AI vs AI Kill ratio

[major.kudo]

I kept thinking DT would settle this problem sometime and I waiting so far.
But it wasn't settled at all, so I decided to propose.

This is the picture I drew before.
http://gachopin.no-ip.info/kudo/Fighter_2.jpg
Maybe it was insufficient to tell my true meaning only with this picture.
The one I wanted to say is Kill ratio by a war between AI.
I experimented. In Fw-190s vs P-51Ds fight. skill is average.

16 vs 16 = 10 Fw-190s killed : 13 P-51Ds killed
32 vs 32 = 18 Fw-190s killed : 22 P-51Ds killed

The next things finds out that AI fighter in v4.13 will be observed attentively.

1. Pilots of average and rookie are shoot there guns strangely correctly.
2. AI can't avoid the bullets by AIs gun fires. Even if maneuver is performed, bullets hits.
3. A lot of planes are falling down one after another consequently.

I think Kill ratio between AI in v4.13 is too distant from a real war.
I hope hard repair this problem.

Everyone. What do you think?

-

Maybe this will be helpful reference data.
https://en.wikipedia.org/wiki/Confirmation_and_overclaiming_of_aerial_victories_ during_World_War_II
Airplanes aren't shotdown so much in once of air combat.

Sorry my poor English.

[gaunt1]

I think you are completely right about this! Average and especially rookie AI are way too accurate. They also really like head on shooting, and they are also extremely accurate in this.

[Furio]

In my opinion, overkill has little to do with the topic of AI kill ratios. Personally, I always considered kill claims simply not credible, period.

Returning to topic, I think the problem is combat duration. In real life, air combats were usually very brief, often lasting just a few seconds, and usually ending with opponents losing sight of each other. In game, AI pilots never lost sight of each other, and never quit combat when they should, because of damage, bad tactical position or low fuel level. In my opinion, these are the main reasons of abnormally high kill ratios. As a consequence, I don’t believe an easy fix could be implemented in game, at least one that pleases everyone. A typical, realistic mission will entail long, boring navigation and very brief combats, often inconclusive, followed by another long and boring return home.
My preference, for what is worth, would be to try some steps toward realism. AI’s eyesight could be gradually reduced or – better yet – player could have the option to choose different distances, the same way as AI pilots experience is chosen.

Also, I would like a “combat quitting” routine implemented. I know that’s difficult to determine an exact sequence of conditions that trigger quitting, but at least some could be tried. Any engine damage, particularly in the cooling system, should trigger a “Return to base immediately”.

[Pursuivant]

Its no secret that AI vs. AI kill ratios reflect unrealistically aggressive pilot behavior.

That is necessary to make a fun game, but isn't at all realistic historically.

Historically, at least for the USAAF, only 1 in every 10 fighter pilots shot down a single plane during their entire military career! Only something like 1 in 1000 made ace. Numbers were roughly comparable for other nations. That means you have a very few "eagles" and a whole lot of "turkeys."

Furio makes good points, but in addition to all the other advantages that AI aircraft have they never have moral problems about killing, they never get scared, they never get tired (air combat - especially a hard turning dogfight - was quite fatiguing), and they never think about their overall mission or tactical situation (i.e., "Am I going to get attack while I'm attacking?" "Will I have enough fuel after this dogfight to get home?")

There should be an option in the QMB or FMB to allow historical levels of pilot aggression, with the vast majority of pilots being quite cautious, and very aggressive pilots (the sort that become aces or die trying) being unusual.

[major.kudo]

- Gun shooting of AI more inaccurately. In particular, Rookie and Average.

- AI Decision to "return to base immediately" by more less damage.

- If the prudence of AI can be set at QMB and FMB, it's more better.

Isn't there a way which raises the survivability of these other AI?

[Pursuivant]

- Gun shooting of AI more inaccurately. In particular, Rookie and Average.

Possibly. I think that AI accuracy levels for fighters are more or less right, but rookies still seem to be too good at estimating distance and deflection.

I've also argued that there should be a level of quality below Rookie to simulate pilots who have no business being in a combat zone. (e.g., RAF fighter pilots fresh from training school in summer of 1940, many Soviet replacement pilots in 1941-42, and most Japanese and German replacement pilots from 1944 on).


- AI Decision to "return to base immediately" by more less damage.

Two issues here.

First, in the heat of combat you might not be aware that your plane has has taken some types of critical damage - other than obvious signs like fire or sudden loss of power output or maneuverability.

Second, in a dogfight, unless you can plan your disengagement and successfully implement it, you don't have any option other than to fight until the end.

But, I don't think that IL2 is modeling either of those things. Any engine damage, any fuel leak, any pilot injury, or serious damage to any other aircraft systems should be an automatic disengagement from a dogfight for all but the most aggressive or desperate pilots.

Average or rookie pilots should have the strong possibility of panicking when their plane is damaged, meaning that they disengage even when it is tactically unfavorable to do so.

Exceptions might be made for bombers, where often it made more sense to stick with the formation than fly back to base alone through hostile territory.



- If the prudence of AI can be set at QMB and FMB, it's more better.

I'd love to see this and I've been nagging TD for it for years. Currently, individual AI pilot qualities are modeled within the game, but there's no way for mission builders to access them. I don't know if the limits of the game engine prevent this, or if TD just doesn't have the resources to build the necessary user interfaces.


Isn't there a way which raises the survivability of these other AI?

In the game, no, other than aggressive management of your wingmen to make them break off attacks.

For the future:

1) Option in FMB (possibly QMB) for certain aircraft to fight more defensively than normal.

- Extremely Defensive - Avoid enemy planes, break off combat if attacked, don't take anything other than the easiest shots (e.g., less than 10 degrees of deflection, within 100 meters for most weapons, 50 meters for LMG).

RTB if pilot or crew is injured, if there's engine damage or fuel leaks, if the radio is shot out, or if there's serious damage to other aircraft systems.

- Defensive. Avoid enemy fighters unless you've got numeric, altitude, and positional advantage, don't make repeated attacks, break off from dogfights if it is safe to do so, don't take anything other than easy shots (e.g., less than 20 degrees of deflection, within 200 meters for most weapons, 100 meters for light MG).

Against bombers, only attack if you've got altitude and positional advantages. Make a single attack from a quarter where the bombers' defensive fire is weakest, at maximum effective range for your weapon (300-500 meters for cannons, 300 meters for HMG, 200 meters for LMG) and then disengage.

Return to Base if there's engine damage, injury to pilot or crew, serious fuel leak, or serious damage to any other aircraft system.

- Aggressive. Avoid enemy planes unless you've got at least one of numeric or altitude advantage, break off from dogfights if you're disadvantaged and its safe to do so.

RTB if there's engine damage, serious fuel leak, serious damage to pilot or crew, or serious damage to any other aircraft system.

- Very Aggressive. Current AI.

2) Introduce more "Human Factors"

- Rookie and Average pilots might "freeze" or hesitate rather than taking shots. (This can also simulate pilots forgetting to charge their guns due to over excitement.)

- Rookie and Average pilots might lose their nerve - especially when attacking large formations of enemy planes. They either refuse to attack or take shots from excessively long ranges.

- Rookie and Average pilots, and possibly hard pressed Veteran pilots, might panic, causing them to break off combat even though its tactically disadvantageous to do so. This is particularly true if the pilot is wounded, his plane is damaged, and/or he is attacked by surprise.

- All pilots should suffer the effects of fatigue during prolonged combat, particularly at altitude, or in fights where they pull lots of Gs. Once fatigue reaches a certain level, maneuverability and ability to pull further Gs suffers. When fatigue reaches a critical level, pilots will attempt to disengage from combat.

- Introduce Discipline Levels.

- Undisciplined. Pilots will regularly ignore their superiors' commands, particularly if doing so would endanger their lives. They will automatically break formation if attacked. If they break off combat, or refuse to engage in combat, they will ignore commands to return. If they engage in combat, they will refuse commands to disengage. If they break formation, they will eventually rejoin in their own good time.

Undisciplined + Aggressive pilots will break formation to chase enemy planes, assuming other conditions are favorable.

- Disciplined. Current AI.

- Very Disciplined. Pilots automatically obey their superiors' commands unless physically unable to do so. If they RTB, they can be ordered back into combat, even if it is suicidal.

3) Introduce more tactical options in the FMB.

- Alter course to avoid: fighters, bombers, flak, ships, ground vehicles, cities. This option allows planes to be even more defensive, as well as simulating planes which are trying to avoid being spotted.

- Seek & Destroy: fighters, bombers, flak, ships, ground vehicles. Planes will go out of their way to find targets of the specified types, and will attack them preferentially.

- Ignore: fighters, bombers, flak, ships, ground vehicles. This option allows for suicidally heroic "straight in" attacks.

- Hold formation: Planes hold formation no matter what, unless incapable of doing so, or ordered to break formation by their commander. Typical behavior for most bombers, doctrine for early war British and Soviet fighter pilots, and for early to mid-war Japanese fighter.

- Maneuver to avoid: fighters, flak. This option allows planes to maneuver within formation, or break formation, if attacked. If "hold formation" and "maneuver to avoid flak" options are combined, the entire formation will randomly "zig zag" to spoil flak gunners' aim.

[gaunt1]

I'd add one point: Eliminate the willingness of taking other planes head on, except for ace AI. Or except maybe against bombers.

[Pursuivant]

I'd add one point: Eliminate the willingness of taking other planes head on, except for ace AI. Or except maybe against bombers.

I wouldn't eliminate it outright, but I agree that it needs to be modified.

In addition to the situations described, a head-on attack is a safe shot if taken from outside of the enemy's cone of fire. It's also a "fairly safe" shot if you're flying a heavily-armored, radial engine plane against a lightly-armed opponent.

For example, if I'm flying a P-47, F6F or F4U vs. a Ki-43, then I won't hesitate to take 12 o'clock level shots. It saves the hassle of trying to shoot an agile target using BnZ tactics.

