any work planned on dogfighting AI?

[PatrickAWlson]

1 hour ago, AEthelraedUnraed said:

Good explanation on why it's tough to program, but I just want to add that again, a lot of the things that people complain about are mission logic rather than AI.

 

Compare, for instance, a late-war Allied "armed recon" flight with a German attack on the bridge at Remagen. The Allied planes should attack anything and everything in sight, regardless of if that makes them deviate from their flight path. The German planes, on the other hand, should only deviate from their assignment as a last resort, i.e. if they're actively attacked by enemy aircraft. That bridge needs to be destroyed, no matter what it costs!

 

All this is possible with the current AI. As some people above have said, the AI itself is pretty good already in simple dogfighting scenarios (i.e. QMB). In more complex scenarios too, it's possible to make the AI do what you want. But that takes some mission building effort, as anyone who's ever created a somewhat complex mission knows. Add in all kinds of optimisations necessary to keep the game playable, and of course the mission logic itself. Hard enough for a mission designer to do ("Now, do I need to set the waypoint priority to Medium or High? And what if [...] happens, does that change things?"), harder for someone to program. Programming mission creation is not just programming, it's writing a program to do the programming for you.

 

That aspect of mission making, hand or programmatic, is not really all that hard.  That lone parameter, waypoint priority,  is about all of the control a mission maker has over the AI.  I use low for fighter ops and medium for everything else.  I would like to use high for level bombing but doing so shuts down the gunners, a situation that has been true since RoF.  I have integration level tests that generate missions and verify that WP priority is as I want it to be.

 

What this means in PWCG is that fighters returning from a ground attack mission will not be aggressive.  I want it that way.  They did their job and now they're going home.

 

What I have seen is planes not behaving in accordance with their WP priority settings.  Fighters that simply do not engage because the internal AI decisions override the mission planning decisions incorrectly, for reasons unknown.  It is IMHO not a pervasive problem (it used to be in RoF) but I have seen it happen.

[AEthelraedUnraed]

19 minutes ago, PatrickAWlson said:

That aspect of mission making, hand or programmatic, is not really all that hard.

And how many hours have you put into PWCG? With that in mind, tell me how writing a good mission generator is not a lot of work. It might not be "hard" as in "difficult", but it's certainly "hard" as in "laborious".

Edited April 25, 2022 by AEthelraedUnraed

[BlitzPig_EL]

My big issue with "them" isn't so much their flying, but their shocking ability to land long range hits on you.  I'm speaking of enemy fighters here, or even WW1 two seaters using their fixed forward guns.

 

To get satisfying flying from the AI I need to set them at ACE level, but to get believable gunnery, they need to be a step lower than that.  I don't want them to be the drunken and blind gunners that bombers have, but not every AI should be Carlos Hathcock either.  Believable outcomes should be the goal, not only for our beleaguered devs, but for those of us that toil to make missions for our fellow players to enjoy.

 

One area where I am pleased and surprised about is that, at least the AI tanks will, often give different out comes when running the same mission multiple times.  I know it's nothing I have done, because I'm still pretty much a n00b at mission building, it was very surprising to see.

Edited April 25, 2022 by BlitzPig_EL

[Gambit21]

2 hours ago, PatrickAWlson said:

 

That aspect of mission making, hand or programmatic, is not really all that hard.  That lone parameter, waypoint priority,  is about all of the control a mission maker has over the AI.  I use low for fighter ops and medium for everything else.  I would like to use high for level bombing but doing so shuts down the gunners, a situation that has been true since RoF.  I have integration level tests that generate missions and verify that WP priority is as I want it to be.

 

What this means in PWCG is that fighters returning from a ground attack mission will not be aggressive.  I want it that way.  They did their job and now they're going home.

 

What I have seen is planes not behaving in accordance with their WP priority settings.  Fighters that simply do not engage because the internal AI decisions override the mission planning decisions incorrectly, for reasons unknown.  It is IMHO not a pervasive problem (it used to be in RoF) but I have seen it happen.

