The AI thread

[Leifr]

5 hours ago, Avimimus said:

This is inaccurate. The AI can evade in different directions and has at least nine evasive maneouvres (based on the wingman commands menu shown in some of the videos).

P.S. There was an issue in Great Battles which goes back to Rise of Flight, where it is possible for AI to try to turn after running out of energy and altitude (at which point it could get locked into a right-hand turn)... but that is a different issue which can only really be fixed by keeping the AI from trying to do minimum radius turn fights at low altitude. But that isn't really even an evasive manoeuvre - just a loss in flexibility that happens once you've exhausted your kinetic energy.

I think the issue is, as well, that the AI have to be told to perform the evasive manoeuvre rather than using the intelligence part of their thinking. 

Well, everything else in Korea looks to be a moderate to excellent upgrade on an old and leggy edition of Great Battles, so here's to some post-launch improvements!

[Dash,Polder]

We have yet to see the other side of the aisle when it comes to AI.  Nothing on the ground war or it's air defenses so far.  Makes one wonder why it isn't being show cased, can't be I'm the only one eagerly waiting to see it in action.  

[Aapje]

According to the FAQ and Enigma, they are going to still work on the AI, so it might still get a bit better.

[Koziolek]

1 hour ago, Aapje said:

According to the FAQ and Enigma, they are going to still work on the AI, so it might still get a bit better.

I am still working on my dancing and can honestly say - it's not getting any better

[Aapje]

Well, according to Wikipedia, a Koziołek skalny is a weird antelope that constantly walks on its toes, and does not wear ballet shoes, so it makes sense that dancing is hard for you. 😉

But lets see what the devs can make happen.

Edited June 22 by Aapje

[Reisen]

2 hours ago, Aapje said:

According to the FAQ and Enigma, they are going to still work on the AI, so it might still get a bit better.

I feel like if this was a priority for the devs they would have already done it. Crazy that so much improved other than this.

[CatteMoto]

I was watching some historical aviation videos on youtube and it recommended some gun camera footage from USAF planes over Korea. None of the victims appear to pull any "insane roll maneuvers" or really any abrupt movements as one would see online. They appeared to try to retain their energy and climb or dive away from danger, circling somewhat gently and without resorting to split S moves. 

 

The AI could certainly be improved for GB and beyond, but it seems that an unrealistic and overly "epic" single-player experience is imagined by many of the doubters. The Ace Combat series is on its eighth installment and is surely chock full of such bombastic moments. Maybe their forums could use some complaints about historicity.

Personally I will assert my own unrealistic demands and expectations after the product is launched in full at the end of the year. For example if there is no USS Pueblo Incident module I will leave a scathing review on Steam. And no I don't care if it happened in 1968

[Catch]

3 hours ago, Aapje said:

According to the FAQ and Enigma, they are going to still work on the AI, so it might still get a bit better.

That's great. Let's just not say anything official on the official forum and it will go away.

 

 

[[I.B.]-=ViRUS=-]

2 часа назад, CatteMoto сказал:

None of the victims appear to pull any "insane roll maneuvers"

Have you considered that guncams shutter speed is might be higher than 24? And what happens on guncams is just a slowmo of what really happened?

[kodamcity]

4 hours ago, CatteMoto said:

I was watching some historical aviation videos on youtube and it recommended some gun camera footage from USAF planes over Korea. None of the victims appear to pull any "insane roll maneuvers" or really any abrupt movements as one would see online. They appeared to try to retain their energy and climb or dive away from danger, circling somewhat gently and without resorting to split S moves. 

I think the World War II gun camera footage also seems similar to what you said.

[tattywelshie]

On 6/21/2026 at 9:59 AM, gecko said:

I think for the dev teams in the past, the answer is not how easy or difficult the AI was to improve, but what importance they attached to it, which was apparently not very much. Pretty much all combat flight sim development for the past two decades has heavily leaned into the multiplayer side of things at the expense of making a single player experience that was actually good.

BSR seems to not to value AI development as much either since it isn't as good as a human. To each his own, I do understand the draw of multiplayer. But let's not pretend this is impossible to improve, or hide from the widely recognized success smaller developers have had in this area just because you watched one video (in which the AI was in fact doing something a lot better than in other sims) and for some reason decided that's all there is to see.