When flying a well-armored but sluggish plane like the Buffalo Mk I, F2A2, or IL2, head-on shots are just about your only hope of shooting down faster or more agile opponents.

* In all cases, if AI initiates or accepts a head-on attack, an attack from 12 o'clock high or low (or whatever angle avoids the most fire) is preferred.


- AI Head-on attack preferred:

Average or better quality fighter pilots vs. heavy bombers (as long as the bombers have effective rear quarter defenses).

Aggressive or Very Aggressive Veteran or Ace pilots flying planes which are inferior to their opponents in both speed and maneuverability. (Cautious or Very Cautious will attempt to flee/hide in the clouds and use ambush tactics. Aggressive or Very Aggressive Rookie or Average pilots will attempt to use maneuver or BnZ tactics to gain a rear-quarter shot and die trying.)


- AI Head-on shot accepted if initiated by enemy/occasionally used:

Aggressive or better Veteran or Ace pilots flying planes which are superior in armor, firepower and engine durability. Chances increase if the plane they are flying is inferior in speed or maneuverability to their opponent.

Average or better pilots flying planes against light or medium bombers, but only if the plane they are flying is superior in armor, firepower and engine durability. (Rookies will attempt rear quarter or beam attacks.)

Sometimes, Very Aggressive or Aggressive Average or Rookie pilots will initiate or meet head-on shots rather than breaking, even if their plane is inferior in armor, firepower, or engine durability, or is superior in speed and/or maneuverability. (That is, their aggression will get the better of them, so that they make the mistake of playing the enemy's game.)


- AI Head-on shot avoided/never initiated:

Almost any pilot flying a plane which is inferior in frontal armor, forward firepower or engine durability, or superior in speed and/or maneuverability, except as described above. (The idea here is that AI should almost never accept a clearly unfavorable tactical situation if it's possible to get a better firing solution on the target.)


* In any case, all but the most aggressive and experienced pilots will start head-on attacks too soon (at 1,000-600 meters) and break off the attack to avoid fire and maneuver for position before they get within about 200 meters. Aggressive or better pilots of Veteran or higher quality will start the attack within 600 meters and will only break off the attack at the minimum distance required to avoid a collision.


* Rookie and Average pilots will have extra trouble with range and deflection estimation for head-on shots, since it wasn't typically part of gunnery training. They might even fail to line themselves up for the shot!

* All but Aggressive or better pilots will break off head-on attacks if they take any damage during the attack run.

[majorfailure]

First, in the heat of combat you might not be aware that your plane has has taken some types of critical damage - other than obvious signs like fire or sudden loss of power output or maneuverability.

Second, in a dogfight, unless you can plan your disengagement and successfully implement it, you don't have any option other than to fight until the end.

But, I don't think that IL2 is modeling either of those things. Any engine damage, any fuel leak, any pilot injury, or serious damage to any other aircraft systems should be an automatic disengagement from a dogfight for all but the most aggressive or desperate pilots.

Average or rookie pilots should have the strong possibility of panicking when their plane is damaged, meaning that they disengage even when it is tactically unfavorable to do so.


What totally works against any and all attempts to disengage is the omniscient AI, who when once has acquired a target only loses it when out of range. Humans and AI cannot disengage at will - and even trying is pointless. If there would be a chance that AI loses track of their target, because of it flying into a blind spot, or getting distracted by other potential targets, or because their leader commands them to do so - then RTB could be worth the risk when you are damaged.

I'd add one point: Eliminate the willingness of taking other planes head on, except for ace AI. Or except maybe against bombers.
Against bombers with good defensive fire from the rear all but dumbest rookie should do head on or slashing passes, given plane allows for that.
Against fighters it should not depend on skill level -only on agressiveness - I'd bet the average Japanese pilot was quite willing to take the head on shots - despite sitting in a lightly armored plane. And while sitting in a heavily armed, heavily armoured plane may make you more confident to go head-on, an ace in just such a plane may avoid it unless any other maneuvre will favour the enemy.

I do not accept head on exchanges against AI fighters - except if I'm in
a) a massivley better armed plane - and not in a very good position, e. g. the possible quick kill is worth the risk. E.g. Fw190 vs. Yak.
or
b)my plane is just so much worse, I can only go for head-ons. Typically used for IL-2 vs fighters.

[Pursuivant]

What totally works against any and all attempts to disengage is the omniscient AI, who when once has acquired a target only loses it when out of range.

I'm not sure this is true. 4.12 finally got AI "target acquisition" more or less right, and it seems that Rookie or Average pilots can be pretty clueless if you get into their blind spots.

What doesn't seem to be modeled is loss of Situational Awareness (SA) due to target maneuvering, information overload, and other factors.

A "simple" way to make loss of SA possible would be this:

Baseline ability to retain SA is based on pilot quality.

When an enemy aircraft flies into a plane's blind spot, there's an x% chance every second that AI will lose SA with respect to that plane, based on the target's range and speed. (Elements like target size, camouflage, visibility, etc. don't apply because this algorithm attempts to model the human ability to mentally track targets you can't see.)

This base chance is modified upwards if the spotting plane is damaged, under attack, or pulling Gs, or if the pilot (or some of the crew) are wounded. If the other plane is pulls Gs, or gains or loses altitude while in a blind spot, there's also an upward modifier. (This is also a simplification which represents that its harder to keep track of a fast or wildly maneuvering target.)

If the other plane is smoking, leaking fuel or coolant, or contrailing, there's a big downward modifier. (This represents a pilot's ability to track a plane, and follow its path, by using its smoke or vapor trail.)

There's also a big downward modifier if the other plane is firing on the plane attempting to sight it. (Simulating the fact that a human can easily extrapolate a firing plane's position from the angle of the tracers going past the cockpit.)

Finally, there's a cumulative penalty to keep track of targets after the first, based on pilot skill. This represents "information overload" of SA, with more distant and faster-moving targets being lost first.

A simple but arbitrary formula for maximum number of objects a pilot can track at once might be (pilot skill level)^2. With Rookie pilots being assigned a skill level of 1 and Ace pilots being assigned a skill of 4. (The very high number of objects an Ace can potentially track represents the fact that ace pilots tended to excel at the sort of spatial reasoning tasks represented by SA.)

If AI loses SA and is attempting to disengage, Rookie pilots will fly straight and level at top speed. (They assume that they're safe, even if they're not.) Average or better pilots will dive or climb at full power, as appropriate, while gently maneuvering to establish visual confirmation that they're not being pursued.

All pilots will use clouds and terrain as cover when attempting to disengage.

All pilots will fly towards friendly flak and fighter formations as a method of discouraging pursuit.


If AI loses SA and is attempting to engage, it will attempt to reestablish visual contact by maneuvering.

Rookie pilots will maneuver to establish line of sight to the target's last known position. Average pilots will attempt to regain line of sight by turning towards the target's last known line of travel.

Veteran or better pilots maintain some degree of SA with respect to lost targets and will usually turn towards the quarter of the sky which the target currently occupies. The exact percentage depends on pilot skill, say 60% for Veteran, 90% for Ace. Otherwise, they behave like Average pilots and turn towards the target's last known line of travel. (This represents a gross simplification of the mental calculations that a human would make regarding the opponent's energy state, speed, direction of travel, maneuverability, etc. Its not intended to make AI omniscient, just to give them a better than even chance of being able to reestablish Tally.)

Against fighters it should not depend on skill level -only on agressiveness.

As I imagined it, Aggression doesn't mean stupidity or bad tactics, it just means willingness to do battle. In some cases, it also means a willingness to force the opponent to play your game rather than accepting his.

While any pilot will take front quarter shots if they come his way, like you said, most pilots won't go out of their way to set them up.

If an enemy sets up a head-on pass where both planes can shoot each other up, even an aggressive pilot might not choose to reciprocate if it puts his plane at a disadvantage.

In such cases, the "aggressive" maneuver is to briefly go defensive, rolling or diving out of your opponent's line of fire and using your opponent's commitment to the attack to gain an advantage so you can attack him from a more favorable angle.

For the Japanese, it really depends on the plane. I could imagine a skilled and aggressive A6M2 pilot taking a head-on long-range shot against an allied fighter - particularly one with an inline engine - before rolling or diving out of the line of fire. The Zero driver is betting that his superior gunnery and heavy weight of his cannon fire will let him get in a fight-ending shot before his opponent can bring his guns to bear.

But, I can't imagine a sensible Ki-43 pilot doing the same thing. Unless he's got laser-like gunnery skills, he's got little to gain and everything to lose.


I do not accept head on exchanges against AI fighters - except

That's the sort of "accept/refuse the head-on attack" decision tree I'm trying to outline as suggestions to improve AI programming.

Factors that weight AI decision towards "accept" (in order of importance): Presence of frontal armor/armor glass, superior engine durability (i.e., radial vs. inline, includes current engine damage), inferiority maneuverability vs. opponent, weight of forward firepower, self-sealing fuel tanks, superior airframe durability, inferior speed vs. opponent, inferior climb vs. opponent, fighting over friendly territory, fighting over land.

Factors that weight AI decision towards "refuse" (in order of importance): Lack of frontal armor, inferior engine durability, superior maneuverability vs. opponent, inferior forward firepower, lack of self-sealing fuel tanks, inferior airframe durability (includes damage), superior speed vs. opponent, superior climb over opponent, fighting over hostile territory, fighting over water.

[Furio]

I don’t know if Daidalos Team guys read all of our discussion. Sometimes I hope they do, because they would read interesting things to ponder about. Sometimes I hope they don’t, because they have limited time and surely more useful things to do.:confused:

However, in this case I have a simple question that – perhaps – could have a simple reply. Have they ever tried to limit AI field of vision? If yes, have they obtained any meaningful result?

[Pursuivant]

However, in this case I have a simple question that – perhaps – could have a simple reply. Have they ever tried to limit AI field of vision? If yes, have they obtained any meaningful result?

