Regarding the AI issues - a request for help

[LukeFF]

18 hours ago, Dr1falcon500 said:

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.

 

What happened on takeoff? This is what the person who examined the files says:

 

"I saved the mission from the track, launched, took off, nobody's circling (watch video). There's no leader (Ludwig Haas) in the track. He probably crashed on takeoff and the group is waiting for him. Then it's a bug. It is better to record a track during this situation, not after. But that's my guess." 

[plepew]

I think it might be helpful to think about AI in specific bins in this discussion.  For example, 1) defensive (how well enemy aircraft maneuver out of your sights), 2) offensive (how well enemy AI get and keep you in their sights), 3) collective (how well enemy AI work together - for example come to each others aid or sneak up on your tail while you are trying to attack another aircraft).  And then a final catch-all of other aspects of AI like deciding when to engage/disengage, adapting an attack strategy by exploiting the weaknesses of specific aircraft types and leveraging the strengths of one's own, using clouds/sun to conceal oneself during an attack, etc.  

 

Another issue which needs to be better defined is what standard are we comparing against.   One standard would be for AI to fly in a historically realistic way.  This may not necessarily be challenging, but would leverage the tactics of the day. Remember scout/fighter pilots were literally learning on the job in 1915-18 and not every pilot was an ace.  Furthermore some of the higher scoring aces like McCudden and Richtofen got most of their kills by carefully choosing their engagements and only attacking when they had the advantage.  Both pilots saw little value in what they called trick flying. McCudden would spend hours stalking a single observation plane and usually get a kill in a single burst after maneuvering onto its tail unseen.   Other pilots were much more aggressive.  Rhys David and Ball, both British aces,  were known to go after enemy aircraft with little regard to the odds.  Both met an early end after racking up impressive kill counts.   

 

By 1918, as large formations of fighters would engage each other, maneuvering skill meant less and less.  In these furballs of 20-40+ aircraft, mid-air collisions were not infrequent and pilots evading attack by one EA were just as likely to pass across the sights of another one in the process.  Richthofen met his end in such a furball.  CPT Brown recounts diving on a formation of around 20 EA and soon had two on his tail, after evading his attackers, he happened to see a red triplane attacking LT May's Camel below him.  Brown was in the right place at the right time; diving, he fired a long burst at the triplane but lost sight of it after pulling up.   Richthofen crashed sometime later and of course there is quite a bit of controversy over whether Brown or a machine gunner on the ground, SGT Popkin, deserve the credit.   

 

Historically, the most successful aces preferred not to get caught in a maneuvering dogfight.  They would rather sneak up behind EA, attack out of the sun, and engage only when they had altitude and numbers on their side.    (In FC, these aspects of "AI" are currently best handled in the construction of the mission file rather than the way the AI pilots fly).

 

The alternative to "historically realistic" is simply to always have the AI optimized.  The latter is more entertaining but perhaps less authentic.  I don't think either option is wrong or right, it is a matter of personal taste.   My point is that it is important to recognize that not everyone is looking for the same outcome regarding the performance of the AI in the game.  There is little point in criticizing someone's personal taste.  Just understand that some people may be happy with the state of the AI while others are not.  Both points of view are legitimate. 

 

Another factor to consider is that AI is a limited resource.  The more aircraft there are flying, the thinner the AI gets spread across them.  This is readily apparent in complex missions with large numbers of AI aircraft.  Throw in all the other entities in a mission (ground units engaging one another, AA gunners, etc.) and you will notice the air to air AI suffers considerably.   This means that comparing the performance of AI in a small QMB engagement where nearly all the AI resources are focused on a small number of aircraft to a larger, complex mission is not an apples to apples comparison.   To make any progress on improving the AI in FC, it is important to standardize the conditions  so that we know to what degree the issues we see are due to the native AI in the game in its optimal use case, vs AI when CPU resources are stretched thin.   Both are important of course but improving AI in a complex mission is a different problem than improving AI in a 4x4 quick mission.   

 

Finally, my thoughts on the AI based on 4x4 QMB using FC aircraft.  (Note, I'll have to send tracks later)

1) Defensive AI - when flying an early war Halberstadt D.IIs AI set to random AI, against later war Sopwith Dolphins - enemy AI set to ace,  I think the enemy AI is decent.  They do their level best to get out of my sights by maneuvering around, diving, and climbing though I'm usually able to shoot them down if I'm persistent.   This is not too different than what I see in WW2 AC using the QMB.

