Regarding the AI issues - a request for help

[453=Whittle]

19 hours ago, Majpalmer said:

One thing that we need to keep in mind is that, I suspect, almost everyone here is a veteran with flight sims. I'm talking decades of experience. What would be interesting would be to get a complete novice, given him 10 hours of training on Il-2 FC, and send him up against an AI pilot, Veteran or Ace. Would he find the AI challenging? What's your guess? When new guys, even with extensive experience, show up on the servers, how long do they usually last? 

 

I'm actually working on a video arguing that the AI is too good. I found one reference that of all the US fighter pilots, all services, during WWII, 90% never even claimed a kill. For the Luftwaffe single-seat fighter force during WWII, the percentage with zero claims was 80%. About that same percentage ended up dead! So, if you go up with a flight of five, and the other four guys do shit, that's actually realistic. 

 

It's not challenging, to be sure. But is that because the AI is bad or because the typical Il-2, or WOFF player, is too good? 

 

Just a thought. 

 

The only reasonable solution that respects customers is to have difficulty settings that allow enjoyment for both newbies and experts.

[Majpalmer]

You can do that, to a degree, using Pat Wilson's Campaign. You can tune up, or down, he AI skill levels. And it can be done in Quick Missions. 

 

Thirty years ago, I flew in an online squadron in Air Warrior. I was the Deputy and had to keep the stats. Our best two guys had kill-to-death (KD) ratios of around 15/1. Me and a couple of others were around 4/1. Then we had the guys who were 1.3/1. You had to be at least half-way decent to get invited. But the skill levels between our best and our worst were enormous. Once I remember one of our better guys went up in a Stuka, using an anonymous second account, and shot down one of our poorer guys in a 190. 

 

And then you had the guys who wanted to join but were rejected. There were poor souls who had 1/12 KDs flying with a Mouse or a Joystick-Throttle-Twist rudder controllers. 

 

It's no easy task, I assume, to program AI to challenge the 15/1 k/d guy, but not overwhelm the 1/12 guy. 

 

One thing I would like to see change is the tendency of the AI to fight to the death. The other day I led a flight of 6t Pfalz DIIIs against 4 SE5as. I shot down 2 SEs, which left it a 2v6. With their far greater speed, the SEs should have run. There's no way we could catch them. But they fought it out and all died. I consider the inability of the AI to disengage to be one of its weakest points. 

[Dr1falcon500]

I've tested a few QM's, my plane on auto pilot, flying against ace AI and they don't fly all that great. One flight Halberstadts v N17's went on for 20 minutes until an N17 collided with my DII. N17 on auto pilot only fired when enemy was directly in front, no effort to turn after enemy planes. Very erratic flying and poor shooting. Enemy planes would turn in from behind and often just veer away. At the end my N17 dived for no reason, climbed and stalled. At the end it did this twice and crashed, no pilot injury or aircraft damage, just bad flying. These are supposed to be Ace AI? In PWCG sometimes you run into an ace Mick Mannock or one of those guys, but they fly and shoot no better than the rest. The challenge in PWCG is that you might be fighting 5:1 sometimes, the AI is more likely to get a bit more aggressive.

[Aapje]

I would like for the AI to have more realistic behavior, like having similar limitations as humans for spotting. So they can't look through the plane, can't look everywhere at once, can't always afford to look around/behind them or may choose to favor the kill, etc.

 

I would like for them to weigh different goals, like the mission, self-preservation, ego (getting a kill over defending the bombers or letting another ally take the kill), etc.

 

I would like for them to be able to fly planes according to their strengths and weaknesses, like energy fighting vs turn fighting, also based on what the opponents are flying.

 

I would like it if the AI would model the human behavior and make predictions based on that.

 

And I would like for AI to have strengths and weaknesses in more dimensions than just good/bad. A pilot may be great at spotting, but poor at shooting. He may be great at maneuvering, but suffer from tunnel vision to an extreme extent. He may be a coward who declines any fight that looks the slightest bit risky, or too confident and enter a fight at a big disadvantage. He may know the strengths of his planes well or not. He may know the strengths of other planes well or not. Etc.

