AI opponents cheat in dogfight?
-
I’ve done something that some people frown on, which is that I have made a few mods to aircraft data for one specific type of F-16, but the original unmodded file is safely backed up.
I will NEVER use modded aircraft data in any online engagement of any sort. I won’t cheat against anybody else.
I modded the block 52 aircraft data by increasing the thrust multiplier and relaxing the AOA limits.
I then selected this aircraft type and set up a Block 50 as my adversary. (Which used different aircraft data.)
My modded aircraft can maintain the corner airspeed range forever in a 9 G max turn. That’s understandable given the edit I made to the thrust multiplier.
I should have an overwhelming advantage in a turning engagement due to the ability to maintain corner indefinitely.
But what I’ve found is that the AI opponent in the unmodded block 50 never falls victim to the death spiral. You probably know that if you pull 9 Gs forever, you’ll spiral down all the way to a stall and crash.
I don’t know how slow the AI opponent gets in this death spiral, but I sure don’t have much of a turning advantage over it so it’s probably not dropping below 250 knots.
The AI also never appears to suffer G-LOC or run out of fuel.
In my opinion, these three issues constitute AI cheating and make for an unrealistic dogfighting experience. The AI should replicate the same limitations of a human in the cockpit or it’s not a fair fight.
Granted, I was using a modded aircraft model, invulnerability, unlimited fuel, and no blackouts, but it takes these settings to reveal the fact that the AI apparently is using those same settings AND doesn’t model the death spiral correctly. How would I be able to confirm that the AI uses unlimited fuel if I run out of fuel before he reaches the bottom of an assumed standard tank? Or see that it G-LOCs if I myself G-LOC before observing it do the same thing?
Thoughts, comments, questions, anyone?
-
Well, the AI doesnt use the same flight model that the player does. The AI uses the older flight model from before BMS - its not smart enough to fly with the limitations the player has.
A good example of this is an ordinary takeoff. Fly full afterburner with heavy stores and do a normal takeoff - you will see your wingman ROCKET past you, and fly a circle back into formation. You flew a max performance takeoff, he started rolling AFTER you, and he still overtook you!
They can run out of fuel, but again its based on a different flight model to yours. If you fly at less than MAX AB then you can win just by them running out of fuel. Granted dogfighting in MIL means you need to play it smart, but its not impossible.
Its not a fair fight, by the way. The AI is not smart enough to kill you without missiles.
-
many wrong information in this thread
First take ACMI and you will se AI is not pulling 9G all the time except if you are flying WVR high speed
-
the AI does not use the same flight than human model BUT the turn / accel/ climb performances are SIMILAR
-
never heard about death spiral, for your information i can sustain infinite 9G turn without modding data and without entering into a “death spiral” …if you wonder how, you should leard to read Energy management diagrams. the AI anyway is able to control its G and roll so that it is able to maintain in flight enveloppe.
-
the fuel consomption of AI in WVR is the same as human and yes of course the AI runs out of fuel !
-
the only true information : the AI indeed does not suffer from G lock
-
-
many wrong information in this thread
First take ACMI and you will se AI is not pulling 9G all the time except if you are flying WVR high speed
-
the AI does not use the same flight than human model BUT the turn / accel/ climb performances are SIMILAR
-
never heard about death spiral, for your information i can sustain infinite 9G turn without modding data and without entering into a “death spiral” …if you wonder how, you should leard to read Energy management diagrams. the AI anyway is able to control its G and roll so that it is able to maintain in flight enveloppe.
-
the fuel consomption of AI in WVR is the same as human
-
the only true information : the AI indeed does not suffer from G lock
Cheers for the corrections. I do note that their acceleration performance is not THAT similar… seems to be enough fudge room to let wingmen stay in their positions even when they shouldnt be able to
-
-
Cheers for the corrections. I do note that their acceleration performance is not THAT similar… seems to be enough fudge room to let wingmen stay in their positions even when they shouldnt be able to
the performance are similar , but AI suffer from a poor thrust management AND they dont use the same atmosphere (AI uses standard atmo while Human use standard evolving atmo) , which is very visible in transonic ans supersonic regimes….
will be fixed
-
- never heard about death spiral, for your information i can sustain infinite 9G turn without modding data and without entering into a “death spiral” …if you wonder how, you should leard to read Energy management diagrams. the AI anyway is able to control its G and roll so that it is able to maintain in flight enveloppe.
can you elaborate on this or point me to some more detailed info? thx!
