Mig-15? You're kidding me...

[Zychon]

We need human opponents .

 

There is no way around it. To do it right we need a flyable Mig15bis.

 

The AI doesn't know its planes strengths vs your planes weaknesses. I have played many missions now and the AI always dives down to the deck and trys to turn fight me. As long as I keep my speed above .7 mach I win every time till i run out of ammo.

 

We really need a human opponent to see what the mig can really do.

 

I was dying for the Sabre, now I'm dying for the MiG. ...I think the 15 is my favorite MiG. :D

 

Yeah, I think it's the same AI in the seat regardless of the plane. At least it seems that way. We can say we want that or this from the AI, but I personally have no idea what the cost is in terms of processor. At the end of the day, though; I can't think of any AI that I went up against in any game that you couldn't maneuver into a predictable pattern. It's just a matter of time figuring it out. If they didn't have sniper vision and super E management, you could just kick 'em over like cardboard cutouts.

 

To have these two A/C which were so alike, but with significant differences that change the dimension of the fight. ...in MPlayer?! Yes, please.

[ralgha]

We can say we want that or this from the AI, but I personally have no idea what the cost is in terms of processor.

 

The bottleneck with AI is almost never processing power. It's spending the development time to identify and solve the various problems that need to be solved. AI is very tricky and can usually only be created by humans, who invariably take a long time to figure out solutions. Any claim similar to "if we just had x times more processing power..." should be viewed with great suspicion at this point.

[Zychon]

Zychon ,one track would replace a thousand words :)))) thanks)

 

Here's one from this morning. This is a really protracted fight because of all the mistakes I made. All of you born killers out there will probably spot a ton of other mistakes too, but that goes to prove my point. I'm not really terrific at this and I can beat the AI MiG-15 on Excellent ...eventually anyway. :D They can't climb forever, nor are they bullet-proof. So if you are having trouble, maybe something in this clip will help. My first thought it that you are constantly chasing the pipper instead of conserving E, only because that's how I used to fly.

 

http://youtu.be/cVliS8lTITA

Edited August 13, 2014 by Zychon

[mjmorrow]

I was dying for the Sabre, now I'm dying for the MiG. ...I think the 15 is my favorite MiG...

 

To have these two A/C which were so alike, but with significant differences that change the dimension of the fight. ...in MPlayer?! Yes, please.

 

I really like the F-86F a lot. The F-8F is definitely my favorite current DCS ride, by far. I could see myself really falling for the Mig-15bis, too. I like that both machines are highly competitive with one another, yet they are not entirely the same. The Mig-15bis is going to bring something new to the table. She is a bomber interceptor, best adapted to climb particularly fast and bring down large multi-engine strategic bombers.

 

When facing the Sabre, I am going to enjoy trying to leverage the advantages and mitigate the challenges of a machine best suited for downing B-29s. I don't know if I will like the Mig-15bis more than the F-86F, but I bet I will like the Mig-15bis, at least as much as I like the F-86F. I too can't wait to get my simulated hands on a Mig-15bis. :thumbup:

Edited August 13, 2014 by mjmorrow

[ivo]

Hi to all, first I'm not a very good pilot, but since last patch 1.12.10 I Always shot down Mig-15 in average and good AI before the last patch I never shot down....then I love this patch, however in excellent Ai for my is impossible...

 

Bye

[Basco1]

I was dying for the Sabre, now I'm dying for the MiG. ...I think the 15 is my favorite MiG. :D

 

Yeah, I think it's the same AI in the seat regardless of the plane. At least it seems that way. We can say we want that or this from the AI, but I personally have no idea what the cost is in terms of processor. At the end of the day, though; I can't think of any AI that I went up against in any game that you couldn't maneuver into a predictable pattern. It's just a matter of time figuring it out. If they didn't have sniper vision and super E management, you could just kick 'em over like cardboard cutouts.

 

To have these two A/C which were so alike, but with significant differences that change the dimension of the fight. ...in MPlayer?! Yes, please.

 

Yup,totally agree,I would purchase a flyable MiG-15Bis for sure,DCS take my money now :thumbup:

[kontiuka]

Yup,totally agree,I would purchase a flyable MiG-15Bis for sure,DCS take my money now :thumbup:

Has this actually been mentioned anywhere that the Mig-15 could possibly be made a flyable module?

