Missile Dynamics - A discussion

[GGTharos]

That looks like a reasonable exchange ratio ... but for AI it means little. R-77's are about three times as effective in the hands of humans. Or were.

Wait for the tuning to be mostly, if not completely done.

[danilop]

There are only two possible conclusions:

 

1. Russian AI is worse

2. Russian planes + weapons are inferior

 

Just saying ...

 

Would like to hear from devs on this ...

[GGTharos]

The devs will tell you that the missiles are being tuned. We're aware of all this stuff. There's a reason why we don't use AI in these tests, and Alfa nailed it.

[Exorcet]

I'm honestly surprised. I tried this in DCS World pre FC3 and the F-15's were about useless. They would lose even if the other side didn't have ARH.

 

As for these tests though, the F-15 should be coming out on top. It's a better BVR plane with better missiles, and the AI only knows how to light the burners and hold down the trigger. Even in the F-15 vs MiG-29+77, the Eagle probably should have a better than 50/50 ratio. The MiG's aren't going to take into account the fact that their missile has shorter range, nor will they use EOS (although AI is all seeing anyway), and nor will they care about R-73 and HMS until it's in range.

 

If this was real life, I'd think the Eagles would probably win too as both planes are out in space conducting a battle without the support of any other assets, although the final result is not as clear. The F-15 loses AWACS and support like SOJ, but is otherwise OK as it still has a big radar and good missiles. The MiG, being a short range interceptor would probably have more trouble finding the F-15's than the other way around, however now it gets murky as real MiG pilots would be able to employ tactics the AI would never dream of.

 

Now throw these planes in real life in a US vs Russia situation and then I think you'd get an even performance from both of them.

 

There are only two possible conclusions:

 

1. Russian AI is worse

2. Russian planes + weapons are inferior

 

Just saying ...

 

Would like to hear from devs on this ...

 

Or that the AI scripting simply favors the F-15 at the moment. No one with experience is going to confuse an AI fighter for a player.

[Alfa]

Look at ARH results ;)

 

I did and while I find the results of that scenario more suspect, I don't quite understand your choice of opponents.

 

If you are conducting the whole thing as AI vs. AI anyway, then why not choose more appropriate platforms - e.g. F-18 with 4x AIM-120, 2x AIM-9 vs. MiG-29S with 4x R-77, 2x R-73.

 

At least that way you would exclude the radar range factor - i.e. F-15s possibly enjoying a "first shoot" advantage(more powerful radar).

[danilop]

I am afraid that the whole logic behind your test is faulty - i.e. that leaving out the human factor will help better determine the state of missile modelling. It will do the contrary - AI isn't sophisticated enough to employ tactics necessary to counter the obvious disadvange of using SARH vs. the other side using ARH missiles.

 

An AI controlled aircraft will simply drop everything and start evasive manounvering when a missile is launched at it - this is the case for both sides, but with SARHs a lost lock means a lost target, while with ARHs there is still a good chance that the missiles will find their targets autonomously....especially if fired at relatively close range.

 

Well he only mentions SARH vs ARH case and that may be (highly unlikely) true, however in the second case including 120c vs R77 it's obvious that there is significant difference between sides.

 

Some difference is to be expected, but this is way, way to much.

 

As I said before, there are two possible scenarios really - bad Russian AI or modeling of the planes and missiles.

[danilop]

....

 

At least that way you would exclude the radar range factor - i.e. F-15s possibly enjoying a "first shoot" advantage(more powerful radar).

 

Almost all kills are WVR

[Alfa]

There are only two possible conclusions:

 

1. Russian AI is worse

2. Russian planes + weapons are inferior

 

Just saying ...

 

Would like to hear from devs on this ...

 

There are really two factors involved when matching up the Russian fighters with the Eagle in the sim:

 

1). the Su-27 and Su-33 have a large radar, but employ an earlier generation missile which is inferiour to an AMRAAM - and not just because its SARH.