 

Waypoint priority primarily, but in conjunction with the "Force Complete" MCU, along with proximity triggers for instance, you can achieve some complex behavior...relatively speaking. So imagine giving the "force complete - high" command, plus a high priority waypoint. This is the "bug out" logic that I use. I can however also put a 'medium' priority waypoint in the same place. By using a time plus a proximity trigger, I can tell the AI to reengage if the player is pursuing and within a certain range. This would keep the 'bugging out' AI from simply being a target drone at this point. There are many variations of this, and the only downside is the logic is difficult to test/debug.

 

I know you're aware of this, just writing for the benefit of others.

 

All that said, I don't want to confuse anyone. The "dog fight" AI (decision making once engaged) is hard coded, and the same across the board (Career, QMB etc etc)

We do have quite a bit of control over engage, disengage decisions etc if we want to take the time. (I can make a 190 decide to attack you from the other side of the map) The AI also has it's own hard-coded decision making logic. We can't set parameter in the editor for this other than AI skill level.

[PatrickAWlson]

@Gambit21 I do use something similar for for damaged planes - bug out back to base at high priority.  Therefore my statement "I never use high priority" was not accurate.  More accurate is "I never use high priority during the normal course of a mission". 

 

I also use a proximity trigger to keep enemies from following you deep behind your lines, with the exception of offensive patrols - which are supposed to follow you deep behind your lines.  Not sure if I put healthy planes that I want to turn back on high or medium priority, but medium makes more sense.

 

Still, mission defined behavior is basically an outline or suggestion.  In game AI is where the details are decided  and, sometimes, overrides happen.  So there are three basic sources of inputs: mission, use command, and in game decision making.  To the extent that things go wrong I think it is simply wires getting crossed.

Edited April 26, 2022 by PatrickAWlson

[silent_one]

I just know when I was playing BOB2 and Mig Alley the fights were more fun.  There were fights in Mig Alley where the maneuvers were as good as any human player. Il2 might be more realistic but its not more fun. Having enemies just fly round and round till you get bored or your engine overheats isnt fun.

[Eisenfaustus]

54 minutes ago, silent_one said:

I just know when I was playing BOB2 and Mig Alley the fights were more fun.  There were fights in Mig Alley where the maneuvers were as good as any human player. Il2 might be more realistic but its not more fun. Having enemies just fly round and round till you get bored or your engine overheats isnt fun.

Endless circles aren’t fun - but neither in qmb nor in career I experience them anymore.

[Yogiflight]

1 hour ago, Eisenfaustus said:

Endless circles aren’t fun - but neither in qmb nor in career I experience them anymore.

This is a definition thing. What is endless circling for you? I experience it quite often, that AI fighters fly several circles. But this might be, because they know, that I can't get a shot on them, flying a Bf110 E2, as long as I follow their circles. So circling makes sense for them, boring or not. If they go out of the circle before I do, they give me a chance to hit them.

[Dragon1-1]

Well, the classic Luffberry circle is the most basic component of a dogfight in this era. You keep turning until either you or your opponent fall out of the sky, or, if there's a large disparity in performance, until one gets on the other's tail. It's sometimes possible to use the low yo-yo to close up in the circle, although I'm not sure how well it works in Il-2.

[Eisenfaustus]

1 hour ago, Yogiflight said:

This is a definition thing. What is endless circling for you? I experience it quite often, that AI fighters fly several circles. But this might be, because they know, that I can't get a shot on them, flying a Bf110 E2, as long as I follow their circles. So circling makes sense for them, boring or not. If they go out of the circle before I do, they give me a chance to hit them.

You are right - if I have the worse turning aircraft and start following an enemy in a turn he‘ll continue it. If I have the better turner or more energy to burn they usually either dive away or try to force an overshoot. 
 

Yet I remember the times when AI regardless of aircraft or tactical situation would conduct an endless max performance turn… always…

 

This I haven’t seen in ages

[PB0_Roll]

I think delaying any further purchase until fixes (pilot harness not tightened enough for surviving anything but a kiss landing and smooth braking, nav lights seen from the moon, career missions illogical, career friendly AI illogical) may help focus the low manpower ressources on this fixing before thinking of producing more planes and maps with the same fun breaking issues remaining forever.

