Is AI cheating, or am I just bad at dogfighting? (Or both)

[MAXsenna]

I wonder if adding weight would help to tame the MiG-21 a bit.....

Some DLC campaign creators do exactly this.

I'm midly curious about @Katmandu 's lua edits. Wouldn't it be wonderful if those that have some knowledge could edit this more realisticly, and then we could petition ED to implement them until we get the GFM?

Sent from my SM-A536B using Tapatalk

[diego999]

The AI MiG-21 is a bad joke of an UFO capable of 9G sustained turns with no speed loss. It would give a F-22 a run for his money.

We've tested and reported it, answer is always the same "wait for GFM" for years now, like dynamic campaign, Vulkan, you get my drift.

At least tweak it a little so it's not so absurd, but no, too much effort.

[Katmandu]

The glass is half full, people  Yes, there are issues with some(!) AI SFMs, but isn't it cool that we have access to make changes as we see fit?

Here is a table of DCS aerodynamic parameters that are used in SFM (on top of the power output per Mach band tables - those are self explanatory) from page 30 here https://forum.dcs.world/topic/97337-beginners-guide-to-dcs-world-aircraft-mods/.

The location of the SFM lua file is "DCS World OpenBeta\CoreMods\aircraft\MiG-15bis\MiG-15bis.lua" and so on for other aircraft. There just two small tables to edit in the file

To conclude, the

12 hours ago, MAXsenna said:

Wouldn't it be wonderful if those that have some knowledge could edit this more realistically

part is more accessible than it may initially appear. We load up the AI Climb Test/AI Turn Test missions with our jet having identical fuel, weapon and speed settings - missions link: https://forum.dcs.world/topic/326798-ai-turn-and-climb-performance-impossible-for-a-human-aircraft/#comment-5226532

We try our best to keep up with the AI. Then exit the game, tweak the parameter that we feel is out of whack, load back in, test again. Rinse and repeat until there is a good match with our PFM plane and its AI twin. The world is our oyster  Or, we could wait for ED, the second coming and the heat death of the universe

Edited June 13 by Katmandu

[MiG21bisFishbedL]

On 6/12/2025 at 7:35 AM, Hiob said:

I wonder if adding weight would help to tame the MiG-21 a bit.....

It absolutely does. It isn't perfect, but it's a temporary fix.

1. Put down a MiG-21 and name it. Let's use 'Fishbed' as the name.
2. Under the 'Triggers' box, make a new one. Set it to 'MISSION START" for event.
3. Under the 'Actions' box, make a new one, Set it to "DO SCRIPT'
4. Put this in the DO SCRIPT text box below the 'Actions' box: trigger.action.setUnitInternalCargo('Fishbed', 1600)

This adds an additional 1600kg of weight to the Fishbed. I've seen most recommend 2000, but I find that really makes the Fishbed lazy. Play with it to ascertain your own preference, but the sweet spot is between 1600kg and 2000kg.

You'll have to do this for every Fishbed in a mission, but it beats trying to fight the communist Magic School Bus.

[Hiob]

48 minutes ago, MiG21bisFishbedL said:

It absolutely does. It isn't perfect, but it's a temporary fix.

1. Put down a MiG-21 and name it. Let's use 'Fishbed' as the name.
2. Under the 'Triggers' box, make a new one. Set it to 'MISSION START" for event.
3. Under the 'Actions' box, make a new one, Set it to "DO SCRIPT'
4. Put this in the DO SCRIPT text box below the 'Actions' box: trigger.action.setUnitInternalCargo('Fishbed', 1600)

This adds an additional 1600kg of weight to the Fishbed. I've seen most recommend 2000, but I find that really makes the Fishbed lazy. Play with it to ascertain your own preference, but the sweet spot is between 1600kg and 2000kg.

You'll have to do this for every Fishbed in a mission, but it beats trying to fight the communist Magic School Bus.

Thanks for this!
This is awesome specific advice!

I will try that.

[Katmandu]

Adding weight via a script limits this workaround to missions which you edited yourself. The same result could be achieved by nerfing its thrust/weight ratio via dropping the thrust numbers across all Mach bands by the same percentage. This would automatically carry over to ALL missions and campaigns. 

