What's with LMG in AI's hand these day? Or rather AI accuracy in general?

[mmm 35]

Edit:Actually how about a challenge: beat AI MG team with a pistol. Tips including normal use of cover and simply going prone.

 

Plainly put, my observation is that they can't hit the side of a barn with LMG, almost literally.

 

Originally I was doing some really simple experiment in editor with regards to AI engagement range. But the time I moved to LMG type weapons is when I notice something's off. Even the first shot within a burst generally misses the target by a mile, something that doesn't appear on other weapons. Subsequent shots, as one might expect has next to no chance to hit. As the distance draw closer, the size of burst increases, ensuring the fire rate of aimed shots remain low. A behavior I can only describe as deliberately trying not to hit anything while wasting as much ammo as possible. All in all LMG used by AI amounts to not much more than a prop for atmosphere.

 

On the other end of the spectrum you have DMR/sniper rifle of various flavors. One trait they share when used by AI is that they're unconditionally more lethal than any other weapons under practically any condition. It also seems stance no longer have any meaningful effect on AI accuracy. I have always accepted AI gaining more accuracy and fire more frequently in prone as a kind of abstraction of the resting/deploy weapons mechanics AI can not take advantage of. It's a bit immersion breaking to see AI snipe with the most unwieldy weapon in unsupported standing stance from half a mile away, and actually hit the target first shot.

 

I'm not sure what happened in the last year or so when I was not very actively following the development. Was this change documented somewhere? The nature of the changes? And its stated purpose?

[pierremgi 4890]

Yes, that's absolutely weird and ridiculous!

I just tested in Arma vanilla 1.78, one player BLUFOR, one AI OPFOR Autorifleman with Zafir (LMG) , facing on Tanoa runway, 120 m distant.

 

BI dev, please just test that!

 

[mmm 35]

Higher caliber machine guns are the worst offender it seems. You can have rifle teams wipe out machinegun squad twice the size easily at most ranges. MX SW fares slightly better. For Type 95 LSW and SPAR-16S they're just assault rifles with deeper magazine.

 

I don't know when this happened but I very much prefer when LMG was a serious threat to contend with when AI drops to prone. Sure in someway it was so lethal it overshadows other weapons in AI's disposal. But at least that corresponds to its efficacy in players hand very well. This is one time I wish I could revert to earlier version, with the official campaign crash and all.

[pierremgi 4890]

 

[lexx 1387]

You know, the other situation is "AI 180° NOSCOPE ME!" so I guess it's never right. :>

 

Always thought AI / MG vs. player was just for ambient effect anyway. I mean, as a mission designer, you don't want to kill the player, you want to have him stay on the edge. How much of that is valid in the case above though, no idea.

[Beagle 684]

The MG or LMG is the mainstay of a squad in neutralizing targets. A lot of Armies still use the MG as the main weapon of an infanrty squad that all other members have to support, not the the other way around. When MG and LMG lose the superiority in an squad vs. squad firefight, something is badly wrong and defies all light infanrty doctrines from the real world.

 

I really can't imagine that this was supposed that way.

[mmm 35]

5 hours ago, pierremgi said:

 

You know that wasn't too bad actually, he did kill you within 30 seconds(another minute or 2 maybe if you manage to hit prone....) I had 3 on 3 both side with Mk200 starting from 350ish meter away, they didn't really start hitting much before 150m and eventually just limping away from each other because both side ran out of the 600 rounds. That's a great way to make peace of people.

 

5 hours ago, lexx said:

You know, the other situation is "AI 180° NOSCOPE ME!" so I guess it's never right. :>

 

Always thought AI / MG vs. player was just for ambient effect anyway. I mean, as a mission designer, you don't want to kill the player, you want to have him stay on the edge. How much of that is valid in the case above though, no idea.

By all means you should have the tool to make AI shoot without hitting. 1st mission in official campaign already did that pretty adequately IMO, I think putting AI aim skill to zero will do. I just don't believe having such inaccuracy inherently reside with MGs all the time make any sense.

 

The flipside of the coin is that you're leading an AI squad, and you want to maximize the effectiveness of your man to the best of your knowledge. If you don't know any better, just base on your personal experience with ARMA weapons or common sense the choice might be having more MGs in your team. Well... Let's say they could have been better served as pack mules.