It's also worth pointing out, that the intelligence of a human opponent is not going to be matched in flight sim AI in the foreseeable future, and that is an attractive challenge for multiplayer. On the other hand, a guy who's been playing flight sims in his gym shorts for 20 years with the benefit of 80 years of historical hindsight, a comfy chair, and a bag of chips, but who has never flown a real airplane of any kind, and mainly wants to increase his stats, is not going to act the same as an actual pilot in a real cockpit who *is* 20 years, and knows a rash decision has a good chance of resulting in him dying a horrible death within the next five minutes. 

Getting modern players to discipline themselves en mass to fly from a more historical perspective, and operate in some semblance of following procedures used at the time in order to recreate the experience of an actual mission is going to be vastly more difficult than getting the AI to do those things. Kudos to a few groups who have made impressive efforts, but even they still rely on AI aircraft to fill in the gaps, and most of them still lack the training, discipline, and incentives of a real pilot. The shift towards multiplayer has made those of us who preferred the big picture mission experience be dependent on whether a bunch of other people decide to fly together in a way they have neither the training, discipline, nor motivation to do. I'm not saying that's even the point of multiplayer or that it should be, but we have lost something we used to have. That's why some of us prefer single-player, and are frustrated that 20 years later, sims are not doing any better at this, and are arguably a fair bit worse, because it hasn't been something developers have cared about in any meaningful way.

A lot of what's been shared indicates the developers now are trying to correct some of that imbalance with Korea. We're now at the point where we find out if they made meaningful progress and if they have the commitment to iron out the issues that will inevitably pop up and make it work really well. Gritting my teeth and hoping for the best.

Hit the nail on the head here. Multiplayer, whilst you obviously get the quick fix of battling another human, doesn't really come close to the actual immersive feeling of a well structured career mode, dynamic or non. I agree, the advent of multiplayer has rendered single player and with that AI to the backburner. Developers have got lazy almost in terms of not really having to please the single player community. Not just flight sim developers either, think of the amazing single player experiences of games like Medal of Honour, Battlefield Bad Company etc etc. Anyway, regarding AI, it was always going to be lacking on launch, hence the total lack of proper info from the devs, all we can hope for is significant updates in the future.   

[Steini18778]

2 hours ago, tattywelshie said:

Hit the nail on the head here. Multiplayer, whilst you obviously get the quick fix of battling another human, doesn't really come close to the actual immersive feeling of a well structured career mode, dynamic or non. I agree, the advent of multiplayer has rendered single player and with that AI to the backburner. Developers have got lazy almost in terms of not really having to please the single player community. Not just flight sim developers either, think of the amazing single player experiences of games like Medal of Honour, Battlefield Bad Company etc etc. Anyway, regarding AI, it was always going to be lacking on launch, hence the total lack of proper info from the devs, all we can hope for is significant updates in the future.   

We hope for 17 years now. As Battle of Kuban was on the horizon we all hoped that the solid campain mode will change the AI for any better. For the release date i bought books summarising the events on Kuban only for preparations. I started the campaign and after some missions my expert JG52 squadron was wiped out. It is a huge immersion killer managing a squadron having it erased very soon. I gave up hope finally ...

Edited June 23 by Steini18778

[Avimimus]

1 hour ago, Steini18778 said:

I started the campaign and after some missions my expert JG52 squadron was wiped out. It is a huge immersion killer managing a squadron having it erased very soon. I gave up hope finally ...

I still cite a conversation with Starshoi (who built the original dynamic campaign generator for the original Il-2) about this. The basic observation is:

1. Real pilots are less aggressive.

2. Players want the AI to be aggressive.

3. Therefore historical attrition rates are impossible.

Real pilots will often avoid combat - especially if they don't have an advantage: Out numbering the enemy, catching the enemy by surprise, or having more altitude/energy.

Real pilots would also try to regroup or would return to base if separated or disoriented (rather than risk being jumped).