 

But I'm so weak that if Pacific is the next theaterI might buy nonetheless (and still not fly due aforementioned issues), oh well ?

Edited April 26, 2022 by PB0_Roll

[AEthelraedUnraed]

42 minutes ago, PB0_Roll said:

I think delaying any further purchase until fixes [...] may help focus the low manpower ressources on this fixing before thinking of producing more planes and maps with the same fun breaking issues remaining forever.

And without the sales associated with new planes and maps, who's gonna pay to fix these issues?

 

Not sure if you read, but according to Jason, if they fail to make their June deadline for BoN, their "ability to produce new products will be jeopardized". They can spend all their time fixing the issues you describe - and that'd be the last they'd do.

[PB0_Roll]

I'm fine with it, as long as these issues are finally fixed, so I can enjoy the game. 

 

I have no desire of any more maps or planes until they are, those I already bought (and don't fly due fun breaking issues) will do. I've paid enough.

 

Jason is an astute businessman, he'll succeed with any team he ties with or builds, if this one fails him.

 

Maybe Pacific needs a fresh start and a new engine, after all ?

 

 

 

[354thFG_Leifr]

7 hours ago, silent_one said:

I just know when I was playing BOB2 and Mig Alley the fights were more fun.  There were fights in Mig Alley where the maneuvers were as good as any human player. Il2 might be more realistic but its not more fun. Having enemies just fly round and round till you get bored or your engine overheats isnt fun.

 

Buddye really did wonders with the AI in BoB2, it is still the best I've encountered and experienced in any video game.

[Ribbon]

1 hour ago, AEthelraedUnraed said:

 

 

Not sure if you read, but according to Jason, if they fail to make their June deadline for BoN, their "ability to produce new products will be jeopardized".

I really hope that is not true, kind of doubt looking at recent Jasons posts and DD that they'll manage to release 3 flyable light bombers, two AI planes, map and career in two months timeframe.

Also in the past when WIP expansion got near release we already got next expansion announcement following preorder option.

 

About AI, it simply isn't challenging enough, you can set unlimited ammo and fuel, go alone and clean the map from enemies!

Current AI comes at very high performance cost undergoing realistic FM same as human but still being non-challenging at all (only good thing it's doing is running away in circles so it's more like pursuit rather than dogfight).

Which makes no sense, i'd rather have AI with cheating FM but more dangerous and challenging, plus ypu get better game performance.

What we have now is very bad tradeoff imo (AI FM vs.game performance vs.skill/challenge)!

 

When it comes to SP flying bombers and ground attackers is only choice for me now (and even that i stopped flying), that's why i hope we'll get something naval focused in next dlc.

 

 

 

[PatrickAWlson]

Per circling - the AI has improved massively.  If the AI has altitude it will not circle.  It does quite a bit of vertical maneuvering.  The AI circles when it gets low.  Sometimes it scissors.  Vertical maneuvers at this point would be a bad idea without the altitude or energy to pull them off.

 

In real life this is what happened when a dogfight got that low.  I think, in real life, it often did not get that low because the aggressor did not follow.  You know, suicidally exposing yourself to every machine gun and rifle on the front - friendly and enemy alike.  Those who did often paid (von Richthofen, Mannock, Preddy to name a few).  

 

For me the next step in AI is not endless circling when the AI gets low but giving me, the mission builder, the ability to command the AI to RTB when they drop beneath a certain altitude.  Further, improve performance such that I can flood the front with machine guns, making it historically dangerous to follow.  Finally, enhance the AI not to follow an escaping aircraft below a certain altitude, also definable in the  mission.

[tattywelshie]

AI is an interesting topic that’s for sure. I think to be fair it’s not too bad, yeah it could do with improvement, but really I think it can pose a decent challenge, especially when you have 9 AI planes trying to hunt you down. Re AI skill, I’ve read plenty of books from Luftwaffe pilots who’s massive amount of victories on the Russian front come down to the poor quality of Russian pilots, so one could argue, at least if your flying as Luftwaffe that the AI might be quite accurate on the Russian side ?

[AEthelraedUnraed]

32 minutes ago, =VARP=Ribbon said:

I really hope that is not true, kind of doubt looking at recent Jasons posts and DD that they'll manage to release 3 flyable light bombers, two AI planes, map and career in two months timeframe.