-
Can you elaborate on which part you wanted elaborated on?
If you mean about how to sustain 9G for long periods of time, try this setup - 9.0G at sea level, entering the turn starting at 600 knots.
-
I mean the diagrams -
-
many wrong information in this thread
First take ACMI and you will se AI is not pulling 9G all the time except if you are flying WVR high speed
-
the AI does not use the same flight than human model BUT the turn / accel/ climb performances are SIMILAR
-
never heard about death spiral, for your information i can sustain infinite 9G turn without modding data and without entering into a “death spiral” …if you wonder how, you should leard to read Energy management diagrams. the AI anyway is able to control its G and roll so that it is able to maintain in flight enveloppe.
-
the fuel consomption of AI in WVR is the same as human and yes of course the AI runs out of fuel !
-
the only true information : the AI indeed does not suffer from G lock
set g_bAIGloc 1 // Toggles AI GLoc
Doesn’t work ?
-
-
set g_bAIGloc 1 // Toggles AI GLoc
Doesn’t work ?
I just wanted to ask the same. BTW it is pointless. The G tolarence is so high comparing to original F4.0 or FreeFalcon 5 (which restored the original state) that you never experience even grey vision…
-
-
many wrong information in this thread
- never heard about death spiral, for your information i can sustain infinite 9G turn without modding data and without entering into a “death spiral” …if you wonder how, you should leard to read Energy management diagrams. the AI anyway is able to control its G and roll so that it is able to maintain in flight enveloppe.
No, you can’t. I just ran a series of tests using a block 50 with unmodded data, and at 330 knots (bottom of corner airspeed range), the very best I could pull was 6.2 Gs. At 440, 8.8 Gs is the best I could manage, and it dropped steadily to 6.2 @ 330 kts in a sustained max turn.
Continuing the max turn, I stabilized at 2.7 Gs @ 197 kts maintaining full aft pressure on the stick.
Starting high and fast, as fast as it’ll go, I pulled for max Gs and tried to hold it as long as possible. You can’t do it. Once your airspeed drops to around 450 knots, you can’t hold 9 Gs.
There is no possible way that you can maintain a 9 G turn indefinitely with unmodded flight data. You don’t have the surplus energy to maintain more than 200 knots in a max turn. You would need about 90,000 pounds of thrust to maintain 9 Gs, as I’ve found out in my experiments with the thrust multiplier. (At 3.0, you can stay in corner at max pull, all day long.)
I can’t believe you haven’t heard of the death spiral. It’s one name for what happens if you maintain a hard turn to gain position, starting at altitude, and trading altitude for energy while continuing the turn. You end up chasing your opponent’s tail while he does the same to you, and ultimately you end up with leaves in your intake.
If I’m wrong, prove it. Show me the screenshots and tell me how to maintain 9 Gs indefinitely. Maybe my whole BMS installation is busted.
-
No, you can’t. I just ran a series of tests using a block 50 with unmodded data, and at 330 knots (bottom of corner airspeed range), the very best I could pull was 6.2 Gs. At 440, 8.8 Gs is the best I could manage, and it dropped steadily to 6.2 @ 330 kts in a sustained max turn.
Continuing the max turn, I stabilized at 2.7 Gs @ 197 kts maintaining full aft pressure on the stick.
Starting high and fast, as fast as it’ll go, I pulled for max Gs and tried to hold it as long as possible. You can’t do it. Once your airspeed drops to around 450 knots, you can’t hold 9 Gs.
There is no possible way that you can maintain a 9 G turn indefinitely with unmodded flight data. You don’t have the surplus energy to maintain more than 200 knots in a max turn. You would need about 90,000 pounds of thrust to maintain 9 Gs, as I’ve found out in my experiments with the thrust multiplier. (At 3.0, you can stay in corner at max pull, all day long.)
I can’t believe you haven’t heard of the death spiral. It’s one name for what happens if you maintain a hard turn to gain position, starting at altitude, and trading altitude for energy while continuing the turn. You end up chasing your opponent’s tail while he does the same to you, and ultimately you end up with leaves in your intake.