 

You could have AI pilots that differ on all of these dimensions.

 

To really do this well almost certainly requires a machine learning AI.

 

There is a risk that doing partial fixes just exposes other issues. For example, the current obsession by planes with their mission mitigates the omnivision to some extent, as you can 'sneak' up to planes (in reality, the AI does see you, but doesn't care). The good old sucker punch was an important aspect of real flying.

 

But if you fix the obsession with the mission without making the vision more realistic, the game will become less realistic in that aspect.

Edited February 9, 2024 by Aapje

[LukeFF]

Guys, let's please stay on the discussion of the AI. ?

[Dupxo]

After the latest patch there is still much work to be done with AI behaviour. OK, the AI enemy planes take head-on now like they should but the most irritating thing is that AI scouts don't stay straight on enemy's tail for more than 2 seconds. This is absurd that they are always doing dodges when they are on enemy tail like they want to dodge tail gunner's fire even if the enemy is one seater scout :/.

 

Bots should be more "focused" when they are on enemy tail, they shouldn't do this dodging maneuvers crap at all OR do this only when tail gunner is firing at them. Bots should SHOOT almost all the time when they have the ability to hit the enemy and should stay focused on enemy six when they outmaneuver him. Now they break maneuvre almost all the time when they reach the good position to shoot enemy plane at the back OR they fire VERY short burst and then do dodging crap maneuvers.

I am not a programmer but maybe look at RoF AI code. I know that RoF AI is average but at least bots in RoF stay more focused and they shoot longer and more accurate bursts when they can hit the enemy. Letting the bot to catch your tail in RoF is mostly the deathly situation. In Flying Circus it is pure... circus.

[Majpalmer]

Today I had a very different fight. I was one of 4 Alb DVs near Lens and we were jumped by a lone Camel with Alt. For ten minutes we dueled with this Brit Major who repeatedly dove on one of us, but then zoomed back up and maintained his Alt advantage in a fashion I'd not seen before from the AI. He'd dive in, HI YOYO out and up, fly away, reverse, and repeat. Eventually I started diving to get some E when he reversed, and then would take nose-up shots at him. After ten minutes of this I got a lucky hit. 

 

The AI with a 1v4 surviving for 10 minutes? 

 

Of course, better AI would have had him disengage at some point, but....

 

Was something changed the other day? 

[Majpalmer]

All the aircraft are holding their altitude better. 

 

And I have to relearn co-alt head-to-head merges. The AI now comes in firing. 

[Dr1falcon500]

The last update change 15, Fixed a problem with AI pilots pulling away from frontal attacks too early. The result of this seems to be near suicidal AI flying. These attacks are direct head on, not angled. Would real pilots do this? I've made a fair few head on attacks from angles where enemy planes couldn't return fire, not direct front assaults. It isn't too difficult to counter at WW1 speeds. After these high risk attacks AI goes back normal evasion. 

[Majpalmer]

Decades ago I flew online in Air Warrior and the ... unskilled often went head-to-head at you. I learned how to avoid it. Unfortunately, I've been spoiled since. Now I have to relearn. 

BTW, do you see them holding their ALT advantage better? Or is it just me? 

[Dr1falcon500]

Some of them seem to be a bit better at climbing away fairly quickly. It seems that an enemy plane is level with you then a few second later back to having an altitude advantage. It depends a lot on what plane you're flying, mostly I'm flying slower planes like Albatros DVa and Dr1. Last mission was Camel  flying mostly against DVII so it wasn't that noticeable. 

[Majpalmer]

Okay. Here's an AI issue.

 

Flying Circus. Flight of four Pfalz D3as. We encounter 2 Alb DVas fighting 2 SPAD VIIs and 3 Camels at a higher altitude. We close to help. 

 

Obvious move for the Albs as we approach is to drag the enemy down to our altitude. That would make a 2v5r a 6v5. 