Well, he also said recently that they're more or less on schedule. So I guess that everything will be alright, but it does mean that they have little room for extra stuff that cannot be postponed.

 

34 minutes ago, =VARP=Ribbon said:

About AI, it simply isn't challenging enough, you can set unlimited ammo and fuel, go alone and clean the map from enemies!

Current AI comes at very high performance cost undergoing realistic FM same as human but still being non-challenging at all (only good thing it's doing is running away in circles so it's more like pursuit rather than dogfight).

Which makes no sense, i'd rather have AI with cheating FM but more dangerous and challenging, plus ypu get better game performance.

What we have now is very bad tradeoff imo (AI FM vs.game performance vs.skill/challenge)!

I don't agree. When set to Ace level, the AI poses quite a challenge (when flying an aircraft of similar performance than you). Perhaps you are a better pilot than I am and that's not the case for you, but given that the average IL2 player has much more dogfighting experience than the average WW2 pilot, I think the AI is fine.

 

12 minutes ago, PatrickAWlson said:

For me the next step in AI is not endless circling when the AI gets low but giving me, the mission builder, the ability to command the AI to RTB when they drop beneath a certain altitude.  Further, improve performance such that I can flood the front with machine guns, making it historically dangerous to follow.  Finally, enhance the AI not to follow an escaping aircraft below a certain altitude, also definable in the  mission.

With a bit of work, the altitude part is already possible by using a number of CheckZone Spheres. It's not ideal, and the altitude won't be exact, but it *is* possible

[Eisenfaustus]

1 hour ago, Leifr said:

 

Buddye really did wonders with the AI in BoB2, it is still the best I've encountered and experienced in any video game.

BoB2WoV had outstanding AI and the best campaign system I've ever seen. I even played it sometimes as a strategy game

 

A pitty it doesn't work properly from Windows 10 onward.

 

45 minutes ago, =VARP=Ribbon said:

Also in the past when WIP expansion got near release we already got next expansion announcement following preorder option.

That so? I think to remember the followup announcement only came AFTER the release - might be wrong though ^^

 

47 minutes ago, =VARP=Ribbon said:

Which makes no sense, i'd rather have AI with cheating FM but more dangerous and challenging, plus ypu get better game performance.

I get your drift - and the performance part is acually quite convincing - yet overall I have to disagree. Although the aforementioned BoB2 is a stellar example of immersive simplified AI the other Aspects of simulation have far advanced since then. The FM/DM nowadays are far more complex - and the G-System of IL2 ist extremely well done. I actually can't imagine the same tricks that worked so well in BoB2 would offer the same immersion in IL-2.

 

And I say that knowling that this design decision will prevent me from ever being part of a Furball of 80 fighters dogfighting around a combat box 100 flying fortresses in a GB environment.

[KevPBur]

@PatrickAWlson I've noticed, that in WW2 anyway, the fleeing from the a mission along the deck with a long line of enemy ai doing some form of conga across the map have gone. Rather now I find the ai stay at 2000m and spread out allowing you no chance to turn on them whilst they follow taking chances to drop on you in turn. Of course sometimes that can end at your home base other times they are happy you are out the fight and leave you alone. I've even manged to give them the slip with 'unexpected' changes of direction so they may not be as all seeing now either?

 

@=VARP=RibbonYou, I'm sure, are a far better virtual pilot than I, as I find the ai plenty challenging. Try in an La5 in a Stallingrad career, constantly outnumbered by 109's 3-4 to 1 on every mission. Each 109 pilot is an experienced veteran or ace and every friendly ai a rookie. I'm classing each mission I get back over the Volga a success and if anyone else in my flight returns as a minor miracle. Yes I wish my overly enthusiastice Soviet friends would run away from vastly superior odds rather then charging to thier patriot deaths but even that appears to be born out by historical accounts.

[Gambit21]

34 minutes ago, PatrickAWlson said:

Per circling - the AI has improved massively.  If the AI has altitude it will not circle.  It does quite a bit of vertical maneuvering.  The AI circles when it gets low.  Sometimes it scissors.  Vertical maneuvers at this point would be a bad idea without the altitude or energy to pull them off.