It would also be good to validate the change against player controlled PFM plane with tests like the above, otherwise over-nerfing is like putting AI on easy. If one wants simulation of course.

Edited June 13 by Katmandu

[Hiob]

Well, it’s nice to have options. For a quick sortie, the script approach is certainly easier.

However for a real mission build you’re right of course.

i’m soaking all of this up, because so far, I didn’t really mess with either and I find the possibilities really interesting! 

[Kang]

I'd say the worst offender to me is the MiG-15 actually. To the point that dogfighting it in the Sabre - mind you, pretty much the only 'pairing' that ED ever really released - is so little fun that I had taken to tangling with more modern enemies instead during my Sabre era.

The problem of the simplified flight models of the AI being rather inaccurate in these regards obviously becomes more and more visible when you are actually dogfighting. DCS seems to much prefer BVR engagements of modern jets in which this doesn't matter much. But what is worse is that ED in all their combined wisdom have decided to include AI that pretty much exclusively and constantly sticks to manoeuvres and tactics that exploit/showcase this weakness of the simulation to the max.

[MiG21bisFishbedL]

55 minutes ago, Kang said:

I'd say the worst offender to me is the MiG-15 actually. To the point that dogfighting it in the Sabre - mind you, pretty much the only 'pairing' that ED ever really released - is so little fun that I had taken to tangling with more modern enemies instead during my Sabre era.

 

I actively recommend against buying either, just because they're without peers and have so many longstanding and unaddressed issues. They are not worth a purchase.

Edited June 13 by MiG21bisFishbedL

[Katmandu]

19 minutes ago, Lidozin said:

 

Nice video! So, at pitch angle below 20 deg and above 500km/h the SFM and PFM have good correspondence - this is cool and finally some good news for Mig -15 SFM haha!

But!  In my vid, the pitch angle was about 35 deg, so both planes were losing speed, all the way to around 200km/h - and the AI had big advantage in climb rate. This kind of zoom climb is actually more reflective of actual combat, as planes do not do shallow climbs when fighting.

So your vid does not seem to disprove anything with regards to combat/steep climbs for AI, but does prove that SFM and PFM do have good correspondence for shallow climbs.

Edited July 16 by Katmandu

[Lidozin]

10 hours ago, Katmandu said:

Nice video! So, at pitch angle below 20 deg and above 500km/h the SFM and PFM have good correspondence - this is cool and finally some good news for Mig -15 SFM haha!

But!  In my vid, the pitch angle was about 35 deg, so both planes were losing speed, all the way to around 200km/h - and the AI had big advantage in climb rate. This kind of zoom climb is actually more reflective of actual combat, as planes do not do shallow climbs when fighting.

So your vid does not seem to disprove anything with regards to combat/steep climbs for AI, but does prove that SFM and PFM do have good correspondence for shallow climbs.

Thank you — I agree it's good to see that SFM and PFM align well under stable conditions. However, zoom climbs of the type shown in your video are difficult to verify without detailed data. The results depend heavily on maintaining the same speed-energy profile and minimizing oscillations or excess control input.

To make conclusive statements about energy performance in steep climbs, a TacView recording or a comparable export of time history for TAS, altitude, and G-load would be ideal — for both aircraft. That would allow direct comparison of energy rates and drag profiles.

Even if we assume that the thrust curve in the low-speed regime was deliberately adjusted for some internal purpose (though I’d argue that it's not simply a "no-loss" curve, since the shape doesn't fully match that either), the difference in equivalent vertical velocity at the worst point (IAS ~250 km/h) does not exceed 8–9%. Given that the AI spends almost no time at those speeds, the contribution to its total energy gain is negligible. At higher speeds, the difference becomes virtually zero.

So, even replacing the base thrust table with one that strictly matches the theoretical curve would not result in any substantial change to the outcome of dogfights against the AI.

[Katmandu]

4 hours ago, Lidozin said:

To make conclusive statements about energy performance in steep climbs, a TacView recording or a comparable export of time history for TAS, altitude, and G-load would be ideal — for both aircraft.