If I’m wrong, prove it. Show me the screenshots and tell me how to maintain 9 Gs indefinitely. Maybe my whole BMS installation is busted.
you want me to proove you that i can sustain an infinite 9G turn without modded data ?
jeez you really need to learn reading EM charts and apply them
take a block 30 with a DI of 38 and GW of 26 000 Lbs Sea Level
Fly mach 0.9 and pull
-
But anyway, the AI did not pulled infinite 9G or whatever
they follow the exact same rules and flight model as human …
the fact that they dont follow a “death spiral” is just because they know how to keep enough energy to avoid this area.
take an ACMI of your dogfight and you will see that the AI manages its G (still poorly managed …this will be fixed in 3 to 4 weeks)
just be aware that pulling more G’s than the opponent does not necessarly means you will catch it ,it is also a question of turn radius…
and by the way, your testing method is wrong , what you have done is to shift the PS curves but you have not changed at aLL the turning capacity of your aicraft.
i mean for the same speed and alt, you just pull the same G’s than with untweaked model, the only difference is that you dont loose speed.
if you want to change the capacity in turning, fake your CL multiplier instead
-
If I’m wrong, prove it. Show me the screenshots and tell me how to maintain 9 Gs indefinitely. Maybe my whole BMS installation is busted.
My last post was literally the information on how to do just that…
Lets have a look at the same graph again, with an example turn highlighted.
This is the doghouse graph for a block 52 at 5 thousand feet, assuming GW 22klbs, DI 0. Mach 0.8 at 5 thousand feet AMSL is roughly 530 knots I believe, so there’s that.
a 19 degrees per second sustained turning rate is not too bad - and it is at 9G. The thing is, its not indefinite. Eventually, you run out of fuel.
Once your airspeed drops to around 450 knots, you can’t hold 9 Gs.
Below 450 knots, you dont have the speed to pull 9G. The cool thing is, the turn rate stays roughly the same down to that 6G at 330 knots.
Above 450 knots, you have the speed required to pull more than 9G. Im sure you have noticed that you are limited to just 9Gs. You will also have noted from the above diagram that pulling 9G at 530 knots actually gives you a slower turn rate, and a wider turn radius, than pulling 8.8G or even 6.2G, at 440 and 330 knots respectively.
While the aircraft is aerodynamically capable of that mid 20’s degree per second turn rate at say 600 knots, you would need more than 9Gs to achieve that kind of turn rate. And as the FLCS limits your turning performance to ‘just’ 9G, you will note that above 9G, you are unnecessarily limiting yourself.
-
I have confirmed that a light block 30 can hold 9 Gs indefinitely if you start the turn at or above mach 0.9. Interesting.
Since I wasn’t looking at older block types, it’s easy to understand why I would have missed that. I’ve been spending all my time flying block 50 and 52 aircraft.
Actually that brings up a pretty good question: Is there any guide available that defines the relative strengths, weaknesses, and capabilities of each block as compared to the other? I can not assume that a block 52 plus is going to be the best possible choice for every possible mission. But I guess I have made that assumption.
It’d also be good to know if certain weapons systems work on specific blocks but not on others.
Oh, quick note: Pick the block 15 (A model) in dogfight mode, and the model you get looks like a C model. It has the larger tailplanes with clipped corners of a C model. In the interests of accuracy, shouldn’t that be changed to the proper A model tailplanes?
-
Block 52 or 50 can achieve a constant 9G turn , it all depends on mass and fuel quantity.
The best weight to thrust ratio is the block30, it is by far the best for dogfight. Block 42 is the worst.
As Bluewolf pointed out pulling 9g at mach 0.8 is less interesting than pulming 6G at mach 0.5 in term of turn rate and radius.
This is why your test based on max G against Ai is not a good one
Read HFFM manual to see all capabilities of the different blocks. Then you will judge which is the best in each area
-
Oh, quick note: Pick the block 15 (A model) in dogfight mode, and the model you get looks like a C model. It has the larger tailplanes with clipped corners of a C model. In the interests of accuracy, shouldn’t that be changed to the proper A model tailplanes?
The bigger aft horizontal stabs came in on Block 15s so that sounds correct.
-
Correct me if I’m wrong, but I’ve always understood that the revised stabs were introduced with the C model, and not a moment before. They are positive identification of a C model or updated earlier model. So I have always understood it.