2) Offensive AI - If they get me in their sights they do hit me with a good burst. Though the damage seems less substantial than what happens with WW2 AC,  I think that the guns during WW1 do less damage than the weapons of WW2, so I think the offensive AI is probably on par with what I see in WW2 QMB missions.     

3) Collective AI - This is the biggest difference I notice between FC and the WW2 QMB missions.  In any WW2 mission, if I'm chasing after an enemy AC for more than 1 min, I'm almost always attacked by another EA.  In fact, at the higher (veteran, ace) AI settings, ALL other EA will be after me.  While I'm cursing my wingmen for not helping out, it makes the QMB's very challenging to survive at those settings for me.   In FC, the EA rarely gang up on me making it not so difficult to pick them off one by one.   So, in my opinion, that's where the focus should be regarding improving AI.  

 

Hope this helps the discussion.  

 

 

 

Edited November 27, 2024 by plepew

[plepew]

I did some testing flying duel quick missions.  In each one, I flew the Nieuport 28 vs several German aircraft with their AI set at Ace.   I started the mission testing the defensive AI simply attempting to maneuver into a firing position to see how well the enemy AI was able to counter my attempts.  I did not fire on the enemy AC.   After several minutes of testing the defensive moves, I flew straight and level and waited for the enemy AC to get on my tail and shoot me down.  I did some low G turns to give it a little bit of a challenge.   After several turns, I just flew straight and level.   Overall, the enemy AI did a good job at defensive maneuvering.  They generally turned to counter my attempts to get on their tails and did a fair job only allowing me opportunities for deflection shots.  

 

Offensive AI was another story.  The DR1 and to a lesser degree the SSW DIV and the DVIII were able to shoot me down when flying straight and level.   The Albatros DV and Fokker DVII were unable to shoot me down when I was flying straight and level.  They approached from behind flying a serpentine pattern and as they got closer, they would take passing deflection shots as they turned back and forth behind my aircraft.  Very few shots hit my aircraft and none were able to shoot me down.  The attached track file illustrates this pattern of attack by the DVII.  

DVII_weavingattack.zip

Edited November 29, 2024 by plepew
corrected some errors

[plepew]

The track I saved of the Albatros DV was too large to upload.   Attached is the DR1.  In several tests it was always able to press the attack and flew straight and level when pursuing me from behind.  In this particular track when I was able to get on its tail, it seemed to be unaware and didn't take much evasive action.  When I dropped back a bit he resumed maneuvering again.   Generally, the DR1 does a good job in defense and in offense.  

DR1_defense_offense.zip

This is the track of the DVIII.  This aircraft has good defensive AI and the offensive AI is a mixed bag.  He did a lot less weaving and was effective at deflection shooting.  When I went straight and level, he first started weaving but as he got closer, he straightened out and took me out with a long burst. 

DVIII_defense_offense.zip

Here is the track with the SSWDIV.   He had excellent defensive AI and at one point when I was on his tail, he chopped his throttle and pulled up dropping behind me.  I thought that was pretty slick!   His offensive AI was good too although when I started flying straight and level, he pulled right up behind me and watched me for about a minute before opening fire to finish me off.  

SSWDIV_defense_offense.zip

[plepew]

Several tests in the WW2 era (me P51D vs ME109G14) using the same approach showed that the ME did no weaving but went straight in for the kill when I flew straight and level and with gentle turns.  The track files exceeded the 5MB limit so I couldn't upload.   

 

I'll continue testing more aircraft using this approach.  I think duel (1 vs. 1) is a good way to start picking apart the fighter AI because they have only one job  - to shoot you down.  As more aircraft are added it gets harder to replicate findings.  Also following a script of testing defensive vs offensive AI is helpful because it is easier to replicate the behaviors you observe.  