 

In real life this is what happened when a dogfight got that low.  I think, in real life, it often did not get that low because the aggressor did not follow.  You know, suicidally exposing yourself to every machine gun and rifle on the front - friendly and enemy alike.  Those who did often paid (von Richthofen, Mannock, Preddy to name a few).  

 

For me the next step in AI is not endless circling when the AI gets low but giving me, the mission builder, the ability to command the AI to RTB when they drop beneath a certain altitude.  Further, improve performance such that I can flood the front with machine guns, making it historically dangerous to follow.  Finally, enhance the AI not to follow an escaping aircraft below a certain altitude, also definable in the  mission.

Having an “altitude greater than/less than” trigger is indeed handy. I wish we had that here.

On the other hand using check zones lends a nice bit of unpredictability.

[tattywelshie]

9 minutes ago, KevPBur said:

@PatrickAWlson

@=VARP=RibbonYou, I'm sure, are a far better virtual pilot than I, as I find the ai plenty challenging. Try in an La5 in a Stallingrad career, constantly outnumbered by 109's 3-4 to 1 on every mission. Each 109 pilot is an experienced veteran or ace and every friendly ai a rookie. I'm classing each mission I get back over the Volga a success and if anyone else in my flight returns as a minor miracle. Yes I wish my overly enthusiastice Soviet friends would run away from vastly superior odds rather then charging to thier patriot deaths but even that appears to be born out by historical accounts.

 

Agree totally, I’m in a Yak1 in Stalingrad and was smashed by around 8 Romanian 109s, out of the flight of 8 that took off, 7 of my AI buddies bought the farm which really does add to the immersion as I reckon a similar thing happened in real life.

[PatrickAWlson]

2 hours ago, Gambit21 said:

Having an “altitude greater than/less than” trigger is indeed handy. I wish we had that here.

On the other hand using check zones lends a nice bit of unpredictability.

 

I'm thinking ... 

For each plane, what if we had a huge checkzone that covered the whole patrol area.  One such CZ for each aircraft.  

Activate the CZ when the flight achieves altitude.  In PWCG that would be the ingress WP.

Deactivate the CZ when the flight is coming home.  In PWCG that would be the egress WP.

The CZ is configured to encompass a very large diameter, but with an altitude limit

If the CZ triggers it tells the plane to RTB, just like the "onDamage" threshold.

 

Therefore, any plane configured this way will RTB when it gets below a certain altitude.  Planes that are flying low altitude missions would, obviously, not be subject to this configuration.  

 

Looking at the CZ it does not seem to have a ceiling but it does have a floor.  Therefore the CZ has to be defined as where the plane should be and not where the plane cannot be.  Set the CZ to trigger on "further".  So ... set the floor to 1500 and the zone to be a pretty large area.  If the plane goes outside of the zone or below the zone then it RTBs.  One nice thing about this is that it causes RTB not only for altitude loss but also for straying too far from where the plane should be.

 

Thoughts?

Edited April 26, 2022 by PatrickAWlson

[Gambit21]

Yep. only I’d be tempted to use maybe 4, with one chosen randomly. That way the low point isn’t always in the same place. Further the RTB response randomized as well.

 

I  created a radar for use over the continent like this, only instead of a single check zone, I made a grid of large ones. All connected to the player. When the player is up in the grid, 500’ AGL +-, he is swept by the radar every handful of seconds (I actually have it rotating/pulsing) and he hears the interference over the radio.

 

If he drops out of the grid, the radar can no longer detect him, the pulse-tone stops. The radar logic is piped into the “plotting table” that may or may not send a fighter to intercept depending on how busy it is (dice roll)

 

Then that dispatched fighter will lose contact if you subsequently drop below 500’ or so.

 

Don’t know if you’ve had the experience, but it’s the kind of thing where I go “I don’t know how I figured all of that out!” and it would take me some time to duplicate this thing from scratch if I even could. 
 

Using a single check zone was problematic because even a very large sphere in the map area still has a relatively small “low” point. So I chose to use multiple overlapping spheres.

 

For what your describing I think a single one would do - but again randomized.