 

But, no! The 2 Albs keep fighting high as we climb until 1 gets shot down. Then the survivor breaks off, leaving the SPADs and Camels to dive on us. 

 

I've attached the mission. I have no idea if it will play out the same way. 

 

Nor do I have any idea if that kind of AI thinking can be programed. 

Jakob Rosenmeier 1917-06-08.zip

[Dr1falcon500]

On 2/29/2024 at 7:53 AM, Majpalmer said:

Today I had a very different fight. I was one of 4 Alb DVs near Lens and we were jumped by a lone Camel with Alt. For ten minutes we dueled with this Brit Major who repeatedly dove on one of us, but then zoomed back up and maintained his Alt advantage in a fashion I'd not seen before from the AI. He'd dive in, HI YOYO out and up, fly away, reverse, and repeat. Eventually I started diving to get some E when he reversed, and then would take nose-up shots at him. After ten minutes of this I got a lucky hit. 

 

The AI with a 1v4 surviving for 10 minutes? 

 

Of course, better AI would have had him disengage at some point, but....

 

Was something changed the other day? 

Very similar to an encounter between our Jasta 11 patrol and 4 Camels. 1st 2 Camels went down after colliding. 2nd flamed. The last Camel was flying above and diving to try a few head on attacks. Managed to avoid these and inflict some engine damage. Camel still kept above my Dr1, but downed after PK burst fired from near stalling position below. Seems like something new and different.

[Dr1falcon500]

I hope something will be done about the general reckless and erratic AI flying. Usually there is a collision in each mission. Two of my planes almost smashed into each other while taking off. It's common to have one of my own planes turn across my path or dive in front of me. If my plane was AI it would have been splattered. There was supposed to improvements and work done on AI if customers bought FC Volume 3. I don't see much has improved. Could we please not have this level of AI idiocy as in this screenie N17 dived right in front of me, nice.

[ST_Catchov]

3 hours ago, Dr1falcon500 said:

in this screenie N17 dived right in front of me, nice.

 

This has been happening FOR EVAH! I often wonder if it has something to do wth WWII co-habitation with WW1 and residual unwanted effects filtering into FC? Otherwise, Luke tells us the devs have been struggling to find a suitable AI expert over the years .... or some such thing.

 

Some like to say it's realistic behavior from novice target obsessed pilots.

 

Who knows, except to say the AI issues are a major fail. Go MP they say, which entirely ignores the problem.

[LukeFF]

Guys, you know what I (and Han) am going to say: post track files showing what the problem is. That's the only way this is going to be addressed. 

[Dr1falcon500]

10 hours ago, LukeFF said:

Guys, you know what I (and Han) am going to say: post track files showing what the problem is. That's the only way this is going to be addressed. 

So the usual "we need track files". This sort of implies that people are making all this up for some reason. Most people  who regularly fly any sort of missions know  the state of AI.

The original post in this thread was 3 or 4 YouTube videos all zoomed out to maximum and played at extremely high speed. This was supposed to convince us that AI is okay, but what could anyone tell from those. That was okay for you to post videos, but not us?

[Aapje]

4 hours ago, Dr1falcon500 said:

So the usual "we need track files". This sort of implies that people are making all this up for some reason.

 

This is not very charitable. They probably need the track files to be able to debug the issues efficiently

[LukeFF]

13 hours ago, Dr1falcon500 said:

So the usual "we need track files". This sort of implies that people are making all this up for some reason. Most people  who regularly fly any sort of missions know  the state of AI.

The original post in this thread was 3 or 4 YouTube videos all zoomed out to maximum and played at extremely high speed. This was supposed to convince us that AI is okay, but what could anyone tell from those. That was okay for you to post videos, but not us?

 

You also left out the part in my OP where I said "At this point, we need more information from you in the form of mission and track files that show what planes are having AI issues and how those issues can be repeated. That sort of information will help the team greatly to see where there may be issues that need to be rectified."

