AI vs AI Kill ratio

[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]

Originally Posted by taly001

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]

Originally Posted by Storebror

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]

Originally Posted by majorfailure

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]

Originally Posted by 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.

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).

Originally Posted by Storebror

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]

Originally Posted by Storebror

[*]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.

Originally Posted by Storebror

[*]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.

Originally Posted by Storebror

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]

Originally Posted by sniperton

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.