 

 

[LukeFF]

4 hours ago, tattywelshie said:

Agree totally, I’m in a Yak1 in Stalingrad and was smashed by around 8 Romanian 109s, out of the flight of 8 that took off, 7 of my AI buddies bought the farm which really does add to the immersion as I reckon a similar thing happened in real life.

 

I was about to say pretty much the same thing - if people think the AI sucks, play a Soviet fighter career on the Stalingrad map on no lower than Medium difficulty. I'm lucky to survive a handful of missions before I inevitably am captured or killed.

[tattywelshie]

31 minutes ago, LukeFF said:

 

I was about to say pretty much the same thing - if people think the AI sucks, play a Soviet fighter career on the Stalingrad map on no lower than Medium difficulty. I'm lucky to survive a handful of missions before I inevitably am captured or killed.

Definitely, I’m about 5 missions in, and so far haven’t returned to base with an intact aircraft. It’s only a matter of time before I’m finished off, I don’t think I’ve shot a single Jerry/Romanian down without another one rather unsportingly shooting at me and putting holes in my kite, swines! 

[Gambit21]

@PatrickAWlson

 

Thinking still. You could pipe that zone into a proximity trigger as well. You wouldn't want an enemy AC just cutting for home when the player or another if your wingman is 200 meters on his 6, just wouldn't happen really as that's suicidal. So your 'lower than' logic, plus Proximity/Greater than (500m?) = bug out.

Then you can get really fancy but activating another proximity trigger of the player closes, forcing him to re-engage.

 

In actual prictice I normally don't do this kind of thing just because testing/debugging becomes cost prohibitive from a time standpoint.

[PatrickAWlson]

19 minutes ago, Gambit21 said:

@PatrickAWlson

 

Thinking still. You could pipe that zone into a proximity trigger as well. You wouldn't want an enemy AC just cutting for home when the player or another if your wingman is 200 meters on his 6, just wouldn't happen really as that's suicidal. So your 'lower than' logic, plus Proximity/Greater than (500m?) = bug out.

Then you can get really fancy but activating another proximity trigger of the player closes, forcing him to re-engage.

 

In actual prictice I normally don't do this kind of thing just because testing/debugging becomes cost prohibitive from a time standpoint.

 

I would have them cut for home on a medium priority WP.  That way they would continue to fight if threatened.  Similarly, if they were close enough that they were about to kill you they would probably try to finish the job.  It wouldn't be perfect but still (hopefully) a significant improvement.

[Gambit21]

14 minutes ago, PatrickAWlson said:

 

I would have them cut for home on a medium priority WP.  That way they would continue to fight if threatened.  Similarly, if they were close enough that they were about to kill you they would probably try to finish the job.  It wouldn't be perfect but still (hopefully) a significant improvement.

 

That works. Then when/if the player closes again would you repeat the logic? Can go down a logic birds nest very quickly.

I do like that about the editor though, to an extent you can do almost anything you want. We're only missing a few key triggers.

 

[FuriousMeow]

11 hours ago, PB0_Roll said:

I'm fine with it, as long as these issues are finally fixed, so I can enjoy the game. 

 

I have no desire of any more maps or planes until they are, those I already bought (and don't fly due fun breaking issues) will do. I've paid enough.

 

Jason is an astute businessman, he'll succeed with any team he ties with or builds, if this one fails him.

 

Maybe Pacific needs a fresh start and a new engine, after all ?

 

 

 

 

And none of that will happen. They will finalize and close out the remaining main items for BoN - remaining planes, map, campaign, and maybe a bug or two - and that's it. The majority of issues you care about won't be fixed. There's no new engine, no Pacific, no further development. It ends there. This fails, that's likely it for WWI/WWII/Tanks (not that I care about tanks, but I bought to fund the team) from this team and Jason. This publisher will see this as good while it lasted and that's it from them. Aviation/wars from 80+ years ago aren't massive money makers. It was incredibly difficult to get this far, and something this advanced.