[Dr1falcon500]

AI Nieuport 17's flying against Halberstadt DII's all planes Ace AI. My N17 is on auto pilot all I'm controlling is the camera. Conclusions, very poor erratic flying,, shooting mainly when enemy is in front of gun sights. My AI pilot made only one attempt to turn after enemy. Stalled a few times and crashed from stall though no damage to cause this. Sorry no track files as I have removed Flying Circus from HD.

https://www.mediafire.com/file/yge8bkc0k95dzd1/Ace+AI+N17+FC.zip/file

[plepew]

Continuing with my testing.  My goal is to find repeatable instances of AI "foolishness" so that the Devs have something concrete to evaluate and test any fixes against.   To do that I'm continuing with my simple duel use case of me against an Ace flying various aircraft.  Friday I reported that the Albatross DV and DVII exhibited a "weaving", serpentine attack when I was flying straight and level but that  the DR1, SSW, and DVIII were all able to lock in and fly straight at me and attack.   Today I tested the latter batch and instead of their nice straight in attacks, they were all weaving too, so this seems to be a common issue in 1v1 duel mode.  Here are the tracks.  I flew multiple iterations and saw similar results.

SSWDIV_weave2.zip

Here is a track with the DR1 and a second one with the SSW showing the weaving attack.    I also noticed on two  occasions that the SSW would pull up into a steep climb and then dive straight into the ground from 2000m.  Not sure if the pilot blacked out or what happened.   I didn't record either event unfortunately.  

DR1_weave.zip SSWDIV_weave.zip

[plepew]

But something very interesting I observed.  When flying in a Skirmish (1 v 2) against two DR1's, I noticed they snapped out of the weaving attack and came straight at me.   In the first track (1v2_Nieu28vDR1s) you will see two DR1s behind me as I fly straight and level.  They are initially weaving back and forth but about 45 sec in, one of the DR1's stops weaving and levels out and heads straight at me.   10 sec later the other one does the same and once they catch up to me, they finish me off.  The second recording is another instance where they are clearly "locked on" and come straight for me.  I haven't tested flights against other types of aircraft, but so far it seems to be 

 

To keep these recordings short, I'm limiting them to the specific behaviors I'm observing, but I did continue to observe that the defensive AI seems quite good.  It is a challenge to get on the tail of the AI Ace pilots across all aircraft I have tested.  Even when they are doing their inept weave attack, If I start turning aggressively to resume my attack, they revert to effectively countering my turns and making my life difficult.   

 

Clearly more testing is needed against more types of aircraft.  I would encourage anyone out there posting tracks to try to identify specific, repeatable issues and the conditions for eliciting these AI pilot "errors".  The dev's will need these to both try to understand the underlying logic that is at fault and to then test any fixes to that logic to see if they can solve the problem.  

1v2_Nieu28vDR1s.zip 1v2_Nieu28vDR1s_2.zip

[LukeFF]

@plepew do you happen to have the mission files for any of your tests?

[plepew]

On 12/2/2024 at 4:51 AM, LukeFF said:

@plepew do you happen to have the mission files for any of your tests?

Yes.  I’ll post everything to my google drive and send a link.  

[plepew]

On 12/2/2024 at 4:51 AM, LukeFF said:

@plepew do you happen to have the mission files for any of your tests?

 

This folder has .trk files and mission folders.  filenames should be somewhat self-evident.  They include all the .trk file attached above and more.   Let me know if you have any questions.

https://drive.google.com/drive/folders/1GJswePUGwV0P_g8ubBIdvVWX9sW5S9a2?usp=sharing

 

[plepew]

Continuing with some testing.   Instead of just flying straight and level, I'm looking at turning and weaving defenses.   A turning defense is one where I simply maintain a bank angle and slowly fly in a circle.  The turn is very low g turn.  A weaving defense involves alternately banking left and right, also at a low g.   An ace AI should easily dispatch me with either defense.

 

In the folder linked below there are two files recorded while flying a Spitfire mkIX against an ME109G6-late set to ace.  One file is with a weaving defense and one is with a turning defense.  In both cases the AI 109 pilot easily and efficiently kills me.  This is exactly what I would expect.  

 

I did not see this level of efficiency with the WW1 aircraft.  When flying the Nieuport 28 against the Fokker DVIIF set to Ace, the AI DVIIF was unable to get a kill while I was doing a turning defense.  There are two examples of this.  In the second one, I rolled out to level flight halfway through and the DVIIF was still unable to score hits due to his constant weaving back and forth.  It is hard to tell, but this urge to weave seems to be the reason he could not get hits when I was turning. 

 

Interestingly when flying against a DR1 set to ace, the pilot was able to kill me in a turning defense and should have killed me in a weaving defense, however I got very lucky and despite a significant number of hits, my pilot was seriously wounded, but still alive.  Though the DR1 did get a kill, it still involved a rather circuitous pursuit - unlike the 109 that pulls right in behind and blasts away  - nothing fancy but very effective. 