Were you around when DT released the 4.10 patch? Huge improvements in limiting AI visibility:

www.youtube.com/watch?v=nGKtVW1gL9c

www.youtube.com/watch?v=6sx4W9z9MXQ

So, yes, it is possible to limit AI visibility in the game, and most of the work is already done. It was a massive amount of work, and one of DT's triumphs for which I am eternally grateful.

The only problems we've got left with regard to AI visibility is that they still seem to be a bit too good at detecting distant aircraft, correctly ranging distance, and keeping track of aircraft when they lose sight of them.

We might complain about current AI abilities, but DT has actually made MASSIVE improvement compared to what AI used to be like.

[idefix44]

Were you around when DT released the 4.10 patch? Huge improvements in limiting AI visibility:

www.youtube.com/watch?v=nGKtVW1gL9c

www.youtube.com/watch?v=6sx4W9z9MXQ

So, yes, it is possible to limit AI visibility in the game, and most of the work is already done. It was a massive amount of work, and one of DT's triumphs for which I am eternally grateful.

The only problems we've got left with regard to AI visibility is that they still seem to be a bit too good at detecting distant aircraft, correctly ranging distance, and keeping track of aircraft when they lose sight of them.

We might complain about current AI abilities, but DT has actually made MASSIVE improvement compared to what AI used to be like.

I agree with Pursuivant!

[Furio]

Were you around when DT released the 4.10 patch? Huge improvements in limiting AI visibility
Yes, I was, but missed these clips and never noticed much difference between dogfights taking place in clear or cloudy skies.

and keeping track of aircraft when they lose sight of them.
And that’s perhaps the problem. They keep track, and they shouldn’t. They should disappear from the battle scene. Clearly it's easier said than done. I’m aware of the difficulties faced by TD, and I’m no less grateful to them than you are.

[Pursuivant]

And that’s perhaps the problem. They keep track, and they shouldn’t. They should disappear from the battle scene. Clearly it's easier said than done. I’m aware of the difficulties faced by TD, and I’m no less grateful to them than you are.

I think it's fairly "easy" as these things go, although it would require lots of coding to add the necessary "decision trees."

4.12 finally got AI "target acquisition" more or less right, and it seems that Rookie or Average pilots can be pretty clueless if you get into their blind spots. What doesn't seem to be modeled is loss of Situational Awareness (SA) due to target maneuvering, information overload, and other factors.

A "simple" way to make loss of SA possible would be this:

Baseline ability to retain SA is based on pilot quality.

When an aircraft flies into an AI plane's blind spot, there's an x% chance every second that AI will lose SA with respect to that plane (lose "Tally" - visual confirmation of a plane's position & type), based on the target's range and speed. (Elements like target size, camouflage, visibility, etc. don't apply because this algorithm attempts to model the human ability to mentally track targets you can't see.)

This base chance is modified upwards if the spotting plane is damaged, under attack, or pulling Gs, or if the pilot (or some of the crew) are wounded. If the other plane is pulls Gs, or gains or loses altitude while in a blind spot, there's also an upward modifier. (This is also a simplification which represents that its harder to keep track of a fast or wildly maneuvering target.)

If the other plane is smoking, leaking fuel or coolant, or contrailing, there's a big downward modifier. (This represents a pilot's ability to track a plane, and follow its path, by using its smoke or vapor trail.)

Big downward modifier if the "lost" planes are flying in formation, as long as the viewer knows the position of at least one plane in the formation. (This simulates the fact that a human can easily determine the relative location of a hidden plane's position in formation as long as he can see at least one plane in the group.)

There's also a big downward modifier if the other plane is firing on the plane attempting to sight it. (Simulating the fact that a human can easily extrapolate a firing plane's position from the angle of the tracers going past the cockpit.)

Finally, there's a cumulative penalty to keep track of targets after the first, based on pilot skill. This represents "information overload" of SA, with more distant and faster-moving targets being lost first.

A simple but arbitrary formula for maximum number of objects a pilot can track at once might be (pilot skill level)^2. With Rookie pilots being assigned a skill level of 1 and Ace pilots being assigned a skill of 4. (The very high number of objects an Ace can potentially track represents the fact that ace pilots tended to excel at the sort of spatial reasoning tasks represented by SA.)

If AI loses SA and is attempting to disengage, Rookie pilots might fly straight and level at top speed. (They assume that they're safe, even if they're not.) Otherwise, Rookie and Average or better pilots will dive or climb at full power, as appropriate, while gently maneuvering to reestablish Tally & confirm that they're not being pursued.

All pilots will use clouds and terrain as cover when attempting to disengage.

All pilots will fly towards friendly flak and fighter formations as a method of discouraging pursuit.


If AI loses SA and is attempting to engage, it will attempt to reestablish visual contact by maneuvering.

Rookie pilots might maneuver to establish line of sight to the target's last known position (i.e., they will forget that their target isn't likely to be where they last saw it). Otherwise, Rookie and Average pilots will attempt to regain line of sight by turning towards the target's last known line of travel.

Veteran or better pilots maintain some degree of SA with respect to lost targets and will usually turn towards the quarter of the sky which the target currently occupies. The exact percentage depends on pilot skill, say 60% for Veteran, 90% for Ace. Otherwise, they behave like Average pilots and turn towards the target's last known line of travel. (This represents a gross simplification of the mental calculations that a human would make regarding the opponent's energy state, speed, direction of travel, maneuverability, etc. It's not intended to make AI omniscient, just to give them a better than even chance of being able to reestablish Tally.)

If you wanted to get even fancier, you could give AI (and the player!) the ability to ask where other planes are if they lose Tally. Again, not hard to implement, but time consuming, not just due to additional coding, but also due to new commands and voice programming needed.

[major.kudo]

Everyone, thank you for your very good opinion.
I interpreted as follows.


about gun shooting

- Gun shooting of AI more inaccurately. In particular, Rookie and Average.

- The deflection shooting is made more inaccurate.

- head on between the fighters is made more inaccurate.


Behavior

- AI Decision to "return to base immediately" by more less damage.

- The enemy who took Serious damage doesn't attack any more.

- AI loses sight of an enemy more easily.

- AI stop to attack a bomber of heavy armament from just behind.

- If the prudence of AI can be set at QMB and FMB, it's more better.


As far as it's possible, I make all opinion simple.　
And I'm thinking I'll show that to DT in the future.
It's necessary to be change as simple as possible for it.
When being complicated, that wouldn't be achieved.

[majorfailure]

Everyone, thank you for your very good opinion.
I interpreted as follows.


about gun shooting

- Gun shooting of AI more inaccurately. In particular, Rookie and Average.
I still think they shoot too inaccurate - and they correct their aim abysmally.

- The deflection shooting is made more inaccurate.

- head on between the fighters is made more inaccurate.

Yes on both counts - and also the willingness to risk serious damage by offering head-ons is to great IMHO.
Behavior

- AI Decision to "return to base immediately" by more less damage.

- The enemy who took Serious damage doesn't attack any more.

- AI loses sight of an enemy more easily.

- AI stop to attack a bomber of heavy armament from just behind.

- If the prudence of AI can be set at QMB and FMB, it's more better.


And add to that that the better AI should at least when in a favourable plane/position try to disengage when a fight goes south once in awhile.

[major.kudo]

Dear Pursuivant.

I think opinions of Pursuivant are very deep and important good opinions.
But I can't simplify those in my poor English.
I tried several times.　But I think some important points are missing.
So I want you to describe only important part using a sentence of 300 characters degree.
Excuse me, please.

[Pursuivant]

But I can't simplify those in my poor English.

No need to apologize. Your English is far better than my ability to write in your native language!

My post was many ideas for how to improve artificial intelligence programming, by making the computer simulate "awareness of surroundings" that humans have.

The big idea was that the computer should have some ability to know where planes will be if an AI pilot can't "see" them, but not perfect knowledge.

* Rookie pilots don't have as much knowledge as aces.

* It is much harder to lose knowledge of where a particular airplane is if there are many airplanes in the area.

[Igo kyu]

* It is much harder to lose knowledge of where a particular airplane is if there are many airplanes in the area.
Eh, I'm having trouble with that one.

It seems to me that in real life, it would be more difficult to keep track of a particular aircraft if there were more planes in the air, and that is what the AI ought to be emulating.

[Ice_Eagle]

Regarding "unrealistic" head-on attacks:

Rule 6. If your opponent dives on you, do not try to get around his attack, but fly to meet it. The instinctive reaction of many rookies was to turn and flee from an approaching attacker—especially a diving one. This simply presented their tail to the attacker, usually with disastrous results. Boelcke taught that a pilot had to conquer that instinct. Turning to face the attack could force the attacker onto the defensive, or at least keep the situation unsettled, which was far better than presenting your tail. Even though climbing to meet an attack would reduce speed, it was better to try to bring one's own guns to bear than to flee, and approaching the enemy still increases the relative velocity between the two fighters and thus reduces the time during which the enemy can fire. Furthermore, if both fighters miss, the diving attacker must now pull out of his dive, while the defender is now in position to circle around and counter-attack with his own dive.

Still taught to this very day.

Luftwaffe pilots learned early on with the introduction of the -17's in the ETO that head-on attacks were far more effective then attacking from any other position, even though more dangerous. Less time spent in the bomber gunners sight.

The AVG, despite the wonderful BnZ tactics, often turned to meet an enemy head-on rather then diving away. In A2A combat, you don't think, you react. You think, your dead. Thats why these rules are taught over and over so pilots don't have to think.

Overall the 4.12 AI is pretty good. especially when Ki43's & Zero's snaproll and stall, you go sailing right by them, they recover and now there on your 6 :grin:

[Pursuivant]

It seems to me that in real life, it would be more difficult to keep track of a particular aircraft if there were more planes in the air, and that is what the AI ought to be emulating.

That's what I meant, sorry I didn't say it more clearly.

Ace pilots should be able to track lots of planes even if they can't see them. Rookies can lose track of even one.

[major.kudo]

Last suggestion decided as a result of the argument.
The following improvements are needed to make more realistic Kill ratio between AI.