 

9 hours ago, Aapje said:

This is not very charitable. They probably need the track files to be able to debug the issues efficiently

 

What he said. It has always, always, always been the stance of the team that if you want to not just report problems but actually see them be fixed, then posting track files is the only way to go about this. That allows them to see what is going on in ways that a YouTube video is never going to show. 

[Majpalmer]

I'm not a computer expert; I'm an historian. I see the AI issue as a complicated one for the designer of Il-2 because the AI challenge involves what could be termed the micro and the macro, and the two are not entirely symmetrical. In many ways, they are at odds.

 

As for the micro challenge, think of the Quick Mission. You’re in your fighter and you see a Bandit at 3 o’clock level. You turn to engage. What you’re looking for from the AI is simple—a challenge. You want that AI Bandit to have the ability to shoot you down if you make a mistake. You want to have to work for the kill.

 

But in Il-2, along with the QMs, you have your campaigns and your careers. These pose very different challenges. Yes, you still want that AI Bandit to be tough, but you also need the overall outcomes of all the aircraft engaged to approximate a realistic result. Not every Bandit was an Ace! If you’re Hartman or Voss, very few of the Bandits in the air should have the ability to shoot you down in a 1v1.  And the problem with Il-2 is that too many of us, with years or decades of experience, are virtual Hartmanns and Vosses.

 

During WWII, 80% of Luftwaffe single-seat fighter pilots never claimed a kill! The statistics for the American air services were over 90%! The harsh reality was that the overwhelming majority of fighter pilots during the World Wars sucked!

 

The problem for the designer is how do you model that? If you fill the skies with sucky pilots, it’s less than challenging. But if you make the Bandits challenging to the guy who has been playing air combat sims for decades and is a virtuoso of a pilot, while that works fine at the micro, QM, level, that same AI will make a mockery of the outcomes at the macro, i.e., campaign and especially career levels.

 

As things stand, at the macro level, I would argue that the AI is too good! For example, during the Great War Jasta 12, operational from mid-October 1916 until November 1918, about 25 months, lost 26 pilots—about 1 per month. During that same period, the pilots of the Jasta claimed 155 kills—about 6 per month. There is no way you could play a Jasta 12 career in FC and only lose 1 pilot per month. Damn! I lose almost a pilot a day when I play! Nor would your squadron (not counting your own kills) only score 6 kills per month in a career that sends six guys up about 40+ times per month.

 

Two years ago, I ran 2 experiments playing German careers in Battle of Moscow. During that period of the war, 80% of Luftwaffe pilots were not scoring during the average month. In both my campaigns, about 80% of my pilots scored! They were too good, by historical standards.

 

We’ve all complained about the career missions when we go up and the other 5 guys do zip. But that was the reality. Even among Luftwaffe single-seat fighter pilots, arguably the most skilled of the lot, 4/5^th never scored!

 

To give you an even starker example of how harsh the reality was, during WWII 5081 German single-seat fighter pilots (out of about 20000) claimed at least 1 kill. Total claims were over 54000. But of those 54000+ claims, 350 pilots claimed 27000+; the other 4731 claimed the other 27000+. In other words, of the Luftwaffe’s 20000 single seat fighter pilots, 350, or 1.75% of the total, did 50% of all the scoring! Another 18.25% accounted for the other half of the scoring, and the remaining 50% did zip!

 

How do you model an AI that is challenging at the micro level, but at the macro yields historical results? 

[Aapje]

@Majpalmer

 

A mistake in your narrative is that you assume that the campaign is true to reality, aside from the skill of the pilots. It is not. The campaign is gamified, in that you are far more likely to encounter enemy airplanes than in reality. In real combat flying, many sorties resulted in no encounter with the enemy.

 

Quote

Two years ago, I ran 2 experiments playing German careers in Battle of Moscow. During that period of the war, 80% of Luftwaffe pilots were not scoring during the average month.