 

The reason for looking at different aircraft is that I am not sure if the AI pilots are "better" at flying some models than others.  

 

https://drive.google.com/drive/folders/1ISQSyKz-C0hbmXua6GFeNJfAw6bybbaA?usp=sharing

[LukeFF]

Thanks for the files - I have passed them along to our QA team to look at.

[plepew]

On 12/5/2024 at 11:19 AM, LukeFF said:

Thanks for the files - I have passed them along to our QA team to look at.

 

No problem.  Happy to do any additional testing, provide additional info, etc.

[Rick_Rawlings]

Thanks for keeping on this, plepew! I even feel like if the enemy attacked more aggressively overall, you wouldn't notice individual mistakes if you constantly have to avoid being hit by other planes. As it is, enemies will wander off for minutes at a time allowing you to fly along and critique the AI of the plane nearest you at your leisure.

[Varibraun]

Hi @LukeFF - Here is a Google Drive Shared Folder with a Career Mission from yesterday (3 different tracks) with a really clear example of where the AI (enemy and friendly) refused to engage each other or me:

 

 https://drive.google.com/drive/folders/1e6W55rk-MKr-uOODP9azzVIf30rfvT13?usp=sharing

 

FYI - You can also see the SE5s parked in 1916.

[LukeFF]

Thanks! Reported

[Flying_Anchor]

@LukeFF career mode, my own flight ai does not engage any enemies at point blank contact, despite "attack near air targets" and "patrol for air targets" orders have been given.
This thread was made 1 year ago, but FC AI is still disasterously incomplete .
But these nieuports were quite good chasing me, didnt they?

 

https://drive.google.com/file/d/10UhJtwwWM0c-i2Wy8NpTg59-VLmtaSnY/view?usp=sharing

Edited December 28, 2024 by Flying_Anchor

[Flying_Anchor]

@LukeFF same issue, another FC career! My wing doesnt engage enemies.

 

https://drive.google.com/file/d/1GVywG0as4BWUOPYrWHHFl2GL-8o4QldA/view?usp=sharing

[Davesax1965]

The problem (I think) with modelling AI for a WW1 flight sim, as opposed to "game" is that if you want to do it properly, there are far too many variables. A lot of early war RFC pilots were almost totally untrained and would just panic and fly straight if fired on.

With aces, each one is different - compare Voss or Lothar Von Richtofen, who flew like maniacs, to other aces who would carefully calculate and stalk opponents, knocking them down with "energy kills". 

Also. Here's the question - do all Jagdstaffel follow the "Dicta Boelcke" and do all RFC / RAF pilots follow Mick Mannock's Rules - and if so, given that air combat was evolutionary, "from when ?" 

It's probably impossible to accurately model AI for a simulation. Obvious we're discussing "a game" here, but there is an important distinction to be made. What "proper" AI do you include - if any at all ? 

You couldn't model me in AI, by the way. I don't seem to have any "I" when I fly. 😉

 

Edited January 18 by Davesax1965

[artao]

The other members of a flight (when not flight leader, so can't give orders) get too easily distracted off mission. Bomber intercept missions in particular. I'm able to stay with the bombers and keep on-mission, take a couple out, so should they. But the rest of the flight is way back wherever circling around with the escorts. In fact, when the escorts show up, the rest of the flight turns into them in an offensive manner to go fight them. They should try to avoid those escorts and do whatever they can to get to those bombers and stop them. Don't even need to kill em, as long as the bombers simply turn back they can't complete their missions, which is after all our mission primary goal.

[LukeFF]

10 hours ago, artao said:

The other members of a flight (when not flight leader, so can't give orders) get too easily distracted off mission. Bomber intercept missions in particular. I'm able to stay with the bombers and keep on-mission, take a couple out, so should they. But the rest of the flight is way back wherever circling around with the escorts. In fact, when the escorts show up, the rest of the flight turns into them in an offensive manner to go fight them. They should try to avoid those escorts and do whatever they can to get to those bombers and stop them. Don't even need to kill em, as long as the bombers simply turn back they can't complete their missions, which is after all our mission primary goal.

 

If you have a mission file showing the issue I can pass it along to the team to see if anything can be done.