With some ironic tension towards player expectations: This is even more true of experienced units and aces. Pilots like René Fonck or Erich Hartmann preyed on inexperienced or unaware targets caught in disadvantageous circumstances - that is how they survived to get so many kills. A true top ace AI would be one that avoids you or ambushes you (before immediately disengaging).

I'd personally love a 'historical morale' difficulty setting that would cause formations of aircraft to be less likely to engage, be more likely to disengage if separated from allies, and be more likely to disengage/RTB after a single pass on a target. A lot of air combat was like that - one pass followed by disengagement.

[Aapje]

It's the classic conundrum of gamers not caring as much about dying as they would in real life, and wanting more of the fun parts of the simulated world than in reality, and less of the dreary stuff.

In reality, only 5% of pilots made the 5 kills to become an ace, while in the came, pretty much every player is an ace many times over, and if they have the choice, will generally lower the difficulty to ensure this.

 

Quote

Real pilots will often avoid combat - especially if they don't have an advantage: Out numbering the enemy, catching the enemy by surprise, or having more altitude/energy.

Or they have a duty like escorting that doesn't allow them to chase a fleeing plane. I would probably be court-marshalled many a time if I flew the same way in a real air force.

[tattywelshie]

1 hour ago, Avimimus said:

I still cite a conversation with Starshoi (who built the original dynamic campaign generator for the original Il-2) about this. The basic observation is:

1. Real pilots are less aggressive.

2. Players want the AI to be aggressive.

3. Therefore historical attrition rates are impossible.

Real pilots will often avoid combat - especially if they don't have an advantage: Out numbering the enemy, catching the enemy by surprise, or having more altitude/energy.

Real pilots would also try to regroup or would return to base if separated or disoriented (rather than risk being jumped).

With some ironic tension towards player expectations: This is even more true of experienced units and aces. Pilots like René Fonck or Erich Hartmann preyed on inexperienced or unaware targets caught in disadvantageous circumstances - that is how they survived to get so many kills. A true top ace AI would be one that avoids you or ambushes you (before immediately disengaging).

I'd personally love a 'historical morale' difficulty setting that would cause formations of aircraft to be less likely to engage, be more likely to disengage if separated from allies, and be more likely to disengage/RTB after a single pass on a target. A lot of air combat was like that - one pass followed by disengagement.

That's a great observation, a historic morale setting would definitely be workable wouldn't it as well? It's a shame really we can't have a genuine dynamic campaign really, but as has been mentioned, developers haven't really prioritised this. I know DCS is working on one, but who knows if that will ever be released, and if so, it's probably years away. 

[Avimimus]

1 minute ago, tattywelshie said:

That's a great observation, a historic morale setting would definitely be workable wouldn't it as well? It's a shame really we can't have a genuine dynamic campaign really, but as has been mentioned, developers haven't really prioritised this. I know DCS is working on one, but who knows if that will ever be released, and if so, it's probably years away. 

Well, I'd recommend some reflection regarding what we mean by 'dynamic campaign'... originally it meant a campaign with a mission generator (rather than relying on totally pre-scripted missions).

A lot of people seem to think that a dynamic campaign requires a very complex strategy layer to the game world, with persistence between missions - but this was very rare in the history of flight sims. The only ones I can think of are the Enemy Engaged helicopter simulations (EECH/EEAH), Falcon 3.0/4.0/BMS, and some versions of the DiD sims (Tactcom, TAW), and maybe BoBII:WoV.

Some of the early 1990s sims had strategic campaigns - but they were often very limited in functionality. In contrast, having a game mode that procedurally generates missions is what the term 'dynamic campaign' meant for most of the history of flight simulators (meaning that Great Battles has one integrated, along with at least two third party ones as well).

The addition of a squadron commander logistics/strategy layer on top of a procedural mission generator, and damage that persists between missions, firmly puts Korea even more firmly into the category of sims with 'dynamic campaigns' - even though the developers wouldn't describe it that way.

[Leifr]

52 minutes ago, Avimimus said:

Well, I'd recommend some reflection regarding what we mean by 'dynamic campaign'... originally it meant a campaign with a mission generator (rather than relying on totally pre-scripted missions).