 

A lot of them flew transports, did CAP, did escort missions with a small chance to get attacked, etc. If you fly fighters and then also benefit from the game giving you way more enemy fighters to combat than is realistic, you will of course score more and die more.

 

Quote

As things stand, at the macro level, I would argue that the AI is too good!

 

Your statistics, showing that they shoot down a lot, but also get shot down a lot, point more towards them being too aggressive and/or getting in too many fights in the first place.

 

I would personally not call pilots that get relatively many kills, but die very quickly: good. Certainly not if I was such a pilot in real life.

 

Quote

We’ve all complained about the career missions when we go up and the other 5 guys do zip. But that was the reality. Even among Luftwaffe single-seat fighter pilots, arguably the most skilled of the lot, 4/5^th never scored!

 

Reality is not that fun. It's a trap to try to make a game reflect reality too much. You get praise from people who admire it from afar, but no one enjoys playing it.

 

Quote

How do you model an AI that is challenging at the micro level, but at the macro yields historical results? 

 

You don't make it reflect historical results, because that contributes almost nothing to how much fun the game is to play.

Edited April 23, 2024 by Aapje

[EAF19_Marsh]

22 hours ago, Majpalmer said:

During WWII, 80% of Luftwaffe single-seat fighter pilots never claimed a kill! The statistics for the American air services were over 90%! The harsh reality was that the overwhelming majority of fighter pilots during the World Wars sucked!

That regularly engaged hostile aircraft or as % force total? Sounds a little odd. 

[Majpalmer]

I understand about the actual engagement rate. That's implicit in the stats for Jasta 12--155 kills over 25 months, or about 6 Jasta victories for all pilots a month. As for the rest, I don't disagree with most of what you say. And yes, reality is often boring. 

 

My main point was that it's hard to make the AI tough for the single mission and even some approximation of reality for a career. And I suspect a lot of people play SP in career mode. 

 

The bottom line is that even with the current unchallenging AI, in career mode you engage far too often. So, yes, as you wrote you score more, and your guys die more. 

 

For example, in a current FC career, Pat Wilson, I'm averaging 2 kills per sortie. In two months of decent weather, I can rack up enough kills to top MvR. The only reason I haven't caught up with him is that I give away kills and force myself to take leaves for multiples of victories, medals, and promotions. That's a way, without playing around with the AI flight models, to simulate reality. Vanilla career mode has bad weather days but zero leaves. 

 

My aim isn't to have the career player flying around empty skies for reality's sake. I'd like to see a career game where when you did fly you'd have a hell of an engagement, but the program would simulate--bypass--the non-engagement sorties. 

 

I do believe that we need to keep in mind, when we talk about the AI, just how bad most pilots were. As I said, 80% of Luftwaffe single-seat fighter pilots never scored! I actually think the US percentage waws 93! 

 

 

On 4/24/2024 at 11:24 AM, EAF19_Marsh said:

That regularly engaged hostile aircraft or as % force total? Sounds a little odd. 

During the war there were about 20k Luftwaffe single-seat fighter pilots. of all of those pilots, 5081 claimed at least one kill. 

[EAF19_Marsh]

19 hours ago, Majpalmer said:

During the war there were about 20k Luftwaffe single-seat fighter pilots. of all of those pilots, 5081 claimed at least one kill. 

 

That's what I mean. How many in combat, regularly in combat, in high-intensity scenarios, loss rates etc. It is a something of a vague statistic.

[Rick_Rawlings]

Here's my test: Fully fueled Pfalz vs. 3 Ace AI Camels. I don't fly FC all that much and have only flown the Pfalz a few times and not in the last two years. I don't even have the radiator adjustment mapped, apparently, so I am hardly in peak form.