[Ace_Pilto]

Overall the AI prefers to remain aggressive in a fight it is not winning. More maneuverable aircraft always beat less maneuverable aircraft since the AI does not play to the strengths of the aircraft it is flying, it adheres to what seems to be an arbitrary rule set. (Turn this way at maximum rate until dead). I've seen entire flights of 109's die to I-16's because they let the I-16 kill them instead of simply disengaging. Same for the SPAD or Fw-190 or anything American. 

 

I understand that developing an AI for each type of aircraft is probably beyond any reasonable expectation. In a perfect world the AI would be made to fit the aircraft and trained accordingly and if I had millions of dollars I would pay for it myself. So, as an alternative, I would consider implementing three generic types of AI that conform to the tactical strengths of their aircraft in the context of the fight they are in.

Turn and burn (rookie)

Hit and run (veteran)

Hybrid (master)

And then tune the base models to conform to skill.

A poorly skilled pilot will get lured into a disadvantageous tactical position fairly easily, end up low and slow and get killed. A veteran will be more experienced in surviving but less offensively effective and an ace will use their aircraft aggressively and to great effect. What I would change is the minimum entry speed into maneuvers, especially with highly energy dependent, high wing loading, powerful  aircraft displaying a preference for extending, reaching optimal speeds, and climbing/diving away over turning. 

 

I would retain the angles/turn based AI that is currently modeled since it works for aircraft that can use it, but I would add an energy/verticality AI for aircraft that use those parameters better. 

 

Overall the AI isn't actually bad, it's just bad for a lot of the aircraft using it.

Edited April 3 by Ace_Pilto

[Ace_Pilto]

I mean watching a SPAD VII trying to out-rate an Albatros D.II just makes me cry.

[Dr1falcon500]

So since December 2023 the developers have had quite lot of track files and suggestions. Very little change or improvement except for one hotfix.

[ST_Catchov]

 

1 hour ago, Dr1falcon500 said:

Very little change or improvement except for one hotfix.

 

What did the hotfix do?

[Dr1falcon500]

3 hours ago, ST_Catchov said:

 

 

What did the hotfix do?

The one that increased AI head on attacks. That's all since this thread started. If there were any other changes they aren't apparent.

[Dusty926]

The patch loosening the AI's fright of pulling any Gs at all did help quite a bit, but unfortunately they still suffer from one critical behavior that I think makes everything else fall apart:

The swerving, wandering left and right pattern they do whenever following an enemy or attempting to line up anything not a deflection shot.
You'll notice when the AI is trying to get a target solution, that they will start swerving like they are evading gunfire. Makes enough sense when very close to a 2-seater, but 400+ meters behind a fighter makes it quite the comedic display. This not only contributes to their ineffective gunnery (since they will barely ever take any shots in this state), but I think is the #1 reason why they cannot kill the player in many scenarios, as well as failing to kill eachother even in prolonged engagements.

If this single issue were fixed, I would be such a happy camper - The AI obviously has many issues, but this is the one I run into by far the most, and I think is by far the most damaging.
If desired, I can provide track files and video recordings of this. Please message me if this is the case, because I admit I am a little frustrated leaving mine in the thread without prompting. 
Despite my frustration, I really want to do what I can to help find a solution to this problem - I've already really loved the FM reworks and tweaks, and the AI adjustments that have already been made. I want to show my support for more changes any way possible, especially to this IMO critical issue.

[czech693]

If you check the LauScripts/ai settings for each aircraft, the dogfight settings are the same for an Albatros D.II and a Spad VII (also in BoX, a Bf-109G-6 is the same as a P-51D-15), in each category (Novice to Novice, Ace to Ace).  Performance numbers are different.  So the two AI are going to use the same dogfight settings and the performance will determine who wins.  Missing in the Normal category in FC is the MaxPitchRate, which varies between categories in FC, but oddly is the same in all Dogfight categories in BoX.  Can't get my head around what MaxPitchRate represents since it's a measurement over time and plugging different rates into a calculator hasn't provided any insight.  An Albatros D.II Ace uses a MaxPitchRate of 180.0 and a Bf-109G-6 Ace uses a MaxPitchRate of 180.0.

 

Nowhere does there appear to be any setting for an AI to use a Turn and Burn or Boom and Zoom tactic based on his aircraft's performance.

[Dr1falcon500]

Given that IC's priorities are focused elsewhere which is the normal situation, I have to wonder what purpose this thread has served? Maybe to keep people quiet and let them think something will be done?  Customers have uploaded track files and complied with IC's needs. but it seems nothing is happening. This isn't a new problem but goes back a few years. There's always an excuse, not enough cash can't get workers or time or not a priority.