About gun shooting

- The deflection shooting is made more and more inaccurate. In particular, Rookie and Average.

- head on between the fighters is made more inaccurate.



About behavior

- AI Decision to "return to base immediately" by more less damage.

- The enemy who took Serious damage doesn't attack any more.

- If the prudence of AI can be set at QMB and FMB, it's more better.

- AI loses sight of an enemy more easily.
Rookie pilots don't have as much knowledge as aces.
It is much harder to lose knowledge of where a particular airplane is if there are many airplanes in the area.
Ace pilots should be able to track lots of planes even if they can't see them. Rookies can lose track of even one.

- AI fighters does not attack a bomber of heavy armament from just behind.





And, if it's possible more!

- http://gachopin.no-ip.info/kudo/Bombers_formation.jpg

[Ice_Eagle]

Hello,
I seen AI attack heavy bombers from all directions, including head-on
attacks in v4.12, v4.13. I do notice that bomber do not hold formation
on the ground attack waypoint, when they should. Very annoying :x

[Avimimus]

The major problem is the failure of the AI to disengage. If you look at historical accounts, pilots would return to base as soon as they got separated from their allies. In many cases it was easy to lose sight of both allies and enemies.

There is actually already some code that could be repurposed for this: When you call for help the AI automatically checks to see if it is in range (8000m I think) before coming to your aid.

So it should be possible to program in a check that causes aircraft to be have a certain probability of returning to base if they get even somewhat separated from allied or enemy aircraft.

P.S.
There are also checks for how much ammunition is left during an attack. So one could add a random chance of aircraft disengaging after an attack.

Of course, programming AI aircraft to retreat (especially retreating when outnumbered) could be frustrating to some players.

[baball]

Everyone, thank you for your very good opinion.
I interpreted as follows.


about gun shooting

- Gun shooting of AI more inaccurately. In particular, Rookie and Average.

- The deflection shooting is made more inaccurate.

- head on between the fighters is made more inaccurate.


Behavior

- AI Decision to "return to base immediately" by more less damage.

- The enemy who took Serious damage doesn't attack any more.

- AI loses sight of an enemy more easily.

- AI stop to attack a bomber of heavy armament from just behind.

- If the prudence of AI can be set at QMB and FMB, it's more better.


As far as it's possible, I make all opinion simple.　
And I'm thinking I'll show that to DT in the future.
It's necessary to be change as simple as possible for it.
When being complicated, that wouldn't be achieved.

Another behavior I'd be really glad to see added to the game would be the ability for pilots of heavily damaged aircrafts to bail out over friendly territory instead of trying to crash land over some forest. Goddam tree huggers.;-)

[Pursuivant]

The major problem is the failure of the AI to disengage. If you look at historical accounts, pilots would return to base as soon as they got separated from their allies. In many cases it was easy to lose sight of both allies and enemies.

This depends on the mission. In many cases, there would be designated "rally points" where lost pilots could go an wait for their friends to meet them. Usually, these were big landmarks where flak concentrations were low, like a town, mountain, or major road junction.

In any case, a lost pilot, particularly a fighter pilot, is going to form up with any allied plane he encounters which is going towards home, particularly bombers.

Bombers which had to drop out of formation and/or RTB were always happy to have their own personal fighter escort, and fighter pilots were happy to have their own personal bomber to provide navigation, defensive gunnery, and look-out services.

Of course, programming AI aircraft to retreat (especially retreating when outnumbered) could be frustrating to some players.

One way around this is to have a player-controlled option for "realistic aggressiveness". Set it for Yes, and AI aircraft act like the people inside them only get to die once. Set it for No, and AI aircraft behave as they do now.

[sniperton]

One way around this is to have a player-controlled option for "realistic aggressiveness". Set it for Yes, and AI aircraft act like the people inside them only get to die once. Set it for No, and AI aircraft behave as they do now.

I'm eager to know how this 'fear of death' or 'fear of self-destruction' could be programmed. Perhaps the closest analogy is the risk-evaluating algorithm of self-driven cars where the safety of the passenger is first priority. That would give us a 'coward' AI which could be fine-tuned to take more and more risk toward the 'ace' level. But constantly running such an algorithm for multiple AI planes would be resource-heavy, I guess.

[Pursuivant]

I'm eager to know how this 'fear of death' or 'fear of self-destruction' could be programmed. Perhaps the closest analogy is the risk-evaluating algorithm of self-driven cars where the safety of the passenger is first priority.

That seems like a considerably more complex algorithm than I was thinking of.

A simpler way might be for AI to count up all the enemy planes in sight, compare them to the number of friendly planes in sight, multiply or divide by factors such as favorable/unfavorable position (altitude advantage/ disadvantage, advantageous/ disadvantageous attack position as defined by QMB), relative quality of planes involved, damage to plane and crew, and assigned mission.

If the resulting number is above a certain threshold, AI aircraft will attack. Otherwise, they will avoid contact or disengage.

Repeat this algorithm after every attack.

I think that the AI already does something like this.

Average or better fighters will always take a head-on shot against a level bomber if closing from the front. Otherwise, they make high side attacks or high or level beam or flank attacks, depending on relative position (although if you want to get picky, USN pilots were trained to make high side and high beam attacks, not head-on attacks). Only rookies will attack level bombers from the rear.

The exception is that flying boats/float planes will always be attacked from below - from front, beam, or flank. Attack aircraft (including strike fighters and attack bombers) will be attacked from any quarter other than head-on.

I think that one of the problems that IL2 AI suffers from, which might not be correctable, is that it only defines three classes of aircraft - level bombers, attack, and fighters.

It would have been better if every plane had been given a rating from 1-10 for speed, maneuverability, offensive firepower, defensive firepower, armor, and ruggedness, plus yes/no operators for various types of ordinance, commo, and radar (e.g., torpedoes, bombs, mines, rockets, heavy cannon, radio receiver, radio transmitter, tail warning radar), and modifiers for assigned role (interceptor, air superiority fighter, bomber destroyer, level bomber, level attack bomber, anti-shipping, ground assault, dive bomber, torpedo bomber, ASW, ASR, artillery spotting/FAC, low level PR, high level PR, transport, cargo drop, paratroop/agent drop, agent insertion/extraction)

Ratings like these would allow the potential for more sophisticated and flexible AI.

For example, an interceptor fighter with a serious edge in speed (a rating of 10 vs. 6), but inferior maneuverability (a rating of 3 vs. 5), is always going to try to BnZ.

Or, a plane with a high level of ruggedness and armor, plus high offensive firepower, is more likely to take a head-on shot against an opponent with weak offensive firepower.

You could also use these ratings to abstractly determine the outcome of AI vs. AI fights which take place far out of sight of the player, or missions flown just by AI during a campaign.

[sniperton]

I think that one of the problems that IL2 AI suffers from, which might not be correctable, is that it only defines three classes of aircraft - level bombers, attack, and fighters.


If my memory serves, Il2 has the following basic classes:

Transport
Scout
Seaplane
Ground Attack ('Stormovik')
Level Bomber
Dive Bomber
Fighter
TNB Fighter
BNZ Fighter

Each plane is defined as belonging to one or more classes, e.g. the 190 is a 'Fighter', a 'BNZ Fighter', and a 'Stormovik'. New classes could be defined and added, I'm pretty sure, but all this would only affect tactics, not aggressivity.

I think this is the way to go, and this is basically what I had in mind:

A simpler way might be for AI to count up all the enemy planes in sight, compare them to the number of friendly planes in sight, multiply or divide by factors such as favorable/unfavorable position (altitude advantage/ disadvantage, advantageous/ disadvantageous attack position as defined by QMB), relative quality of planes involved, damage to plane and crew, and assigned mission.

If the resulting number is above a certain threshold, AI aircraft will attack. Otherwise, they will avoid contact or disengage.

I would disregard 'relative quality' (think of the pathetic attacks of the Polish and the Finns), but would add
- remaining ammo
- remaining fuel (distance to base)
- territory (friendly/hostile) and, yes,
- self-confidence (aka skill).

Once you calculated the success/risk ratio (which shall be higher for an ace pilot), you can define a kamikaze to attack even with zero chance, a braveheart with 50% chance, and a cautious pilot only with 75% (or higher) chance.

As a result, a cautious British pilot over Kent will be more aggressive than his braveheart German adversary. The crucial point is, I think, to define the conditions when a plane has to disengage and RTB.

[Verdun1916]

My biggest issue with the AI at the moment is the AI's reluctance to strafe groundtargets unless they have dropped bombs or fired rockets first. When flying fighters or fighter-bombers against a ground target, say a column of vehicles or a train, and the only armament is your machineguns and/or cannons the AI doesn't seem to want to play ball. You order your comrades in your flight to attack the vehicles or the train and they answer "Roger" or something similar yet they don't actually attack. They only fly low over the target instead of strafing it. And if there are any kind of AA-protection they usually get themselfs shot down in the process. And this is very annoying!

I know others have mentioned this before but I want to lift this issue again. I would love the AI to be adjusted, if possible of course, so they will strafe ground targets when ordered to no matter if they have been armed with bombs or rockets earlier in the mission or not.

My second biggest issue is as so many others the far to accurate gunners on multicrew aircraft. They are freaking snipers at extreme distances!!! (They even put Simo Häyhä to shame, and he was the best sniper in the history of war! 500 - 742 kills within 105 days during the Winter War) It would be lovely if this "sniper skill" could be dialed down a bit.

[dimlee]

As a result, a cautious British pilot over Kent will be more aggressive than his braveheart German adversary.


And both will be more aggressive at fight over the landmass then over the water...
Good ideas, I agree with everything.

[dimlee]

My second biggest issue is as so many others the far to accurate gunners on multicrew aircraft. They are freaking snipers at extreme distances!!! (They even put Simo Häyhä to shame, and he was the best sniper in the history of war! 500 - 742 kills within 105 days during the Winter War) It would be lovely if this "sniper skill" could be dialed down a bit.