 

Test Zero:

Track file link:

https://drive.google.com/file/d/1V8R_l37yFrdIZzIuHXalatiJ2fbgtYKW/view?usp=sharing

 

Test One:

Track file link:

https://drive.google.com/file/d/11Rj-4pfl3gr0vBxZxjL2gg4oQ7J5yHOC/view?usp=sharing

 

So what I am seeing now would normally be something more like what I would expect to be normal-to-low-veteran AI. They are not making huge mistakes, but they are not eating my lunch like ace AI should. I mean, I am not flying the Pfalz to its strengths or even flying all that well in general. The first overriding issue before addressing anything else seems to me to be the AI tendency to forget why they are there in the first place. I had a few runs that ended in collisions with me (some of which were my fault!) early on so I didn't post those, but they also featured what you can see in these videos: the AI have no problem wandering off on their own, sometimes for minutes at a time!  Then there is like this "Ohhhhhhh yeeeeaaaaahhhhhhhh, I remember! There was a dogfight going on! Wonder how that is going?" moment and they wander back. 

 

What I would like to see from AI, speaking only for myself, are planes that:

 

1. Having increasing ability to detect enemies at a distance based on skill level.

2. Having increasing ability to decide if, when and how to engage based on skill level and numerical, altitude and mission considerations.

3. Be able to fight intensely and perform maneuvers based on skill level, with novices making mistakes and even stalling out and not flying with the best practices for that type of plane up to aces making very few mistakes.

4. Deciding when to head for home based on skill level, health, plane damage, position relative to the lines and numerical and altitude conditions.

 

Someone made a good comment above that being a game that we elect to pay money for and play, we want opportunities for combat on most missions which was not  typical for the real WWI pilot. I get around this in WOFF by advancing time several days or a week between missions, to simulate the missions that my pilot flies where nothing conclusive happens and therefore I don't need to be present. You could add that into the career mode if it isn't already. But also having fights where people break off and go home instead of fighting to the death would go along way to helping out with this problem.

 

Anyway, thanks for looking at this issue. I would certainly play more if it were addressed.

 

RR

[Majpalmer]

Rcik_Rawlings wrote: "I get around this in WOFF by advancing time several days or a week between missions, to simulate the missions that my pilot flies where nothing conclusive happens and therefore I don't need to be present. You could add that into the career mode if it isn't already. But also having fights where people break off and go home instead of fighting to the death would go along way to helping out with this problem."

 

You can do that with Pat Wilson, but not with the Vanilla FC career. I used to like the old Pat Wilson career where you could adjust the spacing between sorties in the confit file. Unfortunately, you can do that anymore, although you can manually use the "leave: function to achieve the same end. 

 

I'd argue that the FC AI is far worse than the RoF AI. In my RoF Pat Wilson careers I go with my kills as scored. In FC Pat Wilson I round down all multiple kills. In other words, if I score 3, I claim 1. This helps build up the skills of your comrades. 

 

I ran an experiment in FC using Pat Wilson's 15.3, starting in June 1917 and pacing myself with MvR. I tied his score by December 1917--7 months. At that point he'd been flying for about 15 months. Then I went into my journal and added up all my multiple kill sorties. If I rounded them all down and adjusted for the fact that I'd flown half as many months as MvR, then the scoring rate would have been about the same. So that's what I do. 

 

[Dr1falcon500]

This is a Jasta 11 PWCG mission in November 1917. Our flight of 2 Dr1's and 2 DVa's were supposed to rendezvous with bombers. 5 Camels showed up so as flight CO I thought I'd better see them off. This is fairly typical, Camels had a good altitude advantage but gave it away by diving and climbing as evasive tactics. I expected them to attack, they didn't. I got 4 Camels and somebody else got 1. Claimed 2. AI very bold if they have a good numerical advantage otherwise just evasion.

https://www.mediafire.com/file/ssjesn3w5gagok9/Tracks.zip/file

[Dr1falcon500]

Escadrille 103 career mission. Our flight only had 3 planes, setting was supposed to parity but there were a lot of Halberstadt fighter. Two of these collided. Later a Nieuport collided almost head on with my plane, so I had to restart. Too many collisions.

https://www.mediafire.com/file/hfs3lu4mx3w0q12/Tracks+2.zip/file

[Martin077]

any news on this?