[Highwayman-Ed]

I've been getting gradually better and better against the MiG starting with with a 'not a cat in hells chance' to a being able to regularly out manoeuvre a MiG on my tail and get the upper hand now, especially with the lower AI levels since 1.2.10.

 

I do however manage to kill the MiG easier in the Dora than in the Sabre though!

Edited August 14, 2014 by Highwayman-Ed

[Basco1]

I've been getting gradually better and better against the MiG starting with with a 'not a cat in hells chance' to a being able to regularly out manoeuvre a MiG on my tail and get the upper hand now, especially with the lower AI levels since 1.2.10.

 

I do however manage to kill the MiG easier in the Dora than in the Sabre though!

 

Hehe,me too,I'm a far better shot in the P-51 :lol:

[Dr_Arrow]

ED has a very difficult task to design a good AI for such a variety of combat aircraft. I really don't know if it is possible or feasible to design special tactics for every aircraft against every aircraft, e.g. you need completely different tactics in Mig-15 against P-15 or Dora and completely different tactics against F-15 or Su-27 or Mig-21. I don't think it will work in close future :D

[shagrat]

Why not? Combat flight maneuvers are a set of standardized actions. Variables can define the performance envelope of different planes for these maneuvers. Now you may call a set of preferred maneuvers against specific planes or plane types quickly...

In essence individual AI flies standard maneuvers, but a separate logic "calls" the actual maneuvers against specific adversaries.

 

What do you think?

[Zychon]

Hehe,me too,I'm a far better shot in the P-51 :lol:

 

The rudder in the F-86 is the thing getting me. I'm used to "cheating" my shots in the P-51. The rudder in the Sabre doesn't have the same authority.

[Zychon]

Why not? Combat flight maneuvers are a set of standardized actions. Variables can define the performance envelope of different planes for these maneuvers. Now you may call a set of preferred maneuvers against specific planes or plane types quickly...

In essence individual AI flies standard maneuvers, but a separate logic "calls" the actual maneuvers against specific adversaries.

 

What do you think?

 

Sounds great to me as a concept, only I don't think it will be that straightforward (if I understand you correctly -not a programmer here). The choice of a maneuver is much more dependent on the relative positions of the combatants than the performance envelopes. Also, the essence of the dogfight is pretty much pilot "a" trying to trick his opponent into a position where he thinks he should do "x", and then closing the door on him after he has committed, instead putting him somewhere where "a" can leverage all of his advantages while nullifying the enemy's.

 

Of course this is obvious, but think about it. The choice of a maneuver wouldn't be simply picking from a list, but a careful "filtering out" of existing choices (angles, paths, etc). These filters would have to slide based on immediate positional data. How often would this check be made? Constantly? If so, then the AI would be essentially thinking zero maneuvers ahead. How do you avoid it falling for the same old trick every time?

 

At least from a unenlightened "game user's" perspective, programming AI seems to be a very tenuous job. Every choice you give it seems to be be another place where it could simply chose to flop around in the corner attempting to gnaw off its own tongue like some half finished homunculus. Adding these calls may not be making it smarter, but simply demanding it be smarter than it actually is.

 

The idea seems like a good starting point. Just playing Devil's advocate here.

[shagrat]

You are absolutely right, but I wasn't suggesting a simple random pick from said list. Think of it as a list of options!

The Combat AI "knows" how to fight the plane (calculates his dogfight with a list of possibly advantageous maneuvers). The "list" just provides the parameters, about the "best" maneuvers against an opponent type / model, about the AI's plane flying characteristics (more or less the SFM), and a skill setting specific "typical error list"... e.g. how likely a rookie pilot chose the second best maneuver to counter an opponents move, a margin of error when not turning in a perfect turn etc.

Basically it's just about separating the Pilot's knowledge about his current plane and the one his opponent is flying, from his actual fighting skills as a combat pilot.

[Zychon]

So it's a matter of prioritizing then? More weight is given to maneuvers that are on the "match up list"?

 

I took a break from fighting the MiG-15 yesterday and switched to fighting against Sabres. It gave me a little more insight into the AI. Going up against the MiG has gotten me into the habit of hanging on to every drop of E like it was my last, and going up against the Sabre like that revealed ...wow! The AI is just god awful at vertical fighting in the Sabre! I had to snap the brakes almost every time. A few times I watched him literally fall out of the sky sideways.

 

If I go up against the excellent AI in Mustang v. Mustang, I never see this happen. So apparently the AI does have to be "taught" (as in specific instructions) to fly a given A/C, and this part of the project is not yet complete. I always assumed it just used a portion the the FM data.