A lot of people seem to think that a dynamic campaign requires a very complex strategy layer to the game world, with persistence between missions - but this was very rare in the history of flight sims. The only ones I can think of are the Enemy Engaged helicopter simulations (EECH/EEAH), Falcon 3.0/4.0/BMS, and some versions of the DiD sims (Tactcom, TAW), and maybe BoBII:WoV.

Some of the early 1990s sims had strategic campaigns - but they were often very limited in functionality. In contrast, having a game mode that procedurally generates missions is what the term 'dynamic campaign' meant for most of the history of flight simulators (meaning that Great Battles has one integrated, along with at least two third party ones as well).

The addition of a squadron commander logistics/strategy layer on top of a procedural mission generator, and damage that persists between missions, firmly puts Korea even more firmly into the category of sims with 'dynamic campaigns' - even though the developers wouldn't describe it that way.

Combat Flight Simulator 3 also had a dynamic campaign of sorts; the player generated the pilot, chose their squadron and then could choose a sector along the frontline every day to attack, different mission profiles generated more or less influence on that sector depending on difficulty. The player could even invest points in to the pilot to improve certain attributes, I forget what they were now, maybe tolerance to g-forces and health points... At the time it was a bit of a novelty but I distinctly remember feeling very 'meh' about it and returning to the structured campaign of CFS2. 

[Steini18778]

3 hours ago, Avimimus said:

I still cite a conversation with Starshoi (who built the original dynamic campaign generator for the original Il-2) about this. The basic observation is:

1. Real pilots are less aggressive.

2. Players want the AI to be aggressive.

3. Therefore historical attrition rates are impossible.

Real pilots will often avoid combat - especially if they don't have an advantage: Out numbering the enemy, catching the enemy by surprise, or having more altitude/energy.

Real pilots would also try to regroup or would return to base if separated or disoriented (rather than risk being jumped).

With some ironic tension towards player expectations: This is even more true of experienced units and aces. Pilots like René Fonck or Erich Hartmann preyed on inexperienced or unaware targets caught in disadvantageous circumstances - that is how they survived to get so many kills. A true top ace AI would be one that avoids you or ambushes you (before immediately disengaging).

I'd personally love a 'historical morale' difficulty setting that would cause formations of aircraft to be less likely to engage, be more likely to disengage if separated from allies, and be more likely to disengage/RTB after a single pass on a target. A lot of air combat was like that - one pass followed by disengagement.

I agres witho you here, but having read many biographies and combat reports, i have to say that at least the core of german aces actively searched for combat.

[belphegor]

2 hours ago, Avimimus said:

Well, I'd recommend some reflection regarding what we mean by 'dynamic campaign'... originally it meant a campaign with a mission generator (rather than relying on totally pre-scripted missions).

A lot of people seem to think that a dynamic campaign requires a very complex strategy layer to the game world, with persistence between missions - but this was very rare in the history of flight sims. The only ones I can think of are the Enemy Engaged helicopter simulations (EECH/EEAH), Falcon 3.0/4.0/BMS, and some versions of the DiD sims (Tactcom, TAW), and maybe BoBII:WoV.

Some of the early 1990s sims had strategic campaigns - but they were often very limited in functionality. In contrast, having a game mode that procedurally generates missions is what the term 'dynamic campaign' meant for most of the history of flight simulators (meaning that Great Battles has one integrated, along with at least two third party ones as well).

The addition of a squadron commander logistics/strategy layer on top of a procedural mission generator, and damage that persists between missions, firmly puts Korea even more firmly into the category of sims with 'dynamic campaigns' - even though the developers wouldn't describe it that way.

To me, a "dynamic campaign" means one in which not only are the missions generated, rather than scripted, but also where your actions have an impact on the outcome of the war.  The first example of this that I personally know of is Lucasfilm's Secret Weapons of the Luftwaffe (1991), where you could designate targets or priorities for the strategic bombing campaign, and then fly the missions that resulted.  If you did well in the mission, then your bombing mission also went well, and for instance if you bombed the BMW engine plants, there would be fewer FW-190s coming up to intercept you, or likewise for the Daimler-Benz plants and Bf-109s, or you could hit the oil refineries, ball bearing plants, etc., with varying effects.  And if you did poorly, you would have fewer bombers and fighters for the next missions.  I'm sure there couldn't have been very much sophistication in the logic that drove the campaign, considering that the entire game was probably smaller than just a single background image in an Il-2 menu, but it did feel like you were really fighting the war through your actions nevertheless.