[LukeFF]

Nothing new to report, sorry. 

[Dr1falcon500]

On 12/22/2023 at 7:45 AM, Han said:

It was before it was hacked by producer's decision. Reason was clear - there were many complaints about "its too hard to damage airframes by my couple .303!” Complaints were loud, so producer's decision was that such complaints are bad for product and team got to multiply airframe damage.

 

In FC we back to realism as the main criteria of simulation quality. Even if some customers want more damage.

So it was a question of the almighty dollar winning out? Where these "many complaints" mostly made by multi players by any chance? Most planes in WW1 were shot down with pilot killed or wounded, engine damage or fire, not airframe damage. I didn't have much trouble shooting down from 40-100m. After the update it was easy to fire a short burst from 200-250m and see the wings fall off. Should knocking down an enemy plane be that easy? Spoilt RoF until I added mods to correct. Thanks RoF producers.

[Avimimus]

58 minutes ago, Dr1falcon500 said:

So it was a question of the almighty dollar winning out? Where these "many complaints" mostly made by multi players by any chance? Most planes in WW1 were shot down with pilot killed or wounded, engine damage or fire, not airframe damage. I didn't have much trouble shooting down from 40-100m. After the update it was easy to fire a short burst from 200-250m and see the wings fall off. Should knocking down an enemy plane be that easy? Spoilt RoF until I added mods to correct. Thanks RoF producers.

 

If you want to discuss the damage model in Rise of Flight you can do that in the Rise of Flight subforum.

 

This thread is for discussing the artificial intelligence in Flying Circus. You are off-topic.

[Dr1falcon500]

7 hours ago, Avimimus said:

 

If you want to discuss the damage model in Rise of Flight you can do that in the Rise of Flight subforum.

 

This thread is for discussing the artificial intelligence in Flying Circus. You are off-topic.

Then why was it bought up Han on this thread?

[LuftManu]

15 minutes ago, Dr1falcon500 said:

Then why was it bought up Han on this thread?

Hey! I think it would be easier to keep track about it there and to further Fixes
If you don't mind to write it there it would be helpful.

Kind regards,

[PatrickAWlson]

@LukeFF SPADs seem to turn too much and use their speed too little. Pretty much any dogfight with a SPAD will show that.  This makes them relatively easy kills as almost anything can turn inside of a SPAD in a horizontal fight.  I have found that SE5as are an example of a plane where the AI  uses its speed more effectively.

[Dupxo]

Are AI pilots skills going to be improved in incoming update?

Edited November 16, 2024 by Dupxo

[Dr1falcon500]

1 hour ago, Dupxo said:

Are AI pilots skills going to be improved in incoming update?

AI was supposed get worked on if customers purchased Volume 3. At least that's what was said. I really can't see any difference or improvement in AI since Vol 3 release.

On 4/18/2024 at 5:26 PM, Aapje said:

 

This is not very charitable. They probably need the track files to be able to debug the issues efficiently

I'd be more charitable if IC was a charity, they're not.

[Dr1falcon500]

Fokker EIII mission.zip

This was a new career mission Jasta 6 Jametz, 27/9/16. Our EIII's started a ground protection patrol. The commander and rest of the flight didn't reach the 1st waypoint but just circled around. Hope this small track file is enough to check.

[LukeFF]

13 hours ago, Dr1falcon500 said:

Fokker EIII mission.zip 3.04 MB · 1 download

This was a new career mission Jasta 6 Jametz, 27/9/16. Our EIII's started a ground protection patrol. The commander and rest of the flight didn't reach the 1st waypoint but just circled around. Hope this small track file is enough to check.

 

Per one of our mission designers, this is not a bug. The leader of the group is waiting for you, so he's circling.

[Dr1falcon500]

5 hours ago, LukeFF said:

 

Per one of our mission designers, this is not a bug. The leader of the group is waiting for you, so he's circling.

I was flying along with the flight not behind, when the rest started circling I kept going. They weren't waiting for me to catch up.