 

[DBFlyguy]

3 hours ago, FuriousMeow said:

 

And none of that will happen. They will finalize and close out the remaining main items for BoN - remaining planes, map, campaign, and maybe a bug or two - and that's it. The majority of issues you care about won't be fixed. There's no new engine, no Pacific, no further development. It ends there. This fails, that's likely it for WWI/WWII/Tanks (not that I care about tanks, but I bought to fund the team) from this team and Jason. This publisher will see this as good while it lasted and that's it from them. Aviation/wars from 80+ years ago aren't massive money makers. It was incredibly difficult to get this far, and something this advanced.

 

 

[I./JG52_Woutwocampe]

I flew in BoBP earlier today, I noticed that my fellow 109's are asleep at the switch very often in this career, so I observed their behavior more closely. 

 

We engaged 4x lightnings after an extremely long flight to the target waypoint, it was a free hunt. Very quickly, I shot down a lightning, I got the mission objective complete message right away and thats exactly when my wingmen entered their coma. The waypoint changed from the freehunt target to the next navigation waypoint to go home and despite still having 3x lightnings in the close vicinity and 4x more thunderbolts jumping in a bit later on, at this point, all my wingmen could do was hover at low throttle above me fighting 7x american aircrafts. 

 

Ordering them to engage or cover me had no effect.

 

So clearly, when the mission objective changes, or when its completed, something, sometimes, completely prevents AI from engaging any enemy anymore. Its not always the case though, but it happens very often in BoBP.

 

A way must be found so the enemies around will remain the priority above new navigation orders or objectives, especially if the flight leader orders it!

Edited April 27, 2022 by I./JG52_Woutwocampe

[FuriousMeow]

Spoiler

1 hour ago, DBFlyguy said:

 

 

 

[PB0_Roll]

6 hours ago, FuriousMeow said:

 

And none of that will happen. They will finalize and close out the remaining main items for BoN - remaining planes, map, campaign, and maybe a bug or two - and that's it. The majority of issues you care about won't be fixed. There's no new engine, no Pacific, no further development. It ends there. This fails, that's likely it for WWI/WWII/Tanks (not that I care about tanks, but I bought to fund the team) from this team and Jason. This publisher will see this as good while it lasted and that's it from them. Aviation/wars from 80+ years ago aren't massive money makers. It was incredibly difficult to get this far, and something this advanced.

 

Read this since 2004 (forgotten battles).

 

Don't believe it any more, never happened.

 

Wallet vote did get efficient quite a number of times though, otherwise you'd still need to unlock planes to be able to fly them in this game, or get early access purchases unflyable for years in the other sim.

[PatrickAWlson]

18 hours ago, FuriousMeow said:

 

And none of that will happen ...

 

 

True and not true

 

Not true.  There have been a steady stream of improvements funded by, but not directly related to purchased product.  AI is the perfect example - you can't help but notice how much better the AI is compared to three years ago.  Various graphics improvements, clouds and plenty more.  

 

True.  AI still has a long way to go.  Performance improvements have been inadequate.  Bugs can linger for a very long time.

 

Here's my take: 1C has to create and deliver the modules to make money.  No money, no business, no product - the end.  They would like to do all of these other things and they do try, but they have to prioritize the modules that actually make money.  The money made is not enough for a staff of hundreds or even dozens, so things linger.  1C does not know how to monetize the other improvements.  Maybe they can't.  This is a complex sim, not easily accessible.  That means limited potential revenue streams.  

 

The team is too small.  The revenue to grow it will never really be there due to limited potential revenue streams.  The only way that I can think to solve the problem is almost certainly a non-starter.  Open source.  Put the code online and accept contributions.  Meanwhile, focus the team on the modules that make money.  Anybody want to take bets on that happening?  

[Irishratticus72]

2 hours ago, PatrickAWlson said:

 

True and not true

 

Not true.  There have been a steady stream of improvements funded by, but not directly related to purchased product.  AI is the perfect example - you can't help but notice how much better the AI is compared to three years ago.  Various graphics improvements, clouds and plenty more.  

 

True.  AI still has a long way to go.  Performance improvements have been inadequate.  Bugs can linger for a very long time.

