Katmandu

We need an official repository with earlier versions, not a link to big Reddit thread where the actual DL link is buried somewhere. Even a simple google drive would suffice for now

Всем у кого НЕТ тензо джойстика, обязательно ставить вот эти настройки (кривые в обычном месте, у меня 20). По умолчанию стоят настройки для тензы и самолет незаметно и непонятно перетягивается и супер легко сливает энергию в вираже. У тензы очень замысловатая ступенька кривой и МЗ, для обыч джоев получается бред

Yep, the omniscient bots are indeed an urban myth in DCS. In your case you've cheated somewhat by spawning in the blind spot, but, although it is a hard thing to do, it is still possible to actually sneak up on the AI: Well, it does not stop climbing. In the test mission, it changes its pitch angle from 35deg to around 25deg and climbs at that pitch angle at around 220kmh IAS. The latter is impossible for the player, I could only maintain around 15deg iirc at that speed and alt. Anyways, most of your graphs also show the default AI Rate of Climb and Acceleration curves to be out of whack relative to the reference and the mod, for the stall to 500kph speeds.

У меня в ДКС в Вин10 и Вин11 зеленоглазый МСФФБ2 ведет себя отлично. Никаких дров и калибровок не делал никогда, plug & play. Твист не пользую, правда. Но менять его на мертвую ручку - никогда. Монстр FFBeast в планах, но он громадный, требует вкручивания в пол и прочее; Моза меньше, но тоже далеко не настольник. Других альтернатив нет просто, ВКБ и остальные пусть сами на свих пружинных палках технологии 19 века летают

I have two versions of DCS installed: 2.9.13.68180 and the most recent one. I have two separate Saved Games/DCS folders, so there is no untoward overlap between versions. I have tested all versions since 2.19.13. 2.9.13.68180 replays work perfect. Every shingle version after 2.9.13.68180 (i.e., every version with mid mission save functionality) results in AI desyncing in replays, even in short single player dogfights. My standard test is a dogfight between player's P-51 vs AI Bf-109. It lasts just a few minutes and always results in AI desyncing and going off script (the player's plane continues to fly like it actually flew). Example track is attached. AI desync.trk

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:

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): 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. See the attachment I'm attaching the test missions as well 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

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

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.

Truly beautiful work! Dogfights vs another Mig and the Sabre of course now make a lot more sense, previously the AI Mig-15 was simply climbing out of trouble even against my PFM Mig-15, my Sabre was altogether reduced to running away and head ons... The mod crashed DCS for me until I removed everything but the SFM changes, maybe @Curly could have a look? Otherwise, the mod is simply supreme, it absolutely transformed my DCS Sabre

He reworked almost all of Mig-15 SFM, way beyond a simple thrust adjustment. Thrust adjustment was my own humble attempt at a mod, this was before I found Curly's thread ( https://forum.dcs.world/topic/351306-sfm-performance-and-inaccuracies/#comment-5459098) Here is comparison of a section of the original lua (red) and its Curly's mod (green):

Оказывается всю дорогу уже был отличный мод СФМ бота Миг-15. С обоснованными правками и графиками. Всячески рекомендую: Тест климба: В бою тоже совсем другой самолет, бот больше не может по сприали уходить вверх от игрока на Миг-15 с одинаковым весом. Класть в D:\Run\DCS World\ (не Saved Games). MiG-15 AI Curly mod.zip

But in our case the leader was ahead - as per my climb test video, the AI starts ahead. So, in his journey towards the hill, he floors the throttle first. I react and floor the throttle second - yes. But, since I was behind him, I was also further away from the hill than he - at the instant that he applied the throttle. By the time I floored the throttle I was approximately the same distance from the hill as he was. (it is also possible to be more precise, the AI starts the climb at 685km/h IAS, so it's a matter of doing the same and maintaining his angle ) Your logic would apply if we were side by side - then yes, my reaction time would delay the throttle application and I would start rolling onto the hill at slower speed that the AI "car" and thus would run out of energy sooner. There is margin for error here, but I could not outclimb the AI to 10km, no matter what profile (speed, angles) I used. PS Anyway, I am super happy with Curly's awesome SFM mod for AI Mig-15, so it's all academic now I only wish I found it much-much earlier...

This is a testable (as well as not at all self-evident!) statement. How about flying this test with two human controlled Mig-15s with both climbing at full power? My wager is on the difference between the leader and the trailing plane being so small as to be negligible. Certainly not the effect of one plane continuing to climb while the other is falling from the sky, like in the Mig-15 AI vs Human climb test video. Moreover, the official document graphs were also taken form real life tests with real life pilots, not robots (i.e., a perfectly flown Mig-15 with no human oscillations would have produced better results that what we see in the official graphs). PS A link to a well researched SFM mod of the AI Mig-15 also deserves a mention in every thread where it's mentioned

Lua file tables provide only the input to the flight model equations. For example, there is are no table values like "climb at 25m/s at pitch angle 45Deg". The tables provide Thrust 20000 at Alt 5000, wing area constant, drag and lift coefs for specific Mach, and so on. The actual aircraft behavior -like the climb rate at some particular altitude, speed, attitude and angle of attack- is not directly derived from the tables, it is computed using ED's SFM model. And, because the SFM is more simplistic than PFM, the 2 models predict different behavior. Jarringly different - in the case of Mig-15. One way to address the problem (outside of waiting for ED's GFM that is X years away): 1) Assume that the PFM model is the much more realistic model of the two - because it actually models fluid dynamics etc 2) Tweak the SFM input tables in the lua for closer match in behavior in controlled tests. Yes, the input would deviate from the real life table data, but the actual SFM model behavior would be closer to a more detailed PFM model.