I love to hunt bombers, offline and online. I learned to accept those uber snipers as another "feature" and I try to stay calm when another B-17 gunner kills the pilot of my interceptor during high head on attack (up to 1000 kmh relative speed). Try...stay...calm... :evil: :grin:
But what really annoys me is a lack of logic. ShVAK gunner freezing in open cockpit of MBR-2 (TB-3, etc.) looking through fogged glasses and aiming with primitive iron gunsight can be more precise than ball turret operator of B-17/B-24 or computer operator of B-29.

[sniperton]

I made some comparative tests and my impression is that 'rookie' gunners in IL2 are better than 'average' gunners in CLOD. One must also take into consideration that damage boxes in IL2 are probably bigger and coarser and this way 'fatal' hits are more common.

Taking down a Blenheim or a Heinkel takes more hits in CLOD, for sure, but it's compensated by smoother controls and better aiming accuracy. A 40% hit ratio is something I could never achieve in IL2. My aiming is either shaky or I don't have the needed control authority in dogfight. The same applies to throttle and trimming. I always feel I applied too much or too less, while in CLOD I find them just OK. This is not only the different AI, but also the different flight physics that counts, I think, and CLOD with its thicker air feels more natural. I wish if IL2 could be made alike.

[dimlee]

I made some comparative tests and my impression is that 'rookie' gunners in IL2 are better than 'average' gunners in CLOD. One must also take into consideration that damage boxes in IL2 are probably bigger and coarser and this way 'fatal' hits are more common.

Taking down a Blenheim or a Heinkel takes more hits in CLOD, for sure, but it's compensated by smoother controls and better aiming accuracy. A 40% hit ratio is something I could never achieve in IL2. My aiming is either shaky or I don't have the needed control authority in dogfight. The same applies to throttle and trimming. I always feel I applied too much or too less, while in CLOD I find them just OK. This is not only the different AI, but also the different flight physics that counts, I think, and CLOD with its thicker air feels more natural. I wish if IL2 could be made alike.

Interesting observations, thanks. But - taking physics model aside - can we really claim what is more "realistic"? Smoother controls of one airsim or shaky ones of another?
I have no experience in CLOD. But then I wonder how many players of CLOD and IL-2 have experience of real life dogfight.

[sniperton]

But - taking physics model aside - can we really claim what is more "realistic"? Smoother controls of one airsim or shaky ones of another?
I have no experience in CLOD. But then I wonder how many players of CLOD and IL-2 have experience of real life dogfight.

I wrote 'natural', not 'realistic'. This is all about the feel, the illusion of reality. In IL2 I feel the need for a long stick to fly these planes correctly, while CLOD seems to be better optimized for short sticks. Another point at issue is the fine balance between throttle/thrust and air drag. A weak engine with little drag (IL2) give a different feel than a strong engine with strong air drag (CLOD). Strong drag also contributes to a better longitudinal stability and to less wobbling when aiming.

A related issue is torque. When speed is low and the stabilizers don't get enough airflow (typically on the ground, when taxiing or during takeoff), the torque effect is very strong in CLOD (and BoS); but once you get airborne and your speed increases, the stabilizers become more and more effective and minor throttle changes do not result in significant and sudden torque changes as they do in IL2. When you reduce throttle in CLOD (or BoS), the nose drops a bit, but there's no urgent need for applying rudder and aileron to re-adjust your aim, which are responsible for most of the wobbling experienced in IL2.

[Daniël]

Interesting discussion about the feel of flying. I only have 30 minutes of flight in a piston powered airplane, so I guess I can't really comment on the effects of throttle and torque, but I have around 200 flights in gliders now (around 30 hours). In my opinion the flight model in CloD resembles the feel of flight better than Il-2 1946. It's difficult to explain, but I think that Il-2 1946 is not as smooth as CloD (and how a glider feels).

In my opinion by far the biggest difference in the flight models of Il-2 1946 and CloD is the handling on the ground. CloD is a lot more sensitive in that aspect and I think it feels better than Il-2 1946.

I like the flight models of DCS and Falcon BMS too. DCS depends on the module of course and I can only give my opinion about the stock planes, the Su-25 and TF-51 as I don't have any modules for DCS yet, but out of those two I think the SU-25 feels the most natural.

Regarding BoS: The planes in BoS are very sensitive (nothing wrong with that), but I think that they're far too wobbly and I feel that the stall characteristics are not realistic compared to what I have experienced in gliders. In a spin in a glider a wing drops and the plane kind of "flips over". However in BoS the "flipping over" action stops way too early I think. I think that the other simulators that I named do a better job at simulating a spin, including Il-2 1946.

Please bear in mind that I have not flown any of the planes in these simulators in real life so my opinion is probably subjective and influenced from flying gliders and I have not played Il-2 1946 in couple of months, so maybe I don't remember all characteristics exactly. I mainly fly CloD and Falcon BMS at the moment.

[dimlee]

Interesting discussion about the feel of flying. I only have 30 minutes of flight in a piston powered airplane, so I guess I can't really comment on the effects of throttle and torque, but I have around 200 flights in gliders now (around 30 hours). In my opinion the flight model in CloD resembles the feel of flight better than Il-2 1946. It's difficult to explain, but I think that Il-2 1946 is not as smooth as CloD (and how a glider feels).

There are no gliders (besides AI) in IL-2 unfortunately. But if you attempt unpowered flight, does it feel natural for you? Probably Me 163 is good for such experiment.

[sniperton]

As to the sensitivity issue: what controller do you have? Mine is a mediocre T Flight Hotas X, where I replaced the pots with Hall sensors. I positively know that the sensors produce a linear output between 1V and 4V, so that the output should be all linear between this range. Still, Thrustmaster’s firmware enforces a broad deadzone around 50%, so that it’s all the same what stick deflection you have, you get the same values from -9% stick deflection (relative to centre) to +9%. So what happens when you try to adjust your aim? You move your stick from -9% to +9% hastily, and consequentially you enter the registered zone with a value which is well beyond what is expected there.

[taly001]

Mine is a mediocre T Flight Hotas X,........... Thrustmaster’s firmware enforces a broad deadzone it’s all the same from -9% stick deflection (relative to centre) to +9%

I bought a T.Flight Hotas X but I gave up on it after a week due to the ~9% deadzone, I sold it and bought a T.Flight Hotas 4, on the recommendations of a hardware tester on the internet, Why Thrustmaster does not advertise the VAST superiority of the Hotas 4 over the Hotas X for flight sims is beyond me.

The T.Flight Hotas 4 has a tiny deadzone and I'm perfectly happy with it. The Hotas 4 even has a dedicated port to directly connect the T.Flight Rudder Pedals, and updateable firmware!

[sniperton]

Then you perhaps can tell me whether the Hotas 4 also has a center deadzone on the Throttle. Because, most absurdly, the X has. Yes, in the center of the throttle! :rolleyes:

[Daniël]

I've done some tests with the Me-163 and I have to say that I'm actually impressed with the feel of Il-2 1946. I haven't flown this simulator for months, but I think I might fly it a bit more now :grin:

I have done some testing with two sensitivity profiles. The first one all sensitivity bars to 100. With that setting I felt that the sensitivity was too "jumpy" in the centre. The second sensitivity setting was 0 5 10 20 30 40 50 60 80 100 and it felt about right. The joystick that I use is an old Saitek Cyborg 3D Rumble Force and I have Saitek Combat Rudder Pedals.

The stall in a Me-163 feels really similar to what I have experienced in gliders. The buffeting seems realistic and it increases when the speed drops and you keep pulling on the stick.

I felt that the spin characteristics of the Me-163 in Il-2 1946 are very forgiving, but it's possible to force a spin by pulling the stick back and to one side and with the pedals in the same direction. The spin recovery felt natural. I don't think that I can comment on the spin characteristics of the Me-163 in particular bacause I haven't flown the real thing ;) , but it might be a bit too forgiving because the stall should start at the tip because of the backward swept wing. I've tried to find data on the airfoil of the Me-163 to find out if the wings were twisted to prevent the tip from stalling first, but I couldn't find data about that.

I still think that CloD for instance has a better feeling of flight, but it's quite close.

[sniperton]

I still think that CloD for instance has a better feeling of flight, but it's quite close.

I agree, but this may have something to do with the visually and acoustically richer environment. At the same time I feel CLOD is the most forgiving as to how you fly, I rarely need to practice my spin recovery skills there, while in BoS it happens to me quite frequently (the LaGG is really nasty in this respect). IL2 is somewhere in between.
I'm relatively new to CLOD and BoS, and what struck me was that a perfect three-point landing is much easier than I experienced in IL2. Although in BoS I can overshoot the runway, in CLOD I don't even have to watch my speed. Planes decelerate very well in CLOD, perhaps too well, I don't know.

[Pursuivant]

If my memory serves, Il2 has the following basic classes:

That's more classes than I recall, but is a good selection. The only really necessary class additions are torpedo bomber, anti-shipping, parachute/agent.

Possibly there could be another flag as to whether an airplane is fully aerobatic, to keep planes like the A-20 from looping and rolling like a fighter.


I would disregard 'relative quality' (think of the pathetic attacks of the Polish and the Finns),

This is a really good point, but could be compensated for by relative pilot quality (both Finns and Poles were good pilots, despite their planes) and "desperation factor" (defending the motherland from invasion). Aggressiveness could also be pumped up for Japanese, and reduced for "less motivated" nationalities like the Chinese.

There should also be player and/or mission builder defined aggressiveness for a particular mission - for both friendly and hostile planes of various squadrons. Every nation had their bad days and their moments of tremendous heroism.


- remaining ammo
- remaining fuel (distance to base)
- territory (friendly/hostile) and, yes,
- self-confidence (aka skill).

More good points.

Additionally, a big factor was risk of capture/death in case of a bailout. For example, American pilots attacking Japan in 1945 could be a bit more aggressive since they knew that there were rescue submarines just off shore. But, for early war missions, both Japanese and Americans had to be a bit careful since a water landing was very likely to be a death sentence.


Once you calculated the success/risk ratio (which shall be higher for an ace pilot), you can define a kamikaze to attack even with zero chance, a braveheart with 50% chance, and a cautious pilot only with 75% (or higher) chance.

Exactly what I was thinking. You could even customize values for various pilots, making pilots in a campaign more or less courageous, ranging from foolhardy to cowardly.

[Pursuivant]

My biggest issue with the AI at the moment is the AI's reluctance to strafe groundtargets unless they have dropped bombs or fired rockets first.

+1!

There should be an command to "attack ground targets with guns" and by preference well-armed fighters and ground attack planes should attack "soft" vehicles with guns or rockets rather than bombs - save the bombs for better targets like AFV, bridges, or trains.

There should also be some sort of option to "strafe along this axis" so that you can make your wingmen strafe down the length of a convoy, the deck of a ship, or along a line of parked aircraft.


My second biggest issue is as so many others the far to accurate gunners on multicrew aircraft. They are freaking snipers at extreme distances!!!

I think that this has gotten a bit better with each new release. In particular, the U-2 (which used to be a flying flak tower) has been fixed in the last release.

Even so, long range flexible gun accuracy is still too good for some planes and some gun positions. I think that slipstream buffeting of guns, vibration, and turbulence aren't factored into gun accuracy algorithms.

Additionally, AI gunners can instantly detect and react to an airplane flying into their view, which makes fire from gun positions with a narrow field of view too effective (like the dorsal guns on the Ju-88, or some ventral gun positions). Realistically, it should take a gunner some fraction of a second to identify and track a hostile plane before opening fire once it comes into view.

[sniperton]

Transport
Scout
Seaplane
Ground Attack ('Stormovik')
Level Bomber
Dive Bomber
Fighter
TNB Fighter
BNZ Fighter

These are the basic IL2 aircraft classes, period. As taken from in-game class files. Later on some others were added, e.g. for planes carrying anti-ship missiles or having only a two-stage prop. But these are the base classes which are defined internally and on which the AI may rely. No more, no less. That's what the AI has.

[dimlee]

I bought a T.Flight Hotas X but I gave up on it after a week due to the ~9% deadzone, I sold it and bought a T.Flight Hotas 4, on the recommendations of a hardware tester on the internet, Why Thrustmaster does not advertise the VAST superiority of the Hotas 4 over the Hotas X for flight sims is beyond me.

The T.Flight Hotas 4 has a tiny deadzone and I'm perfectly happy with it. The Hotas 4 even has a dedicated port to directly connect the T.Flight Rudder Pedals, and updateable firmware!

Thanks for mentioning Hotas 4. I used Thrustmaster Hotas sticks/throttles for many years but stopped to follow their products and was not aware about this new model. Probably it's time to upgrade from Hotas X.

[dimlee]

I've done some tests with the Me-163 and I have to say that I'm actually impressed with the feel of Il-2 1946. I haven't flown this simulator for months, but I think I might fly it a bit more now :grin:

I have done some testing with two sensitivity profiles. The first one all sensitivity bars to 100. With that setting I felt that the sensitivity was too "jumpy" in the centre. The second sensitivity setting was 0 5 10 20 30 40 50 60 80 100 and it felt about right. The joystick that I use is an old Saitek Cyborg 3D Rumble Force and I have Saitek Combat Rudder Pedals.

The stall in a Me-163 feels really similar to what I have experienced in gliders. The buffeting seems realistic and it increases when the speed drops and you keep pulling on the stick.

I felt that the spin characteristics of the Me-163 in Il-2 1946 are very forgiving, but it's possible to force a spin by pulling the stick back and to one side and with the pedals in the same direction. The spin recovery felt natural. I don't think that I can comment on the spin characteristics of the Me-163 in particular bacause I haven't flown the real thing ;) , but it might be a bit too forgiving because the stall should start at the tip because of the backward swept wing. I've tried to find data on the airfoil of the Me-163 to find out if the wings were twisted to prevent the tip from stalling first, but I couldn't find data about that.

I still think that CloD for instance has a better feeling of flight, but it's quite close.

Thank you. I love Me 163 and it's good to know that her gliding in IL-2 is "approved" by real pilot. :)

Airfoil - this one?
http://airfoiltools.com/airfoil/details?airfoil=goe765-il

[Daniël]

I've done some testing with the Bf-109 F-4 and G-2, the Yak-1 and the LaGG-3 in BoS to compare the stall and spin characteristics and I was surprised that the Russian planes spinned more naturally (in my humble opinion). But for both sides I think that the low speed characteristics are too forgiving. When you fly the landing approach way too slow it won't hurt you while from my experience in gliders you have to watch your speed carefully because you will fall hard when you try to flare when your speed is too low.

Dimlee, I found that page, but I'm not an aeronautical engineer ;) so I couldn't find out if the wings are twisted or not. Some terms I do understand but others I have never heard of.

[major.kudo]

I think the game is similar to movie.
One of important thing of the movies is good actors.
Good-looking, number of people enough, and good at acting.
I think these can be applied to computer game.

IL-2 1946 is very old game in today.
So "good-looking" is inferior than the latest games graphic.
This is all right.
And "numbers of actors" is very rich.
There are some actors who would like to request, too.
However, This is also all right.

But "good at acting" is just insufficient.
I think this is a big problem.
Cool planes, abundant variations of weapons, map of the elaborate structure.
Actors at poor acting makes all other good parts bad.

I want movement of develop to pay attention to "AIs good acting" more than appearance of new face.

[Pursuivant]

I think the game is similar to movie.

Interesting comparison.

If you want to compare IL2 to a movie, I'd compare it to an old "epic" movie like "Battle of Britain" or "The Longest Day" - dated special effects, but a "cast of thousands," excellent performances from a number of actors, and a real attempt to get the history right.

Single-plane simulations are more like "character study" or "biographical" movies. One person's life - or actions during a particular period in life - studied in detail, with everyone else as a supporting actor.

Plane-themed video games like World of Warplanes or Warthunder are like those noisy summer movies which feature CGI, guns, and explosions as the lead actors, with humans in a supporting role. Great to look at, but bad writing, painfully bad performances by all the actors, and no historical accuracy.

[dimlee]

Major.Kudo and Pursuivant,
This is interesting comparison indeed and thought provoking one.
I like Pursuivant's words about WT. This is what I used to feel about Gajin's creations but could not name it right. Exactly ! - as to watch some heavily advertised blockbuster in IMAX... and to forget about its content on the next day.

[taly001]

I think the problem with AI in ALL computer games is that they can track the player once spotted with 3D X-ray vision. In real life when sight is lost the player needs to re-find the enemy each time, and then to re-analyse the maneouver the enemy is making. The AI doesn't seem to need this and can track and follow your moves all the time.

This may partly be why AI vs AI is such a kill-fest. They never loose track of each other so will fight to the death. Forcing AI to reacquire the target after loosing sight would require CPU cycles, drawing predicted target flight path from time sight lost and have AI look there first, then scan around to find before been allowed to manoeuver to attack.

[Storebror]

I think your idea bears some useful points to start thinking on.
Already now, IL-2 has the ability the check whether actually AI can see an enemy plane at all, taking own plane obstacles into account as well as buildings, mountains etc.
But you're right that AI will instantly reacquire the target once it comes in it's sight again.
The latter could be changed closer to reality by taking a few parameters into account, e.g.:

The larger the relative movement between own aircraft and target (when it moves out of sight), the longer you need to scan (because it will become hard to predict where the enemy will be).
The smaller the relative movement between own aircraft and target (when it comes back in sight), the longer you need to scan (because it's harder to spot non-moving targets).
The more G's are pulled by enemy aircraft, the longer it takes to scan (the future position will become hard to predict under heavy maneouvering).
The more G's AI pulls themselves, the longer it takes to scan.
The further the enemy distance is, the longer it takes to scan (dots becoming smaller).
If the target is below the horizon when it comes back in sight, add a scan penalty.


With these factors a scan time could be calculated where AI would not immediately track a target when it comes back in sight, but just after this scan time has elapsed.

Best regards - Mike

[Pursuivant]

I think the problem with AI in ALL computer games is that they can track the player once spotted with 3D X-ray vision. In real life when sight is lost the player needs to re-find the enemy each time, and then to re-analyse the maneouver the enemy is making. The AI doesn't seem to need this and can track and follow your moves all the time.

For Ace or Veteran AI it's not unreasonable to have them be able to keep track of several opposing aircraft even if they briefly lose sight of them. That's call "situational awareness" or "spatial memory" and it's a vital skill for good combat pilots.

But, for Average and Rookie AI, ability to maintain a "lock" on a target is perhaps too generous.

One thing that I've never seen for AI for any airplane sim is the possibility of "sensory overload," where if you try to track multiple targets simultaneously you lose track of almost all of them, allowing one or more to get onto your 6.

Forcing AI to reacquire the target after loosing sight would require CPU cycles, drawing predicted target flight path from time sight lost and have AI look there first, then scan around to find before been allowed to manoeuver to attack.

I'm not sure that it would require that many CPU cycles to determine if AI maintains SA with respect to a particular opponent. I think that the real demand on CPU cycles would be constantly drawing lines of sight between various crew stations and other planes to determine if a particular crewman has LoS.

A simplifying factor, which is realistic, is to treat distant formations of aircraft as a single group, and have AI check for line of sight to that group at less frequent intervals, or not at all.

Another way to reduce CPU cycles is to limit the maximum number of aircraft a given crewman can track by skill level (For example, 1 for Rookie, 2 for Average, 4 for Veteran, 16 for Ace). AI will focus on attacking enemy aircraft first, wingman second, other nearby enemy aircraft third, other flights (treated as a single unit) fourth, distant enemy aircraft fifth, and distant friendly aircraft sixth. If there's a risk of collision with the ground, barrage balloons, or similar, attacking enemy comes first, terrain comes second, then wingman, etc.

Finally, you could make the chance of losing a "lock" on an airplane you can't see a function of time that aircraft has spent outside of line of sight - assuming it isn't flying straight and level. Check more frequently for loss of line of sight for less skilled pilots.

[Pursuivant]

But you're right that AI will instantly reacquire the target once it comes in it's sight again.

That removes several important elements from the OODA Loop.

When you reacquire a target, your brain spends a fraction of a second on the Observe, Orient, and Decide portions of the loop before you Act. AI should do the same, if only by briefly hesitating before acting, and with a small percentage chance that AI won't automatically reacquire (because the pilot briefly had his head down in the cockpit, or his vision was briefly obscured by "Fog of War" - literal or otherwise.)

My experience playing the game is that Storebror's list of factors to take into account is very true once you lose Line of Sight (LoS), but sometimes makes maintaining a "lock" on an airplane you can see easier.

For example, a plane which is fast-moving, or committed to a high G maneuver is in some ways easier to predict because a fast moving plane is more or less committed to a particular speed and vector. Likewise, a plane pulling serious Gs has "no place to go" but to a lower G state. Same thing for a plane in a very poor energy state, or a plane with certain types of damage.

But, I'm not quite sure how to get AI to recognize all those things, even if it's possible.

[Storebror]

One thing we shouldn't forget is that AI is comparably "stupid" when it comes to draw conclusions based on the facts they know.
Human players can utilize their experience from previous sorties and can adopt promising options even in uncertain or unknown situations.
AI can't do any of this.
AI has a simple tree/branch decision scheme, sometimes leading to rather stupid moves.
In that regards, letting AI "cheat" in another regime to compensate this lack of experience might be a valid decision to some degree.

Best regards - Mike

[sniperton]

One thing we shouldn't forget is that AI is comparably "stupid" when it comes to draw conclusions based on the facts they know.

It would be intriguing to know what are the ‘facts’ they know.

1. There seems to be a 'visibility' range for contacts. Whether it’s general or depending on AI skill is unclear to me. It’s also unclear under what conditions combat AI kicks in and takes command over from the mission script (or gives the command back).
2. Once a contact is within this range, the AI has no problems to identify it as friend or foe. In this respect the AI always flies ‘with icons on’.
3. How many contacts the AI tracks simultaneously is unclear. It’s also unclear whether and how this depends on skill and/or other factors (apart from blocked LoS). Anyway, the data of all contacts are theoretically available to the AI, even if some of them are disregarded.
4. The AI has exact knowledge of contact position (vector and distance), perhaps with some inaccuracy margin to reflect skill.
5. The AI has a very good knowledge of target travel (direction and speed), probably with an inaccuracy margin to reflect skill. (This ‘knowledge’ is a prerequisite for gunnery, particularly for deflection shots.)
6. The AI seems to be aware of the general type of the target aircraft. There are different attack schemes against fighters and bombers. The differentiation is probably based, in some simplified way, on the types listed in post #46.
7. The AI is aware of the terrain (but not the objects!) it is flying over.

All this is just guessing, of course, and probably far not complete, but brainstorming could be made more effective if we knew the cruel 'facts' about the AI. :grin:

[majorfailure]

One thing we shouldn't forget is that AI is comparably "stupid" when it comes to draw conclusions based on the facts they know.
Human players can utilize their experience from previous sorties and can adopt promising options even in uncertain or unknown situations.
AI can't do any of this.
AI has a simple tree/branch decision scheme, sometimes leading to rather stupid moves.
In that regards, letting AI "cheat" in another regime to compensate this lack of experience might be a valid decision to some degree.

Best regards - Mike

Experience is not needed to have an option to disengage - besides the obvious out of ammo/fuel low option. A script can decide if numerical superiority has gone, a script can decide if positional advantage has gone, and a script can track losses, a script can count friendly planes in sight and so on.
Experience is neither needed to obey your commander. Which current AI does not and most times not at all. I have the experience, and if I tell them we do the mission my way they should not question it -if all goes south, I will have to face the music, not them.

Experience is not needed to shoot semi-accurately at nearly non moving targets, at least if you have been through flight school (possibly with the exception where pilots were needed fast and anyone half capable of making a landing stick was allowed to fly(e. g. Germany end 44).

And you don't need experience to set a limit for maximum contacts to be able to track. Script can do that. Trow in some parameters, as distance to enemy, current LoS, movement across or away/towards, time in LoS, etc and it will be better than all-seeing, and if done decent maybe will let us forget for a few moments we are battling ones and zeroes.

[dimlee]

Ones and zeroes they are but they never stop to surprise us mighty humans. ;)
I like to meet AI online where their skill levels are unknown to me. Sometimes when I relax too much and expect another easy kill... I'm punished by that mysterious super sharp sniper gunner who sends his bullet right into my cockpit from 1000m. I consider that as a "compensation" mentioned by Storebror and continue to enjoy the game.
:cool:

[Storebror]

Right dimlee.
We've lately adopted the 4.13 AI into the good old Ultrapack 3 which we run on our SAS Gameserver (where we're sporting COOP missions mostly, with lots of AI) and it turned the game into something completely new, where we yet have to figure out all the new moves AI has learned.

majorfailure has a lot of good points there.
What I'm trying to say is that regardless of it's superduper abilities in terms of situational awareness, AI already now sucks bad in many other elements of the fight.
Let me give you a few examples:

When AI is outnumbering human players and can keep the fight going, they do a really great job (with 4.13 AI code) to help each other, distribute their planes across human opponents etc., all fine, but...
When AI is outnumbering human players but a single human player has the fastest plane in the set and can run away from the fight, if he's managing the distance to AI planes (keep it at slightly above 1km, never separate further than 1.5km from the closest opponent), he can group up all AI planes behind him in a big swarm and drag them away to a point of choice on the map, then run away and leave them there, completely confused.
When human players are outnumbering AI and AI planes have no option to climb and run away, AI sucks big times in defending their lifes. They will do a couple of weird maneouvres but this will stop quite soon, so all a human player has to do is stay on an AI's six for half a minute and wait for it to go straight - it will keep going straight, even if you start shooting parts off the plane, until you finally kill it.
Outnumbered AI will immediately stop helping each other, instead they'll all fight a fight on their own.
AI bombers don't try to stay in formation when being attacked - when you hit one, he will bounce out, even if he could stay in as well.


This is why I'd suggest not only to think about how and where AI uses superior powers at the moment and how to tone them down, but also how and where AI currently lacks desireable abilities and how to improve them.
Because if we'd just tone down AI's situational awareness, this would have to be compensated elsewhere, and with AI's current abilities the only compensation available would be to give the superpowered AI flightmodel even more super powers of to make the snipers even more sniper like.

I don't think anyone would want AI to become more stupid and more deadly by surprise.

Best regards - Mike

[taly001]

And you don't need experience to set a limit for maximum contacts to be able to track. Script can do that

I just saw a show about a concentration and memory test done with computer graphics object tracking, tracking 4 objects was the "peak point" where high skilled gamers outscored non or low skilled gamers. At less than 3 or more than 5 objects player skill level mattered less.

Storebror AI items listed I agree with, I just add the easy kills of planes flying straight for home that don't evade when hit is the most annoying one, I noticed it worst with recon/patrol fighters.

Il2 AI is now pretty good for dogfighting, and most things.

[sniperton]

@Storebror
What your examples describe can be summarized as 'bad leadership', and an inexperienced or inept human squadron leader would make the same wrong decisions (allowing to be lured away instead of disengaging and regrouping, or issuing 'each on his own' instead of 'together to the Walhalla').

[dimlee]

Regarding AI bombers.

Some observations - since I resumed my online adventures couple of years ago.
- B-17, B-25, He 111, SB hold formations quite well. Slightly damaged bomber still stays in. Quite a task to break a formation without large calibre cannons.
- Blenheims and Beauforts keep well in pairs, but not in a group of 3 or 4.
- B-25 formation of 4 can be broken (occasionally) by flying through it without firing a single shot. Leader turns left, wingmen turns right, etc. Yet to see the same with other types.
- I have not seen AI bomber aerobatics for long time.
(But I see it offline, B-29 and TB-3 are most notorious).

[dimlee]

I just saw a show about a concentration and memory test done with computer graphics object tracking, tracking 4 objects was the "peak point" where high skilled gamers outscored non or low skilled gamers. At less than 3 or more than 5 objects player skill level mattered less.

Interesting... The same was noticed years ago with marine radar operators - before ARPA systems were introduced.

[majorfailure]

I don't think anyone would want AI to become more stupid and more deadly by surprise.

Best regards - Mike
That's exactly what I would like to see, for rookies(maybe average too). Better shooting -especially against targets that do little to no evading -including ground targets. Less SA, target fixation is a typical rookie move IMHO - at least I did that a lot. Veteran and above to me are more convincing, superb SA, mean shooting, but not entirely invincible, and now they sometimes miss the "ridiculous" 90 degree shots, maybe as much as I do.

But still - I don't see why giving the AI an ability to disengage would weaken them and then they need strengthening in another department?

And to add to your stupid AI collection - try attacking a flight as lone wolf, do it as flight leader, they are all over you in seconds. Do it as second in a flight and you can usually pick off one or two planes before reaction is initiated.

[sniperton]

But still - I don't see why giving the AI an ability to disengage would weaken them and then they need strengthening in another department?

I’m with you, but there’s a dilemma. Aerial battles are now more decisive (and more arcade-like) than in RL were, there are more kills for the winner and higher losses for the underdog. This is fine for dogfights and coops, as we all want immediate action, and since we don’t risk our lives, we prefer a bloody massacre to an enemy who ‘cowardly’ disengages as soon as the odds are against him.

Teaching the AI to disengage and to evade combat would involve longer sessions with less action and more advantage-seeking intermissions. I’m personally for it, but it’s a matter of gaming preferences. Perhaps better shooting at non-evading targets would compensate for the more evasive AI and re-establish the sort of balance as we have now.

On the other hand, it’s unpredictable how all this would affect single-player campaigns where mostly the AI is fighting the AI. My point is that what might be desirable against human opponents could easily lead to stalemates in AI-AI encounters and would possibly break many single-player missions and campaigns designed and playtested with the non-evasive AI we have at present.

[Pursuivant]

One thing we shouldn't forget is that AI is comparably "stupid" when it comes to draw conclusions based on the facts they know.

It doesn't have to be. For relatively simple "pure" air combat (i.e., no bomber defense, no ground attack, no strategic objectives, just a dogfight) the "decision tree" can make AI fight fairly realistically. In fact, I would guess that the real trick is to keep it from being too good!

I'd bet that "deep learning" AI programming - were it to be implemented into an air combat sim - would result in some frighteningly effective and realistic AI behavior after a short period of time, to the point that the AI is legitimately unbeatable by all but the most talented human players. Of course, that's a pipe dream given the current time and cost required for machine learning (unless you work for DARPA or General Atomics).


In that regards, letting AI "cheat" in another regime to compensate this lack of experience might be a valid decision to some degree.

I agree, but any AI cheats have to seem fair to the players. In particular, AI "reflexes" can't be any better than a good human player's - no laser-guided gunnery, 360-degree radar vision, or instant, perfect control inputs.

I think that the best way to prevent cheats from being obvious is to have a simple percentage chance based on skill that the AI will screw up and do something random and/or stupid, rather than acting with killer robot efficiency.

[Pursuivant]

When AI is outnumbering human players but a single human player has the fastest plane in the set and can run away from the fight, if he's managing the distance to AI planes (keep it at slightly above 1km, never separate further than 1.5km from the closest opponent), he can group up all AI planes behind him in a big swarm and drag them away to a point of choice on the map, then run away and leave them there, completely confused.

To some extent, that's not unrealistic, since aggressive, inexperienced pilots can get "target fixated." But, a good leader is going to eventually get his flights back in formation, and perhaps position flights so that they can be mutually supportive vs. a faster opponent making repeated diving attacks.

Lack of effective section and flight tactics, like the classic "drag and bag" or "Thatch weave"/beam defense, is where 4.13 AI is currently weakest. This isn't just a fault in IL2; I can't think of any combat flight sim that does AI group tactics well.

When human players are outnumbering AI and AI planes have no option to climb and run away, AI sucks big times in defending their lifes. They will do a couple of weird maneouvres but this will stop quite soon, so all a human player has to do is stay on an AI's six for half a minute and wait for it to go straight - it will keep going straight, even if you start shooting parts off the plane, until you finally kill it.


Outnumbered AI will immediately stop helping each other, instead they'll all fight a fight on their own.

Again, this might be realistic. Information overload and panic can easily overwhelm training in situations like this. Historically, it was very common for formations of aircraft to break down into a "furball" consisting of many confused individual fights. (In fact, many aces took advantage of this feature of aerial combat by exiting the furball as quickly as they could, orbiting above and outside it, and picking off enemy aircraft which had lost Situational Awareness due to getting "sucked into" the furball.)

That said, Veteran or Ace AI should have good methods of being able to retain mutual support while badly outnumbered.

AI bombers don't try to stay in formation when being attacked - when you hit one, he will bounce out, even if he could stay in as well.[/quote]

Heavy or medium bomber AI behavior in IL2 just isn't that good, probably because the sim was never designed to to take those sorts of bomber ops into account.

Bomber crew behavior is grossly simplified, and frustrating for anyone who want to play a bomber crewman. No intercom warnings about incoming bandits or aircraft maneuvers. No ability to aid other crewmen. No ability to coordinate fire. No ability to jettison guns and other equipment to lighten bomber weight. No crew ability to fight fires. No ability for crew to assume proper locations within aircraft for ditching or crash landing. G forces have no effects on gunnery accuracy or bailout ability.

Likewise, bomber "box" or formation behavior isn't modeled well. No loosening up formation when there's flak ahead but no fighters. No tightening up formation when fighters are spotted. No ability for lead bomber to order other bombers in formation to drop bombs on the lead bombardier's command. No ability to command bombers in your formation to do certain things - drop bombs, etc.


Because if we'd just tone down AI's situational awareness, this would have to be compensated elsewhere, and with AI's current abilities the only compensation available would be to give the superpowered AI flightmodel even more super powers of to make the snipers even more sniper like.

I don't think that anyone will object to AI gunnery becoming worse. Most players won't care if AI can fly their planes to the absolute limit of performance and have radar-like vision, as long as they can't shoot straight.

I think it's highly realistic for inexperienced AI aircraft - or even some veteran bomber AI - to start shooting at long ranges. But, if they do that, they should be wasting ammo except for lucky shots.

For shots in dogfights, I think it's realistic for AI to get their lead wrong - sometimes badly wrong - based on relative speed, distance, and G forces involved. But, if there's no obvious effect from their fire, they should stop shooting.

If you want historical accuracy - vs. "realism" based on dogfight server human behavior - it's also realistic for any pilot of less than Veteran quality to absolutely suck at deflection shooting.

Also, unless they're fanatical or otherwise highly motivated, most pilots aren't going to go out of their way to engage in combat. If they lose sight of friendly aircraft, they'll fall back to a rally point. If they can't make contact with friendlies and have any reason to return to base, they will. If they see a big formation of enemy AC over enemy territory, they'll probably find an excuse to not attack.

Remember, just 10% of all US combat fighter pilots accounted for over 50% of all US fighter kills. All the other fighter pilots didn't have the piloting or gunnery skills to get a confirmed kill, or didn't have the aggressiveness and situational awareness to be a winner in aerial combat. To be utterly brutal, they were nothing more than targets, or "turkeys," although they might have been effective pilots in other roles, such as ground attack or bomber escort.

Pilots not selected for fighter pilot training were probably going to be even worse, although there were notable exceptions.

[majorfailure]

I’m with you, but there’s a dilemma. Aerial battles are now more decisive (and more arcade-like) than in RL were, there are more kills for the winner and higher losses for the underdog. This is fine for dogfights and coops, as we all want immediate action, and since we don’t risk our lives, we prefer a bloody massacre to an enemy who ‘cowardly’ disengages as soon as the odds are against him.

Teaching the AI to disengage and to evade combat would involve longer sessions with less action and more advantage-seeking intermissions. I’m personally for it, but it’s a matter of gaming preferences. Perhaps better shooting at non-evading targets would compensate for the more evasive AI and re-establish the sort of balance as we have now.

On the other hand, it’s unpredictable how all this would affect single-player campaigns where mostly the AI is fighting the AI. My point is that what might be desirable against human opponents could easily lead to stalemates in AI-AI encounters and would possibly break many single-player missions and campaigns designed and playtested with the non-evasive AI we have at present.

Solution: Make it a difficulty switch to have evasive AI. And no I wouldn't mind if battles would not end in slaughter on a regular basis. Getting home with only a few kills then would be more rewarding. And imagine what a feat it would be to get home with 10+ kills in one sortie, which now in certain planes is rather the usual case.

[sniperton]

Solution: Make it a difficulty switch to have evasive AI.

Good idea, if it can be done that way. Seems there's an agreement in

1. a skill-based delay in the re-acquisitioning of targets (which weakens the AI);
2. better AI shooting at non-evading targets (which strengthens the AI);
3. an evasive AI which makes it more difficult to overrun AI squads (optional).

[Pursuivant]

Solution: Make it a difficulty switch to have evasive AI.

I've been asking for something like this for years. Mission/Campaign builders should be able to set the level of AI aggressiveness in both FMB and QMB, for both offline online missions.

There are times when you want the AI to be insanely aggressive - like in a QMB fighter sweep mission or online dogfight server. But, there are also times when you want the AI to be cautious, such as during many historical campaigns.

[Marabekm]

It seems the AI perform a little more realistic if you don't set a target.
Example: I want a group of zeroes to intercept a flight of SBDs. If I set the zeroes waypoint to the sbds, and set the SBDs as the target, the zeroes will stop at nothing, and follow the SBDs to the end of the world, unroll they are all shot down or the zeroes are all shot down.

Now if I just set the waypoints close together, but do not assign the SBDs as the target, the zeroes will still attack the SBDs while close, but will not follow and hunt the remainder down once the SBDs are away from that particular waypoint.

[sniperton]

Interesting observations. Could you perhaps figure out how close the Zeros have to be to the SBDs to attack them without any scripted order to do so? Probably this is the range where combat AI is activated in any case, irrespective of the mission script. Learning this threshold would open up new perspectives for mission and campaign design. :)

[dimlee]

I can confirm the observation of Marabekm.

Regarding the distance between attacking a/c and targets - I'm afraid it "depends" and each mission requires some fine tuning and testing. This is at least my impression after creating some missions in CUP/now BAT. (It was based on 4.12, don't know if 4.13 is different in that respect).
Another lesson I learned - randomness does exist. When play the same mission several times and then watch tracks I see that AI behaviour can be strikingly different which leads to various outcome. At least in well balanced scenarios where neither side has decisive advantage at the start.

[Marabekm]

Interesting observations. Could you perhaps figure out how close the Zeros have to be to the SBDs to attack them without any scripted order to do so? Probably this is the range where combat AI is activated in any case, irrespective of the mission script. Learning this threshold would open up new perspectives for mission and campaign design. :)

I'm unsure at how close they have to be. Usually I just overlap the waypoints and try and make sure the red and blue get to the same point at the same time.
On the down side, I have not been able to make this work with the escort fighters. They drop their tanks like they want to engage, but never do and just continue flying like normal, while the friendly bombers get killed right below.

[dimlee]

I'm unsure at how close they have to be. Usually I just overlap the waypoints and try and make sure the red and blue get to the same point at the same time.
On the down side, I have not been able to make this work with the escort fighters. They drop their tanks like they want to engage, but never do and just continue flying like normal, while the friendly bombers get killed right below.

Probably they will engage if timings and/or positions are slightly changed.