 

2). the MiG-29S employs a contemporary and capable missile, but has a smaller radar.

 

Hence my suggestion to try with a Hornet vs MiG-29S instead - in reality that wouldn't be an even matchup either, since the Hornet's radar is more sophisticated than the MiG's, but I doubt that this is depicted for the AI in the game.

[Alfa]

Almost all kills are WVR

 

Even so.

[Angel101]

well, I don´t see any problem to got better missiles in one side. I hate the games were an M16 shoots exactly an Ak47, no reason to choose between them that just aesthetically.

 

Like me, there are a few people that plays Blue sometimes and Red others (most vs IA) and it´s funny having to adapt to each side and their weapons.

[danilop]

Do you think that from 35 fired R77s on average, only four should hit???

 

That's like three to four times less than AIM-120c PK modelled ATM.

 

Completely unrealistic.

 

No one is denying (slight) advantage that F15c should have over the Russian planes modeled in the game - the problem is that this difference is huge!

Edited January 11, 2013 by danilop

[Angel101]

Well yes, I expect a better performance than this for a R77... and also for AIM-9p (I don´t know the reason, but more often one goes away from enemy jet with a good lock and good distance)

 

So, I think that the answer is just wait for 1.2.3.4 and test again the news changes.

 

Greetings

[danilop]

:beer:

[Kuky]

Do you think that from 35 fired R77s on average, only four should hit???

 

That's like three to four times less than AIM-120c PK modelled ATM.

 

Completely unrealistic.

 

No one is denying (slight) advantage that F15c should have over the Russian planes modeled in the game - the problem is that this difference is huge!

 

First I think to have good comparison, you need to have exact same conditions for each missile fired (same range, same aspect, same target type, same AI level, same deployment of cm's etc) and only if this condition for testing is established, then you need to get bit larger sample number to have somewhat decent statistics.

 

Setting AI vs AI and having them fight it out will not work as good testing, so if you want to do good testing you need to test as I described above

[danilop]

Yeah I know.

 

It's obvious that this is not scientifically correct or statistically valid model, but only indication that something is wrong. The idea was to exclude human factor from equation.

 

I ran mission about five-six times, conditions were as equal as I could make them with my modest knowledge of the editor and F15-s were way superior every single time.

 

I mean, we wouldn't have such a long thread and endless discussion if everything is as it should be. As we say here, and I think you will understand: "Gde ima dima, ima i vatre!"

 

Maybe after hundred of samples we would have something statistically valid, and I'm pretty confident that final results would not fluctuate to much.

Edited January 11, 2013 by danilop

[EtherealN]

F-15's get to launch on migs first due to radar.

 

Thus, even if the migs also had AIM-120 (which they do, sort of, since they have R-77), the F-15 should win, and I wouldn't be surprised if they win decisively when using AI since, well, it's not the smartest. For this reason, testing the way you do is actually without value; the confounding factors are too large to draw conclusions of anything from it.

 

What I would suggest is this:

 

Set up a mission where your F-15 takes on another target. Note exact launch conditions. Repeat multiple times. Ensure said conditions are useful also on the MiG-29S and Flanker.

Now do exactly the same thing, but flying MiG-29S.

Now do exactly the same thing, but flying the Flanker.

 

NOW you can start doing comparisons. But you'll have to do each a couple times.

[danilop]

First I think to have good comparison, you need to have exact same conditions for each missile fired (same range, same aspect, same target type, same AI level, same deployment of cm's etc) and only if this condition for testing is established, then you need to get bit larger sample number to have somewhat decent statistics.

 

...

 

I think that conditions where everything is random (like aspect, range, speed) are much better approximation of real world chaos during war - there were even couple of friendly kills during test, which I think is fairly realistic in combat.

[EtherealN]

Ah yes, I think that conditions where everything is random (like aspect, range, speed) are much better approximation of real world chaos during war - there were even couple of friendly kills during test, which I think is fairly realistic in combat.