Tacview is very inaccurate, with error margin up to 100%, ED don't even accept it as evidence in threads about FM. Standard replay taskbar readings are sufficient for ED, and therefore for us also.

 

4 hours ago, Lidozin said:

Even if we assume that the thrust curve in the low-speed regime was deliberately adjusted for some internal purpose

Climb rate at high pitch angle and angle of attack is not JUST about the thrust curve, it's also very much about the airframe- the lift and drag coefficients, Cymax stall coefficient and so on. Which is why I love Curly's mod of the AI Mig-15 SFM so much, he's changed most of the aerodynamics parameters in Mig-15's SFM, not just the thrust at low Mach values (like I did - before discovering his mod).

Here is his mod in action, the 35deg zoom climb rate is very similar for both player and AI. I've recorded the climb to 7K as the AI changes to shallower pitch angle after this. The true zoom climb is to 6900m or there about.

PS Here is a link to Curly's AI Mig-15 SFM Mod for people who might want to try it 

 

Edited July 17 by Katmandu

[Dragon1-1]

1 hour ago, Katmandu said:

Tacview is very inaccurate, with error margin up to 100%, ED don't even accept it as evidence in threads about FM. Standard replay taskbar readings are sufficient for ED, and therefore for us also.

I think ED actually has some debugging tools to analyze tracks, in addition to what the sim normally provides. They might be able to pull better data than we do with them.

[Lidozin]

On 6/13/2025 at 8:45 PM, Kang said:

I'd say the worst offender to me is the MiG-15 actually. To the point that dogfighting it in the Sabre - mind you, pretty much the only 'pairing' that ED ever really released - is so little fun that I had taken to tangling with more modern enemies instead during my Sabre era.

The problem of the simplified flight models of the AI being rather inaccurate in these regards obviously becomes more and more visible when you are actually dogfighting. DCS seems to much prefer BVR engagements of modern jets in which this doesn't matter much. But what is worse is that ED in all their combined wisdom have decided to include AI that pretty much exclusively and constantly sticks to manoeuvres and tactics that exploit/showcase this weakness of the simulation to the max.

 

22 minutes ago, Dragon1-1 said:

I think ED actually has some debugging tools to analyze tracks, in addition to what the sim normally provides. They might be able to pull better data than we do with them.

Does this refer to online TacView sessions, or are there known cases of such 100% mismatch in offline missions as well?

[Lidozin]

2 hours ago, Katmandu said:

Tacview is very inaccurate, with error margin up to 100%, ED don't even accept it as evidence in threads about FM. Standard replay taskbar readings are sufficient for ED, and therefore for us also.

 

Climb rate at high pitch angle and angle of attack is not JUST about the thrust curve, it's also very much about the airframe- the lift and drag coefficients, Cymax stall coefficient and so on. Which is why I love Curly's mod of the AI Mig-15 SFM so much, he's changed most of the aerodynamics parameters in Mig-15's SFM, not just the thrust at low Mach values (like I did - before discovering his mod).

Here is his mod in action, the 35deg zoom climb rate is very similar for both player and AI. I've recorded the climb to 7K as the AI changes to shallower pitch angle after this. The true zoom climb is to 6900m or there about.

PS Here is a link to Curly's AI Mig-15 SFM Mod for people who might want to try it 

 

This is quite interesting, but judging by the graphs posted by Curly, the aircraft's aerodynamics (i.e., the polars), which are solely responsible for energy performance along with thrust, were barely changed in the low-to-mid Mach number region where your test took place.

Therefore, in our case, thrust is essentially the dominant factor in energy gain.

The maximum lift coefficient, unfortunately, has the opposite effect on energy: if you reduce it for the AI — as was done in the mod based on the reference document — the AI will actually preserve energy better than the default version.

Unfortunately, I haven't been able to locate the mod file to try it myself. Would you be willing to share it?

If you still have the track from your test, it would be quite valuable to see video of both runs — one using the default data file, and the other using the modified one. In that case, your piloting should remain the same, and the AI’s trajectory should presumably change.

And, by the way, how can I make the AI perform such a maneuver in a mission?

[Katmandu]

16 hours ago, Lidozin said:

Does this refer to online TacView sessions, or are there known cases of such 100% mismatch in offline missions as well?