 

Got me interested now. I guess I have some reading to do.

[Basco1]

Re the MiG......I am really looking forward to seeing a few more skins,but am not sure if this is going to be possible without the template.

[studUK]

Plain ridiculous MiG15 AI

 

 

very entertaining though

Edited August 19, 2014 by studUK
spelling

[SgtPappy]

Happened to me a lot too, but you can win! I believe in ya! Nice video though. It perfectly sums up my first 100,000 fights.

[Foul Ole Ron]

The Mig is beatable - mainly want to concentrate on not following him up unless you're going at least 320 knots. But there seems to be some screwy things going on with his energy - I've attached a Tacview track which shows it.

 

At 9.02.51 he's at 3492m and going 524.9kmh. By 9.03.04 he's after diving from 3492m down to 2338m while actually managing to lose speed and be going 515.7kmh. I tried to replicate this in the Sabre extending out speedbrakes and cutting thrust to zero and came nowhere close.

 

There's other little oddities that you come across when reviewing in Tacview e.g. the Mig very rarely goes over Mach 0.5 no matter what for some reason. But I guess these are to be expected with a very early release model that was probably rushed out to make the Sabre release date. I'm sure it'll improve over time.

Tacview-20140821-145422.txt.rar

[otto]

F-86 Kills Vs Excellent Ai Mig-15

 

[shagrat]

Well it does not matter if human pilots can beat the AI Mig-15... it seems it is getting more and more obvious, that the AI is not flying the Mig in a very realistic way. Especially if mission building, apart from plain human vs AI is considered.

I would like to have this fixed before learning dogfighting the Mig-15... imagine AI replaced by human pilots and suddenly all maneuvers the Mig-15 AI was flying, well are pretty much different?

And as of now, there is no Sabre AI that can actually beat the Mig, even on average setting. :doh:

[SgtPappy]

Well it does not matter if human pilots can beat the AI Mig-15... it seems it is getting more and more obvious, that the AI is not flying the Mig in a very realistic way. Especially if mission building, apart from plain human vs AI is considered.

I would like to have this fixed before learning dogfighting the Mig-15... imagine AI replaced by human pilots and suddenly all maneuvers the Mig-15 AI was flying, well are pretty much different?

And as of now, there is no Sabre AI that can actually beat the Mig, even on average setting. :doh:

 

Although you're correct, it does matter that we can beat MiGs because it means we've all learned something, primarily about energy fighting. When many of us started, we'd blow E, stall or just run out of gas. Now we've learned to pull low G's and actually go up with the MiG.

 

A more realistic MiG will require different strategies, but there's certainly no harm in learning E-fighting like the current MiG has taught us.

[shagrat]

Sorry, you're right I should have worded that differently. I was referring to the AI's obvious flaws, only.

You are right it trains anyway, but it also does fly in a pretty unreal way, thanks to SFM?

Edit: and the balancing definitely needs to be optimized! It is impossible for mission designers to use a Mig this way as an opponent in an AI vs. AI fight!

 

Not that Belsimtek thinks it doesn't need tweaking, cause it is beatable... :D

Edited August 22, 2014 by shagrat

[Robo.]

F-86 Kills Vs Excellent Ai Mig-15

 

 

Very nice, that's the way to do it. :thumbup:

[grifo]

I would like to spend my 2 cents about that:

I'm not a great dogfighter, and I was going insane trying to beat the 15.

I'm aware that is still WIP, but if Yo-Yo stated that AI is working properly.. well, this is what we have to get down, also if it seems to fly like an UFO as someone stated before and established with useful tracks. I'm also thinking there's something wrong with his energy management, but I never flown a Mig-15.

Anyway.. the point is that after spending hours trying to beat him with poor results at high/excellent level, I decided to try with labels on.

Well.. I beat him every time within 10 minutes and taking no damage as well.

If I switch it off, My success is 1 of 6

As I said, I'm not a great dogfighter, but this made the difference, and I have a 24" at 60cm from me and T-IR.

So I think the big deal with this issue is the poor visibility and the difficulty to visually spot and track targets in DCS. I don't know if targets visibility is scaled correctly, but IMHO it needs MAJOR improvements in order to gain playability.. And with the upcoming of the new WWII modules this will be essential.

I will be glad if someone who has problem with the AI guy would make a try with labels on and let me know if it's just my impression or can be something to think about.:pilotfly: