4.13 development update discussion and feedback

[Derda508]

Again, I think, we have to seperate between historically correct and reasonable for a gameplay. Current fighter AI (below ace level) is much to agressive. Lots of real pilots will have prayed for a mission were they dit not have an engagement with enemy fighters. Also, from what I read, it was extremely rare for a (German) fighter pilot to try more than one attack run on bombers. The stress to face all these dozens of machine guns was just not managable, only a few aces were ablöe to do that. The typical thing was: trying to get on higher altitude (if there is enough time), rushing down on the bombers (if you find them), while hoping that the covering fighters look elsewhere, pouring all ammo onto one target (if it was possible to get close to one) and then dive to get the hell out of there. AI in contrast attacks again and again until each one of them is killed. A behaviour that fat Göring would have loved, but did not happen usually. I remember from an interview with Günter Rall that from 44 on only every fifth rookie survived his tenth mission. The most frequent cause of death was not being shot down, but accidents during landing and take off ...

[Pursuivant]

This article, written by a noted wargame designer and air war historian gives some data on hit percentages for torpedoes and dive bombing against ships.

DT might find it helpful in establishing baselines for dive bombing and torpedo bombing accuracy, since it involves pilots of known skill aiming against targets of a known size from known altitudes.

http://www.ospreypublishing.com/arti...en_it_counted/

Takeaway

Japanese at Pearl Harbor (i.e., excellent torpedoes, Veteran or better crew who had trained extensively for the mission, and who were attacking static, battleship-sized targets) achieved 37% overall accuracy - which was exceptional.

Level bomber accuracy for all other skills and nationalities was 1-3% in 1941, but Japanese pilots preparing for Pearl Harbor were getting hits 14% of the time, from 8,500 feet. At Pearl Harbor, they hit 43% of the time with level bombing attacks.

Up until 1941, German Stukas (all skill levels) achieved 25% accuracy in dive bombing attacks. Japanese dive bombers were hitting practice targets 30% to 34% of the time; at Pearl Harbor, under fire, their accuracy rate was at least 26%. So, 25% dive bombing accuracy while under fire is probably appropriate for Veteran pilots.

By late 1944 the US Navy was pleased to achieve a 40% accuracy rate for practice torpedo bombing (i.e., presumably Rookie level pilots). In training the Japanese torpedoes were hitting 70% to 80% of daylight targets and 50% to 75% at night; at Pearl Harbor their accuracy rate was 51%.

According to the US Navy 20 torpedoes hit American ships. Only the 40 Mitsubishi 'Kate' torpedo bombers of the first wave carried torpedoes. The USS Nevada shot down one 'Kate' before it could release its load, so only thirty-nine torpedoes were dropped. Twenty found their mark = 51% hits.

So, 50% or so can be taken as the maximum accuracy level for a Veteran or Ace torpedo bomber aiming against a static battleship while under light and inaccurate AA fire. Assuming that combat effectiveness of torpedo attacks is reduced by 30% compared to training runs, perhaps 10% hits for Rookies making daylight attacks, and 1-3% for night attacks would be realistic

In the first wave 21 of the 49 800kg high altitude bombs scored direct hits on the battleships USS Arizona, California, Maryland, Nevada, Pennsylvania, Tennessee and West Virginia; three more damaged the USS Oklahoma with near misses. The Japanese pilots targeting battleships had a 43% accuracy rate, 49% including near misses. In either case, the Nakjima B5N 'Kate's' level bombing was far better than predicted and much better than the 13%-14% achieved before intensive training.

In the second wave, of 80 bombs, 21 were effective, a 26-27% accuracy rate which is still less than the 30%-34% scored in practice. The difference is accounted for by the fact that the pilots were diving into smoke and evading flak.

So, medium-altitude level bombing by light bombers piloted by Veteran aircrew should have about 43-49% accuracy in combat, reduced to 26-27% if the target is obscured by smoke and/or planes have to deal with flak. Average to Rookie level level bombing in 1939-41 should have dismal results with just 1-3% accuracy, especially if dropping bombs from higher levels or in combat conditions.

[Pursuivant]

Originally Posted by Derda508

Again, I think, we have to seperate between historically correct and reasonable for a gameplay. Current fighter AI (below ace level) is much to agressive.

Yes. There is a point where "realistic" isn't "fun" and IL2 is a game as much as a simulation.

Right now, I think that DT has gotten the "fun" aspect of rookie fighter vs. fighter AI right. They're clumsy but aggressive, making it easy for an experienced human player to shoot down a lot of them. Other aspects of Rookie AI are very frustrating, however, and should be fixed. Note that "fixing" could be as simple as telling players, "Yes, they really were this bad."

I've always maintained that the solution to different interpretations of "rookie" is to have a level of AI below rookie, called "cannon fodder." They would behave like old, pre-DT rookies in combat, but with no SA or deflection shooting skills at all, and very limited sighting skills.

That way, Rookie AI is reserved for your average "nugget" straight from a well-run training school. Clueless, and needing lots of polishing to turn him into a decent air warrior, but with the basic skills he needs to succeed.

If possible, it would also be nice to be able to set pilot aggression levels in the FMB or campaign setting, from "reckless > aggressive > cautious > cowardly".

Most pilots would fall into the "cautious" range, attacking only when they have several clear advantages like height and position, and they clearly outclass their foes. Ace AI plus "cautious" gives your classic high-scoring "sniper" ace like Manfred Von Richthofen, Eric Hartmann or George Beurling.

Cowardly pilots will avoid contact, will fail to carry through attacks and might well panic when fired on. Combine that with Rookie AI and you get your classic "turkey" pilot. Combine it with Ace AI and you get a "burnout" who's on the ragged edge of Combat Fatigue/PTSD, or who believes that the war is lost and his main job is to keep himself alive (typical of many French pilots in 1940 after the Blitzkreig really started to bite).

Reckless pilots will charge right in regardless of the odds and are otherwise utterly fearless - basically dead heroes waiting to happen. With Ace AI that gives you classic posthumous medal winners like Frank Luke or Werner Voss in WW1 or Greg Boyington in WW2 (I know - not posthumous Medal of Honor, but the Marines thought he was dead when they gave it to him). With Rookie AI, you get kamikazes.

Originally Posted by Derda508

Lots of real pilots will have prayed for a mission were they dit not have an engagement with enemy fighters. Also, from what I read, it was extremely rare for a (German) fighter pilot to try more than one attack run on bombers. The stress to face all these dozens of machine guns was just not managable, only a few aces were ablöe to do that.

This is the effect of morale, which isn't well modeled in IL2. Had the Germans known just how inaccurate all those guns were, they would have charged right in. But, seeing all those tracers, all "aimed at you" freaked out all but the bravest. For this reason, German bomber gun loadout included some practice explosive rounds specifically to make attackers think that they were under a greater intensity of fire than they actually were.

Of course, having Rookie AI refuse to engage bombers would be "not fun." That's where "bravery levels" can be used to balance between realism and fun. Rookie but Reckless = current Rookie vs. Bomber AI, Rookie but Cautious/Cowardly = realistic German rookie bomber intercept behavior.

Originally Posted by Derda508

I remember from an interview with Günter Rall that from 44 on only every fifth rookie survived his tenth mission. The most frequent cause of death was not being shot down, but accidents during landing and take off ...

Rall's interview squares with the data I've posted previously (taken from Spick's "The Ace Factor," but repeated in Bungay's book on the BoB).

The high number of accidents on landing and takeoff might be due to evil habits of the late model Bf-109 - lots of engine torque and a narrow landing gear - plus landing on improvised or bomb damaged fields. But, again, losing 10-20% of your pilots due to landing and takeoff accidents on every mission is "not fun."

[TexasJG]

May have mentioned this before, for 4.13 or 4.14
be nice to have analog controls for the X-4, Fritz, and for the Mistel bomber (along with any other guided munitions).
As is, the X-4 and Fritz are a digital control, and the Mistel has none.

And speaking of the Mistel, it would be cool to have an Mistel FW-190/Ju-88 combo static object, and the ability to use this weapons system in Dogfight mode.

Just one of a million request..
Thanks Daidalos Team!!

[Laurwin]

Best way to deal with tough manouvering AIs is to understand the following tips from real American F-16 pilot Pete Bonanni.

-BFM manouver is always flown "in the future" as it were. In order for a high yoyo to become effective, the bandit must fly according to your prediction of future.

Time happens on a different basis compared to BFM drawing on a piece of paper. Time happens in definite way, the present moment becomes the past, and this continues on and on. The definite version of future is always some kind of abstract assumption of future events. Future in the abstract becomes the present, and present becomes the past. etc...

When you deal with simple and short-term assumption of future event, this is always easier to define compared to long-term future event. When you look at a fast flying bandit, it follows that the bandit's turn circle is larger, therefore, the bandit will not be able to turn so easily... Fast flying bandit = more difficult for bandit to turn aircraft... G force limit of bandit pilot and bandit aircraft form a limitation to turning circles... This assumption derives itself from physics of inertia. But you still cannot exactly predict bandit's future course, you can only assume within certain parameters.

The F-16 modern gunsight operates on this similar kind of principles. EEGS Gunsight. Radar-gunsights should be theoretically better, because radar beams are light-speed, so it doesn't take very long time to find the target in the sky with the radar. But computer still must calculate the lead for the gunshot, at the end of the day, I think...

With WW2 gunsights it's the pilot himself who calculates the lead.

The only consistently possible way to dogfight, is to employ following measures.

1. observe bandit

2. predict bandit's future position

3. fly your aicraft based on this prediction of future.

4. if something changes, react to changes.

Those four lines form the so-called OODA loop by the way.

[Pursuivant]

Originally Posted by Laurwin

-BFM manouver is always flown "in the future" as it were. In order for a high yoyo to become effective, the bandit must fly according to your prediction of future.

This is very good advice in any dogfight, and clarifies comments by WW1 and WW2 aces about the necessity of instantly anticipating your opponent's intentions, position and energy state.

As G-forces go up and airspeed and altitude goes down, the number of maneuver options your opponent has goes down as well. That allows you to predict what they're going to do within certain parameters.

AI is often quite predictable in this regard, doing things like disengaging out to 2 Km and gaining 500 meters of altitude prior to coming in for another BnZ attack, or pulling up sharply when pursued, bleeding off energy and giving you a lovely close range, low deflection shot.

[Derda508]

Originally Posted by Pursuivant

The high number of accidents on landing and takeoff might be due to evil habits of the late model Bf-109 - lots of engine torque and a narrow landing gear - plus landing on improvised or bomb damaged fields. But, again, losing 10-20% of your pilots due to landing and takeoff accidents on every mission is "not fun."

The late 109´s certainly were rookie killers. But it happened to FW 190 pilots as well. In ´44 many rookies were trained on old, obsolete Italian fighters. In these machines you lowered throttle by pushing the lever in. As soon as they sat in much more powerful "modern" combat planes, it was just the other way round. Not much imagination needed about the effects ...
In his novel "Der verratene Himmel" (Skies Betrayed) Rudolf Braunburg gave some vivid accounts of this, based on his own experience as a rookie 190 pilot in ´45.

(By the way: he dedicated this book to all fighter pilots who never scored a kill, but just were shot down ... like himself)

[Pursuivant]

Originally Posted by Derda508

In ´44 many rookies were trained on old, obsolete Italian fighters. In these machines you lowered throttle by pushing the lever in. As soon as they sat in much more powerful "modern" combat planes, it was just the other way round.

The French planes were also built this way - push the throttle to slow, pull it towards you to accelerate. Since the Luftwaffe also used a number of Dewoitine D.250 as fighter trainers, I could see those planes causing problems as well.

It's surprising that the Luftwaffe didn't think to fix this problem. I also have to wonder how many Free French or Italian Allied pilots flying British, Soviet or American types died because in a moment of panic they pushed the throttle the wrong way.

[Treetop64]

Originally Posted by Derda508

Again, I think, we have to seperate between historically correct and reasonable for a gameplay. Current fighter AI (below ace level) is much to agressive. Lots of real pilots will have prayed for a mission were they dit not have an engagement with enemy fighters. Also, from what I read, it was extremely rare for a (German) fighter pilot to try more than one attack run on bombers. The stress to face all these dozens of machine guns was just not managable, only a few aces were ablöe to do that. The typical thing was: trying to get on higher altitude (if there is enough time), rushing down on the bombers (if you find them), while hoping that the covering fighters look elsewhere, pouring all ammo onto one target (if it was possible to get close to one) and then dive to get the hell out of there. AI in contrast attacks again and again until each one of them is killed. A behaviour that fat Göring would have loved, but did not happen usually. I remember from an interview with Günter Rall that from 44 on only every fifth rookie survived his tenth mission. The most frequent cause of death was not being shot down, but accidents during landing and take off ...

I do agree that Rookie AI is too competent. For instance, there is virtually no chance ever of downing Rookie AI fighters from behind without them suddenly becoming "aware" of you before you open fire, even if you initially got the jump on them without them spotting you. Also, Rookie AI are a bit too good keeping up with and downing Veteran and Ace AI, while Veteran and Ace fighter pilots are often too aggressive in turn-fights in aircraft that are definitely not suited to the job, getting themselves shot down in the process. If there is one thing that the old AI got right, it's that AI in BnZ machines stuck to BnZ, and AI in TnB machines stuck to TnB. Yet another issue is, as you stated, a formation will continue a fight against long odds (well-defended bomber formations, attacking well-defended airfields, etc.) until every aircraft in that formation is destroyed.

Another thing that has actually become quite an annoyance is how often the no.4 in the flight goes inverted and performs other gyrations checking his six and what's underneath him, even when the flight is almost at treetop level. It's just too much, and occurs way too damned bloody often. Never cared for it. That needs to be either significantly toned down or, better yet, turned off altogether. Even entire bomber formations was doing this in a previous update of the game when the community made enough noise and said it was ridiculous, before that got removed.

On the whole the new AI is an improvement, but there are some areas that really need addressing.

[Pursuivant]

Originally Posted by Treetop64

Yet another issue is, as you stated, a formation will continue a fight against long odds (well-defended bomber formations, attacking well-defended airfields, etc.) until every aircraft in that formation is destroyed.

This is part of the morale issue that IL2 AI either neglects or can't model. Historically, the sight of all sorts of tracers and/or flak shooting past the formation would have had a demoralizing effect, making all but the most determined attackers break off the attack prematurely, or make one fast attack on the target before bugging out.

There are historical examples of planes bravely slogging through fighters and flak to certain doom (e.g., Fairey Battles at the Albert Canal or the Sedan Bridge, the kamikazes), but I think they're notably rare.

Suicidal morale aside, it's also worth pointing out that all levels AI is still stupid about not using "nap of the earth" flying and other tricks to minimize flak effectiveness.

AI should also selectively take out the first vehicle in a convoy, or the locomotive, when making ground attacks against vehicles or trains. If possible, Veteran or Ace AI should also try to line up strafing attacks so that they can shoot down the length of a train or convoy, or across a line of parked aircraft.

Originally Posted by Treetop64

Another thing that has actually become quite an annoyance is how often the no.4 in the flight goes inverted and performs other gyrations checking his six and what's underneath him, even when the flight is almost at treetop level.

I think that this problem has mostly been fixed. But, the rolling behavior seems a bit odd; it's my impression that the "Tail End Charlie" mostly checked his 6 by making rudder turns and banks.

Finger four formations would often have planes in each section weave across each other's paths to check mutual blind spots.

Fighters in close escort with bombers didn't need to "check 6" since they could rely on all the eyes in the bomber formation to keep a look out for them.

In any case, "check 6" behavior should happen a lot less frequently. Some doctrine said to not fly more than 10 seconds in a straight line in the combat zone, other doctrine suggested no more than 30 seconds. I'd split the difference - Ace maneuvers or otherwise checks 6 about every 10 seconds, Veteran about every 30 seconds, Average about every 30 seconds, but often forgets, and Rookie either doesn't check or doesn't check much beyond every 1 minute.

Originally Posted by Treetop64

On the whole the new AI is an improvement, but there are some areas that really need addressing.

Agreed. AI programming is an art, and DT has mostly got it right. Certainly, it's a joy to fly offline dogfights now. Ace or Veteran AI is quite challenging, and if you can win in a 1:2 or 1:4 fight against an equally matched Ace or Veteran plane, it's a real achievement.