the Atlantic are not aiding much in the grand strategy.
The heavy damage which can be done to the BBs with other ships as you mentioned are not making this easier for the AI as this is more benifical for the counter attacker (which was in this case the human player).

In the end damage was done to both sides in which the UK was the loser in the amount of ships, however their economics are far more able to obtain these loses and the Italians had to abandon their plan to take Malta. So strategically this could be considered a draw for now. A (good) human player will always have a better outcome. Still I will monitor how this will evolve in the end.
User avatar
kirk23
Posts: 2890
Joined: Fri Oct 15, 2010 1:19 pm
Location: Fife Scotland
Contact:

RE: Passive AI

Post by kirk23 »

My point in highlighting the AI decision making regarding the battleships is this,the battleship is supposed to be your sides strongest naval unit,when in fact it is the weakest,in terms of damage it is taking ship on ship of supposedly equal strength. In a destroyer v destroyer you will more than likely see both sides take 1 strength damage each,and that is fine.What is not fine, is when a battleship v battleship inflict 6 or 7 strength point loses per attack,they should be attacking each other,on equal terms,and inflicting similar damage as is the case with a destroyer v destroyer duel.1 strength loss each.Battleships are designed to survive attacks with 15" shells,because of their heavy armour.

What is the point of having battleships in the game,if every time they leave port,they are doomed,taking stupid strength point loss per attack,even when up against cruisers its nuts.They cost a fortune and take 15 turns to build,they are not value for money.I ask you this, has anyone every seen an Army v Army combat, where both sides take 6 or 7 strength loss from one attack,because I have not in all my years of playing these games.
Make it so!
Biker1984
Posts: 27
Joined: Wed Oct 12, 2016 4:08 pm

RE: Passive AI

Post by Biker1984 »

ORIGINAL: kirk23

My point in highlighting the AI decision making regarding the battleships is this,the battleship is supposed to be your sides strongest naval unit,when in fact it is the weakest,in terms of damage it is taking ship on ship of supposedly equal strength. In a destroyer v destroyer you will more than likely see both sides take 1 strength damage each,and that is fine.What is not fine, is when a battleship v battleship inflict 6 or 7 strength point loses per attack,they should be attacking each other,on equal terms,and inflicting similar damage as is the case with a destroyer v destroyer duel.1 strength loss each.Battleships are designed to survive attacks with 15" shells,because of their heavy armour.

What is the point of having battleships in the game,if every time they leave port,they are doomed,taking stupid strength point loss per attack,even when up against cruisers its nuts.They cost a fortune and take 15 turns to build,they are not value for money.I ask you this, has anyone every seen an Army v Army combat, where both sides take 6 or 7 strength loss from one attack,because I have not in all my years of playing these games.

I 100% agree with you regarding that matter.
User avatar
Christolos
Posts: 957
Joined: Wed Apr 23, 2014 10:45 pm
Location: Montreal, Canada

RE: Passive AI

Post by Christolos »

ORIGINAL: Biker1984

ORIGINAL: kirk23

My point in highlighting the AI decision making regarding the battleships is this,the battleship is supposed to be your sides strongest naval unit,when in fact it is the weakest,in terms of damage it is taking ship on ship of supposedly equal strength. In a destroyer v destroyer you will more than likely see both sides take 1 strength damage each,and that is fine.What is not fine, is when a battleship v battleship inflict 6 or 7 strength point loses per attack,they should be attacking each other,on equal terms,and inflicting similar damage as is the case with a destroyer v destroyer duel.1 strength loss each.Battleships are designed to survive attacks with 15" shells,because of their heavy armour.

What is the point of having battleships in the game,if every time they leave port,they are doomed,taking stupid strength point loss per attack,even when up against cruisers its nuts.They cost a fortune and take 15 turns to build,they are not value for money.I ask you this, has anyone every seen an Army v Army combat, where both sides take 6 or 7 strength loss from one attack,because I have not in all my years of playing these games.

I 100% agree with you regarding that matter.

I also agree that battleships could perhaps be buffed to some extent but I was also trying to express (in my own words) the design philosophy of the AI the way Hubert described it (see: tm.asp?m=4162927&mpage=1&key=&#65533; ) , the essence of which can be gleaned from the following excerpt by Hubert:

"It wasn't easy for me to let go of the optimal scripts I had crafted, believe me!, but the arguments against it I now firmly believe to be correct. Reason being is that my thinking on what made an exceptional AI shifted from one that played purely technically well, to one that not only played well but also met majority expectations. Essentially if the AI feels more engaging and there is only a slight drop in overall effectiveness then this was probably a good thing."

In the end the UK AI did succeed in thwarting the Italians' attack on Malta (as pointed out by Biker1984 above) and this in itself is an interesting aspect to consider as well.

I terms of buffing the battleships, it would have to be done in proportion to the units strength such that a full strength battleship would not be an easy target for a light cruiser or a couple of destroyers, but would become more susceptible to torpedo attacks from said ships when damaged. Perhaps Hubert can weigh in on this as this may already be the case.

C
“Excellence is never an accident. It is always the result of high intention, sincere effort, and intelligent execution; it represents the wise choice of many alternatives - choice, not chance, determines your destiny.”

-Aristotle-
Post Reply

Return to “Strategic Command WWII War in Europe”