 

Here's my take: 1C has to create and deliver the modules to make money.  No money, no business, no product - the end.  They would like to do all of these other things and they do try, but they have to prioritize the modules that actually make money.  The money made is not enough for a staff of hundreds or even dozens, so things linger.  1C does not know how to monetize the other improvements.  Maybe they can't.  This is a complex sim, not easily accessible.  That means limited potential revenue streams.  

 

The team is too small.  The revenue to grow it will never really be there due to limited potential revenue streams.  The only way that I can think to solve the problem is almost certainly a non-starter.  Open source.  Put the code online and accept contributions.  Meanwhile, focus the team on the modules that make money.  Anybody want to take bets on that happening?  

 

[Varibraun]

I am late to the party on this thread...but I can echo what most have said above - this DEFINITELY isn't the same circling AI of a few years back.  The improvements have been gradual, but dramatic.  I base this on approximately 2000 hours here, the vast majority in SP.  I just came out of a PWCG mission where a La-5 AI pilot kept me in a grey-out with vertical combat and also used his throttle and flaps to try an force an overshoot.  No circles for this guy.

 

I also think the mission logic that @PatrickAWlson and @Gambit21 are discussing here somehow make the AI fight better and more immersive than what the folks just using the QMB are seeing.  The fight I mentioned with the La-5 in PWCG had my 4 flight of G-4s up against a swarm of P-40s, La-5s, Lagg-3s, and PE-2s (probably 12-14 total enemy aircraft).  It turned into quite the furball with a mixed level of AI pilots and I found the AI behavior realistic with historical accounts.  I do directly attribute this both to the Devs improvements of the AI logic, but also to Pat's tireless work to get them to behave appropriately. 

 

Bottom line - I find the AI experience in PWCG to be pretty darn good (but I am comparing them to historical accounts vs. our cadre of online player Experten).

[dburne]

Yeah the AI has indeed come a long way the last few years in this sim.

[Stonehouse]

4 hours ago, Varibraun said:

I am late to the party on this thread...but I can echo what most have said above - this DEFINITELY isn't the same circling AI of a few years back.  The improvements have been gradual, but dramatic.  I base this on approximately 2000 hours here, the vast majority in SP.  I just came out of a PWCG mission where a La-5 AI pilot kept me in a grey-out with vertical combat and also used his throttle and flaps to try an force an overshoot.  No circles for this guy.

 

I also think the mission logic that @PatrickAWlson and @Gambit21 are discussing here somehow make the AI fight better and more immersive than what the folks just using the QMB are seeing.  The fight I mentioned with the La-5 in PWCG had my 4 flight of G-4s up against a swarm of P-40s, La-5s, Lagg-3s, and PE-2s (probably 12-14 total enemy aircraft).  It turned into quite the furball with a mixed level of AI pilots and I found the AI behavior realistic with historical accounts.  I do directly attribute this both to the Devs improvements of the AI logic, but also to Pat's tireless work to get them to behave appropriately. 

 

Bottom line - I find the AI experience in PWCG to be pretty darn good (but I am comparing them to historical accounts vs. our cadre of online player Experten).

This is WIP mod I put up the other day. Based on an earlier work by Halfoat and expanded in this version with some additional changes. Intended for SP or Coop with a pre agreed set of mods. I'd be interested to get feedback as to whether people feel it was a improvement or detriment to their combat experience. Generally it changes the max firing range of AI for each skill as well as a change to what I believe is the number of hits taken before an AI will dodge around. Note it doesn't change air gunner (aircraft turret) behaviour.

 

 

Edited April 28, 2022 by Stonehouse

[AEthelraedUnraed]

5 hours ago, Stonehouse said:

This is WIP mod I put up the other day. Based on an earlier work by Halfoat and expanded in this version with some additional changes. Intended for SP or Coop with a pre agreed set of mods. I'd be interested to get feedback as to whether people feel it was a improvement or detriment to their combat experience. Generally it changes the max firing range of AI for each skill as well as a change to what I believe is the number of hits taken before an AI will dodge around. Note it doesn't change air gunner (aircraft turret) behaviour.

I haven't tried it yet, but I must say it looks very impressive and well thought-out! I agree that it sounds more like what historical accounts say about aces vs rookies. However I think this thread is mostly about the tactics rather than gunnery of the AI