Another early example was 1942: Pacific Air War (1994), where you would direct the carrier battle groups around the map, send out scouting patrols, and when one of the patrol planes disappeared, do an alpha strike and fly the mission.  If you did well and sunk some ships, those ships would be gone from the campaign.

I have always considered a system where the missions are generated, but the ultimate outcome is fixed, like Il-2 Great Battles, to be a "semi-dynamic" campaign.  No matter how well or poorly you do in the missions, AFAIK, the front line always advances or retreats in accordance with the historical outcome.  Your actions affect your squadron, maybe how many planes are available, or the stories in the newspapers, but you can't affect the overall outcome of the war.

There were also semi-dynamic campaign systems like in Gunship 2000 (1992), where IIRC the missions were dynamically generated, and you could rise through the ranks, become a flight lead, get access to better helicopters, etc., but there was no overarching war that you were fighting, no front line that moved, no supply system, etc.

[Aapje]

In reality, no pilot determined the outcome of the war, so for me it would just be a step away from realism in a way that I don't value all that much. In GB, the bigger issue I have with the career mode is that the missions are often very similar, which is probably true to reality, but are not that much fun. Another issue is that the missions often didn't feel organic, but designed around the player. PWCG improved both of those things for me.

But this is getting a bit offtopic, since it is not actually about the AI.

[Avimimus]

1 hour ago, Steini18778 said:

I agres witho you here, but having read many biographies and combat reports, i have to say that at least the core of german aces actively searched for combat.

Of course! You don't become an ace by avoiding combat. But I'd ask of these biographies:

How many engagements do they actually describe? Looking at Hartmann, he's estimated to have flown over 1400 combat sorties, encountered the enemy in only 58% of them and shot down an aircraft about once every four sorties (and that is assuming there is no overclaiming). In areas with relatively high likelihood of enemy contact, it is quite possible that he would have encountered several enemy formations on those flights - but normally only engaged one of them...

So my argument isn't that Aces always avoided combat - it is that the most successful ones tended to avoid combat in which they'd be at a disadvantage - and they also tended to try to get out of a position of vulnerability as soon as possible (e.g. climbing away after ambushing a target).

[Drewm3i-VR]

8 hours ago, Avimimus said:

I still cite a conversation with Starshoi (who built the original dynamic campaign generator for the original Il-2) about this. The basic observation is:

1. Real pilots are less aggressive.

2. Players want the AI to be aggressive.

3. Therefore historical attrition rates are impossible.

Real pilots will often avoid combat - especially if they don't have an advantage: Out numbering the enemy, catching the enemy by surprise, or having more altitude/energy.

Real pilots would also try to regroup or would return to base if separated or disoriented (rather than risk being jumped).

With some ironic tension towards player expectations: This is even more true of experienced units and aces. Pilots like René Fonck or Erich Hartmann preyed on inexperienced or unaware targets caught in disadvantageous circumstances - that is how they survived to get so many kills. A true top ace AI would be one that avoids you or ambushes you (before immediately disengaging).

I'd personally love a 'historical morale' difficulty setting that would cause formations of aircraft to be less likely to engage, be more likely to disengage if separated from allies, and be more likely to disengage/RTB after a single pass on a target. A lot of air combat was like that - one pass followed by disengagement.

The funny part is this^ is what the best online pilots also do! 🙃

4 hours ago, belphegor said:

To me, a "dynamic campaign" means one in which not only are the missions generated, rather than scripted, but also where your actions have an impact on the outcome of the war.  The first example of this that I personally know of is Lucasfilm's Secret Weapons of the Luftwaffe (1991), where you could designate targets or priorities for the strategic bombing campaign, and then fly the missions that resulted.  If you did well in the mission, then your bombing mission also went well, and for instance if you bombed the BMW engine plants, there would be fewer FW-190s coming up to intercept you, or likewise for the Daimler-Benz plants and Bf-109s, or you could hit the oil refineries, ball bearing plants, etc., with varying effects.  And if you did poorly, you would have fewer bombers and fighters for the next missions.  I'm sure there couldn't have been very much sophistication in the logic that drove the campaign, considering that the entire game was probably smaller than just a single background image in an Il-2 menu, but it did feel like you were really fighting the war through your actions nevertheless.

