[Graf]

"I think that, in most's opinion, if 3.04AI is working as a mod, i shouldn't see successions of barrel rolls from my intended victim, am i right ? (if i'm wrong on this assumption, my test becomes worthless).
"

Exactly. Instead your intended victim will climb...dive...turn...or try to head home. Now I have seen it roll maybe once or twice but thats it. I am testing the AI in offline campaign mode and I definitely see a difference.

Am currently using DCG flying the 109F-2 at the start of Barbarossa. Have it setup that my flight is always outnumbered(2-1 usually) but by of a lesser quality opponent. In several engagements my flight usually downs all enemy a/c and at most loses one 109.

[HolyGrail]

:wink:

[lowfighter]

Hey guys I think it's easy to test if the mod is doing any difference. Set a pure AI fight mission. Play it and record it with mod. Play it and record it without mod. If you can see ANY (might be big, might be small) difference at a given time in the two records then you have the mod doing something
Edit:just saw what Bee wrote, about the same ...

[Klemm]

My best understanding is that those .class files that were provided have got to be hashed first, or they won't be read by the game at all. This can be done with the class hasher by qTim.
I did just that yesterday (the files then go just into mods/aimod/ as the hash name is the encrypted path and name) and lo and behold, like another poster already posted, the game crashed as soon enemy ai got into dotrange. I could watch them from externals, but my game crashed three times trying to confirm this bug.
A reasonable explanation would be that as soon the ai code is activated, due to incompatibilities with some other class files of the game, it just crashed to desktop.
So in the end those files would need to be compared (current and old ones) and the desired changes implemented accordingly. But for that one would need pretty good knowledge of the internal structure game and Java, both of which I unfortunately don't have.
Maybe someone who can do this will surface, but i'm afraid not. :? :o

[vampire_pilot]

so, here is my impression, i added the patch while flying a campaign, remamed the folder for the files to be read first and I do have the strong impression the AI got better, waaaay more aggressive and also ... just overall better.

I achieve less kills than I did before and get hit more often.

this of course is just an impression, no prove but I feel more challengened since the aplication of the patch and also i think I have not experienced the constant roll maneuvre since. one or two rolls, a dive, a turn, a roll...

[poncho]

I still tend to agree with Klemm,

I tried using the hashed 3.04ai from the Mods folder which crashes the game. Using the hashed new ai from the Mods folder ran normally. I suppose its possible that you have found some back channel way of getting the game to read non-hashed class files. I doubt it, but only the guys that wrote the game code would know for sure.

I did try to compare the files in 3.04 ai vs new ai. The 3.04ai has 19 files, the new ai has 21. The code within the common files was identical in about half of them.

One difference I noticed between 3.04 and the new ai was that special behavior was coded for about 5 of the newer aircraft(in the new ai but not in 3.04). The Ta-152 was one of these, I forget the others. I'm not sure what this behavioral difference is but it seems like the Ta-152 ai behavior should be noticably different if the 3.04ai is actually being read.

[lowfighter]

Just did the test with a flight of Bf109G2 against a flight of P40M. No difference and same outcome (2 P40M destroyed and one Bf109 smoked a bit). So now what poncho said is interesting, maybe also it might depend on what type of aircraft, maybe I used aircraft types for which the code is the same in 3. and 4. ?

[Catahoulak9]

Is there any way to definitively test whether the new files are even being read? What about changing some of the figures in the modded files that would cause unmistakable behavior just to see if they are being loaded.

[fly_zo]

guys, you need to test it with some "complete" mission .... takeoffs ...inflight commands.... formation ( +escort) .... when interrupted by fighters .... ground pounding .... landings

you'll see little to nothing in QMB dogfight mission ....


just a thought

Z

[Guest]

You are very right, i just installed using bee's explanation and there is definitively something changed.

I have to explore it further before saying i like it, because i'm concerned on how the energy fighters will work, but i confirm that with the "AI" folder side to the "object" folder there is a change in AI behaviour, and they seem more agressive (even at lower levels).

I start to understand better what your request was.

I'm very happy to read about this m8
Now the top will be if someone will find the code for managing those .class files as I think klemm or poncho are right in their guessing and if possible make a mix of good thing both in old A.I. + new A.I. :roll:

[Bo_Nidle]

I still think the difference is there. I just set up a 4 on 4 dogfight over the coral sea at 3000m between Claudes and F2's.

1/With the mod they go head on and then turn and avoid each other as they merge followed by a twisting and turning fight.

2/Without the mod both opposing lead aircraft collide with each other as they merge with no attempt at avoiding action.

3/Mod activated again and scenario 1 is repeated.

The mod stays for me.

[Graf]

"...you'll see little to nothing in QMB dogfight mission .... ..."

Exactly.

Most people not seeing any differences with the mod are using the QMB as a test bed. Try some offline missions and fly a complete mission.

[lowfighter]

Bo_Nidle, please paste the text of the mission here so that I can try replicate. Maybe it's really dependent on what AC we use. I used old il2FB AC (Bf109 and P40M) and got no difference

I still think the difference is there. I just set up a 4 on 4 dogfight over the coral sea at 3000m between Claudes and F2's.

1/With the mod they go head on and then turn and avoid each other as they merge followed by a twisting and turning fight.

2/Without the mod both opposing lead aircraft collide with each other as they merge with no attempt at avoiding action.

3/Mod activated again and scenario 1 is repeated.

The mod stays for me.

[reflected]

OK, I tested it (in QMB though)

-There is a difference! It's definetly not a placebo effect, the AI is much more creative now! wow!

-however it still does 2-3 barrel rolls sometimes...but not very often.

-flak seems to be much more accurate (?)

-When a lot of AI is "thinking" at the same tme, my high-end PC lags for a split second....


Anyway, I'm so glad to have this MOD

[Xoddam]

[quote="kowi76"]Hi,
just a little mod for anyone who is