Tacview is unreliable for offline as well. In fact, the error margin exceeds 100% and is morel like 400%. E.g., 6g in game was reported as 1.5g in Tacview(use google translate): 

 

15 hours ago, Lidozin said:

judging by the graphs posted by Curly, the aircraft's aerodynamics (i.e., the polars), which are solely responsible for energy performance along with thrust, were barely changed in the low-to-mid Mach number region where your test took place.

I can only say that these small changes made a big difference  Even beyond the simple climb tests, in actual combat, the difference is even more apparent - AI can no longer rely on spiral climbs against player's Mig-15 with equal fuel load.

15 hours ago, Lidozin said:

Unfortunately, I haven't been able to locate the mod file to try it myself. Would you be willing to share it?

See the attachment

 

15 hours ago, Lidozin said:

And, by the way, how can I make the AI perform such a maneuver in a mission?

I'm attaching the test missions as well

15 hours ago, Lidozin said:

it would be quite valuable to see video of both runs — one using the default data file, and the other using the modified one. In that case, your piloting should remain the same, and the AI’s trajectory should presumably change.

I've posted videos of my (admittedly imperfect climb tests)

Default:

Curly's mod :

 

MiG-15 AI Curly mod.zip AI_test_climb_Mig15.miz

Edited July 18 by Katmandu

[Lidozin]

Thanks for the materials!
Small modifications to the aerodynamic polars — especially within the low-to-mid Mach range — have minimal impact on overall energy performance compared to relatively larger changes in thrust. That said, it’s reasonable to acknowledge that changes in thrust, and thus in the bot’s energy potential, can under certain conditions influence its combat logic or maneuvering behavior.

[Lidozin]

As far as I understand, the forum thread was discussing a recording from an online session, since the aircraft names match the nicknames of forum participants. Are there any references showing such large TacView discrepancies occurring exclusively in offline missions?

[Katmandu]

16 hours ago, Lidozin said:

As far as I understand, the forum thread was discussing a recording from an online session, since the aircraft names match the nicknames of forum participants. Are there any references showing such large TacView discrepancies occurring exclusively in offline missions?

They were testing aim-120 missile avoidance maneuvers in single player, high-g barrel roll to be precise (AI shoots missile, player avoids). Tacview was reporting around 1g when in-game readings were 6g+. Some player (Blackfyre) presented Tacview readings as proof, this was rejected by ED team's Chizh and it was subsequently checked by the player himself who concluded that Tacview's readings were indeed very unreliable.

I mean this post:

Quote

 

It seems more complicated here. I left the F-16 from 120 with an overload of 6 with kopecks G (data from avionics and external DCS, takview shows 1 and less G), but the same did not happen on the F-18 (for now, but it worked a long time ago). The difference in the roll rate, which apparently results in acceleration (not overload, but the rate of change of the target's velocity vector). So far I can not understand where and what went wrong.

At the same time, I am convinced that at least with some probability, but the missile should hit such targets, otherwise the same F-16 and the Su-27 family would be practically immortal. And we know for sure that both were shot down, yes, by air defense missiles, but they are even less maneuverable than 120 and lose energy easier (probably).

Track on 6g in the attachment (single player, short).

Тут сложнее кажется. Я ушел на F-16 от 120ки с перегрузкой 6 с копейками G(данные из авионики и внешнего вида DCS, таквью при этом показывает 1 и меньше G), но то же самое не вышло на F-18(пока что, но когда-то давно получалось). Разница в скорости крена, которая видимо выливается в ускорение (не перегрузку, а именно скорость изменения вектора скорости цели). Пока не могу понять, где и что пошло не так.

При этом, я убежден, что хотя бы с какой-то вероятностью, но ракета должна попадать в такие цели, иначе те же F-16 и семейство Су-27 были бы практически бессмертные. А мы достоверно знаем, что и тех и других сбивали, да ракетами ПВО, но они еще менее маневренные чем 120 и проще теряют энергию (наверное).

Трэк на 6g во вложении (сингл, короткий).

 

 

 

Edited July 19 by Katmandu

[Lidozin]