Another early example was 1942: Pacific Air War (1994), where you would direct the carrier battle groups around the map, send out scouting patrols, and when one of the patrol planes disappeared, do an alpha strike and fly the mission.  If you did well and sunk some ships, those ships would be gone from the campaign.

I have always considered a system where the missions are generated, but the ultimate outcome is fixed, like Il-2 Great Battles, to be a "semi-dynamic" campaign.  No matter how well or poorly you do in the missions, AFAIK, the front line always advances or retreats in accordance with the historical outcome.  Your actions affect your squadron, maybe how many planes are available, or the stories in the newspapers, but you can't affect the overall outcome of the war.

There were also semi-dynamic campaign systems like in Gunship 2000 (1992), where IIRC the missions were dynamically generated, and you could rise through the ranks, become a flight lead, get access to better helicopters, etc., but there was no overarching war that you were fighting, no front line that moved, no supply system, etc.

Sounds like what you really want is Tactical Air War and/or Combat Box's Apollo...

[KC23]

On 6/21/2026 at 1:14 PM, Avimimus said:

One thing that is worth keeping in mind is that combat in the early jet era looked quite different from combat in WWII.

I do appreciated that point of view. It certainly is valid. I just find it very hard to believe that a pilot would ignor tracers flying by his cockpit and continue with the same turn rate, speed and aspect geometry. Unless of course it was a clear passing deflection attempt where it was obvious the shots fired could not continue to pose an immediate threat.

Edited June 24 by KC23

[AndreiTomescu]

Regarding the AI that we already have, and some despise so much: from my own experience, with about 3k hrs in GB single player, and myself being an average (at best!) virtual pilot:

- in career, the AI is poor, indeed. Also the hole experience is quite centered on the player

- PWCG improved some, but not fantastic

- some scripted campaigns (not all, but quite some) brought me an completely different AI, much more believable, and even very challenging more than often (remember, i'm a average if not less than average "pilot", aiming surviving and historical accuracy over kills)

Why is that? If the AI is the same, why is it hugey better in some scripted campaigns? What "magic" did their creators do?

And if such "magic" is possible, why wouldn't it be reproduced and enhanced in Korea?

I guess some patience will bring positive changes.

Also, guys, if a truly intelligent AI would be just over the hill, all those damned real drones wouldn't need pilots. I'm no programmer, but i really think getting an intelligent AI opponent is something hard to obtain. Not impossible, but difficult.

Also, the importance of smarter AI might become more relevant for more many players: the new and revolutionary (imho) environment created by LuftManu and his accolites in the new PvE project asks for better AI even in the online realm, not just the off line one.

To end, from my pov the present AI is NOT bad, it's mostly inconsistently good, and depending heavily on the skills of the mission designer. Maybe this will change. But won't be tomorrow. Extending the game to a larger player base (controller-abomination!-, easyer controls, HUDs, graphics, etc) might bring the extra cash wanted and needed for some investment into this elusive AI.

[KC23]

48 minutes ago, AndreiTomescu said:

- some scripted campaigns (not all, but quite some) brought me an completely different AI, much more believable, and even very challenging more than often (remember, i'm a average if not less than average "pilot", aiming surviving and historical accuracy over kills)

Why is that? If the AI is the same, why is it hugey better in some scripted campaigns? What "magic" did their creators do?

And if such "magic" is possible, why wouldn't it be reproduced and enhanced in Korea?

Those are some great questions. Back in the day a very old scripted missions combat flight sim called, Chuck Yeager Air Combat had some pretty darn challenging AI to fight against. Mig Alley by Rowans was another. Of course don't talk to me about the fidelity of their flight models. No doubt IL-2 Korea will blow almost any sim out of the water (not to mention the graphics and damage model).