[cmbishop]

I completely agree with Dr1falcon500.
With FC, we have one of the best WW1 simulators and we are well aware of it. 

We have brought up videos and comments about the AI which is the subject of this post. What a pity to see such a beautiful simulator (great in VR) too often spoiled by an AI too often outdated... What a feeling of frustration! 

Isn't it possible to give modders access to the AI files, being well aware that the developers have a lot to do, especially with the next expansion (Odessa) and the next game (Korea)?

Best regards

[Trooper117]

15 minutes ago, cmbishop said:

Isn't it possible to give modders access to the AI files, being well aware that the developers have a lot to do

 

You've got 2 hopes...

[Aapje]

3 hours ago, cmbishop said:

Isn't it possible to give modders access to the AI files, being well aware that the developers have a lot to do, especially with the next expansion (Odessa) and the next game (Korea)?

 

The configuration settings can already be changed, and we have a mod to tweak the AI a bit.

 

But a fundamental change to the AI would require coding and 1CGS are not going to give people access to their code.

[cmbishop]

Which mod is it ? 

Are you talking about the modification available in PWCG ?

[cmbishop]

I post my message too quickly.

You must talk about Stonehouse's mod ?

[Dr1falcon500]

I don't know if QA team can explain this. PWCG mission.  Four Fokker EIII's took off and 3 curved away and crashed shortly after take off. It seems that AI cannot use a blip switch. If an EIII is behind you its necessary to take off to one side or risk having your plane getting crashed into. Quite a few missions have to be restarted because of these occurrences. The planes shouldn't be starting in two rows anyway. The EIII is almost unusable because of the frequent crashes and need to restart missions.

https://www.mediafire.com/file/s59uxrl3k7ohowd/Paul+Zerner+1916-08-01.2025-05-02_08-36-31_00.zip/file

[LukeFF]

PWCG missions are not our responsibility to troubleshoot. 

[Dr1falcon500]

7 hours ago, LukeFF said:

PWCG missions are not our responsibility to troubleshoot. 

Its not a PWCG issue mate its an issue of AI unable to use blip switch or fly with common sense. Also not a PWCG issue that EIII engine flagged as overheating straight after take off as here and is often the case. Engine overheats for no apparent reason quite often on take off for no reason. Planes start in a double line sometimes causing planes in the first row to be rammed. WW1 planes took of in one straight line. None of these things are written by PWCG. The buck stops with IC. I tried this and and the previous one with auto pilot one both occasions the plane behind hit mine and smashed its prop taking off. Not a PWCG issue. The responsibility is is with IC to address these issues.

Edited May 2 by Dr1falcon500

[LukeFF]

4 hours ago, Dr1falcon500 said:

Its not a PWCG issue mate its an issue of AI unable to use blip switch or fly with common sense. Also not a PWCG issue that EIII engine flagged as overheating straight after take off as here and is often the case. Engine overheats for no apparent reason quite often on take off for no reason. Planes start in a double line sometimes causing planes in the first row to be rammed. WW1 planes took of in one straight line. None of these things are written by PWCG. The buck stops with IC. I tried this and and the previous one with auto pilot one both occasions the plane behind hit mine and smashed its prop taking off. Not a PWCG issue. The responsibility is is with IC to address these issues.

 

Nope sorry. It's the responsibility of PWCG to code the mission correctly so they aren't stacked up behind each other on takeoff - and that is possible. 

 

Give me a stock career mode mission where these issues are present and I will pass it on to the team. 

[Dr1falcon500]

22 hours ago, LukeFF said:

 

Nope sorry. It's the responsibility of PWCG to code the mission correctly so they aren't stacked up behind each other on takeoff - and that is possible. 

 

Give me a stock career mode mission where these issues are present and I will pass it on to the team. 

Planes starting stacked behind each other is the in career and quick missions. Planes overheating on take off also has nothing to do with PWCG. EIII inability to use blip switch nothing to do with PWCG. AI inability to recognize a plane in front of them nothing to do with PWCG. So "nope sorry" doesn't cut it fella. You and IC take responsibility for a change. As the first month of Career Mode has been removed there is only one unit flying the Fokker EIII exclusively I think, Jasta 8. Just have too what happens in career.

Edited May 4 by Dr1falcon500