 

But we're not talking about the chaos of war here. You are looking to establish the relative strength of the missiles. In that case, you need to make sure you are comparing apples with apples.

 

(Doing comparisons that include the "chaos of war" is possible, certainly, but basic rules of statistics mean that you need to get a sample size that overpowers all confounding factors - thus, you're looking at doing thousands of flights with various setups.)

Edited January 11, 2013 by EtherealN

[danilop]

Well, you can do that if you put F-15 against F-15, Or Su vs Su ...

 

One firing AIM-120c - the other firing R77

 

That's the only way to test missile performance excluding all other parameters.

[danilop]

...

(Doing comparisons that include the "chaos of war" is possible, certainly, but basic rules of statistics mean that you need to get a sample size that overpowers all confounding factors - thus, you're looking at doing thousands of flights with various setups.)

 

It's obvious that something is wrong from this four-five examples I've posted. THAT SHOULDN'T HAVE HAPPENED, like ever!

 

I don't want to win Nobel price for mathematics or statistics, just pointing out that something is wrong in the missile department.

 

It's way off, you have to admit that. If AI was equal (it should be) those results shouldn't have happened.

[EtherealN]

No.

 

Situation:

 

F-15C, angels 25, M0.9. Launch on co-altitude hot MiG-29 (or whatever). See how many times missiles is defeated out of (say) 10 attempts when launch is performed at 15nm.

 

then

 

MiG-29C, angels 25, M0.9. Launch on co-altitude hot MiG-29 (or whatever, but same as before). See how many times missiles is defeated out of (say) 10 attempts when launch is performed at 15nm.

 

Tell me which parameter is left to exclude? What I can think of is the fact that we cannot control what the target AI is doing. But at least we have controlled the other side of the equation - the launching platform.

[Exorcet]

What I can think of is the fact that we cannot control what the target AI is doing.

 

Possibly set it to ignore enemies and give it complex closely spaces waypoints?

 

Either way, I think the best way to look at this would be MP stats. That's basically "real world" testing in this case.

[danilop]

F-15's get to launch on migs first due to radar.

 

Thus, even if the migs also had AIM-120 (which they do, sort of, since they have R-77), the F-15 should win, and I wouldn't be surprised if they win decisively when using AI since, well, it's not the smartest. For this reason, testing the way you do is actually without value; the confounding factors are too large to draw conclusions of anything from it.

 

Great majority of kills were in WVR arena, so advantage in radar department is practically excluded. Have a look at the mission. Run it a couple of times.

16vs16.miz

[danilop]

...

Either way, I think the best way to look at this would be MP stats. That's basically "real world" testing in this case.

 

That would be true if sides had virtual pilots of exactly the same abilities which was never the case.

 

In reality, that would be tuning and equalizing for multiplayer - BTW, the side which dispute modelling of the missiles ATM, is accused of wanting just that! Irony, or what? :)

[EtherealN]

It's obvious that something is wrong from this four-five examples I've posted. THAT SHOULDN'T HAVE HAPPENED, like ever!

 

How many times have you ALREADY been told, in this thread and in others, that missiles are still being tuned and there are several known issues with them?

 

I don't want to win Nobel price for mathematics or statistics, just pointing out that something is wrong in the missile department.

 

ED testers have already, multiple times, said that this is known because missile AFM is still being tuned. Everyone already knows, for example, that the ER isn't right. You have been told this long before you started your little exercise, so what are you looking to prove here? That we were right when we told you "yup, missiles are still being tuned"?

 

Please read when people talk to you. If you're trying to say "missiles don't work right in FC3 beta right now", then yes - we know. We've said so many many times.

 

Some quick pickups:

http://forums.eagle.ru/showpost.php?p=1637299&postcount=792

http://forums.eagle.ru/showpost.php?p=1648570&postcount=63

http://forums.eagle.ru/showpost.php?p=1648681&postcount=74