 

AI in the past was not completely "180° no scope" either IMO. Spraying in standing with LMG was more or less like what it is now. They only become exceptionally lethal when dropped down to prone stance, a condition a bit too easy to meet maybe. But then again you have the sniper rifle situation. I think "go prone to be more accurate" was more intuitive and a better starting point than what it is now with machinegun, but that's just me.

[Beagle 684]

I've shot the MG-3 on the 1000 meter combat training range (the one with the moving popup targets) a few tims. ca. 2000 life rounds in total. It is really no problem to bring the targets down up to 800 meters with short burst consiting of 10-12, just with iron sight from standing position with rested bipod. I noticed since the beginning that Shooting an MG in ArmA III was never what i rember it to be. The closed always was the M-200 and Zafir before it got nerfed with more spread.

 

THis is simply going the wrong way....especially with the HMGs.

A Maschine gun weighting around 10 kg is not moving much at full auto resting in its bipod RL.

 

ArmA II is going the wrong direction here way to long, particularly since the DLC MGs are barely of any use for quite a while.

I also noticed that the holding animations for the large MGs is just wierd and anatomically imposible...you need rubber limbs for that.

Another issue in 1.78 is the exageratd recoil animation...the whle upper body of the shooter is compressed like a rubber shock absorber in prone postion from bipod... thats simply odd.

[mmm 35]

I don't have much problem with machinegun accuracy in actual simulation in general. In fact the weight seem to be properly accounted for, with MG suffering far less muzzle climb when deployed than a comparable rifle. On the other hand I don't think anyone can steady a 15kg loaded MG fully laden with attachments unsupported like one do in ARMA, at least not relative to how one can do with a lightweight assault rifle. They could probably have made it in such a way that "hold breath" give you a short period of stability even when unsupported, but far more limited than with lighter rifle while also consuming stamina. While the heavier ones could enjoy even less climb when deployed.

 

On the topic of weight the ammo weight inconsistency and exaggerated weight of certain weapons are other matters I have a bit of issue with. But then again BI didn't really do much to make ammo bearer actually work for AI, not without you as player monkey around for 5 minutes at least.

 

But before all that I want them revert AI accuracy with MG to previous level, if not for any reason make them fire in no longer than 3 rnd burst for even the highest ROF LMG at all combat ranges other than point blank, and shorten the interval between bursts.

[pierremgi 4890]

The problem is probably linked with dispersion coefficient (cfgWeapons config reference). Some dev has corrected the insane accuracy of the Gatling gun mounted on helos. Perhaps a little bit too much!. At this time, it seems to me the AI's "accuracy" changed also for AI's LMG. Perhaps, some advisable dispersion, applied in excess?

 

[mmm 35]

14 hours ago, pierremgi said:

The problem is probably linked with dispersion coefficient (cfgWeapons config reference). Some dev has corrected the insane accuracy of the Gatling gun mounted on helos. Perhaps a little bit too much!. At this time, it seems to me the AI's "accuracy" changed also for AI's LMG. Perhaps, some advisable dispersion, applied in excess?

 

Just taken at face value it is around 20 times dispersion area in general, 1/20 the hit probability. I still think their burst behavior and the removal of additional accuracy and fire rate when prone are factors to this but this explains a lot of the first shot inaccuracy.

 

If I can test some mod weapons that doesn't follow their config guideline on LMG then it will be more conclusive.

[Jvin 0]

I also find the lack accuracy of the heavy gunners and machinegunners to be annoying and the fact that a squad of heavy gunners and machinegunners vs a squad of riflemen will lose to the riflemen almost all the time to be highly unrealistic, it really ruins the immersion somewhat. I'm actually in this forum looking for answers on how to resolve this problem after I noticed it in the editor. At least I know I'm not the only one that is bothered by this.

 

Additionally, I noticed that the suppression effect on ai from autoriflemen easily makes up for the fact that it takes longer for them to hit their target compared with riflemen (which is nice), however after testing in the editor in plain vanilla arma 3, I noticed that for machinegunners and heavygunners there doesn't seem to be much of a suppression effect on the ai from their fire. If ever suppression from heavy weapons like machineguns should actually be far greater than that from rifles with large magazines. I'm not sure if it's because of their accuracy or something else. From what I know machinegunners and heavy gunners are all from the DLCs of arma 3 and not the base game.

 

Did anybody else experience this problem with suppression from lmgs or just me?

[pierremgi 4890]

You're right. Suppression is overwhelmed by Arma engine, anyway, even if you order a duration. Here is a loop to make something durable.

On the other hand, I tested to suppress on AI enemy leader, Arma vanilla, regular difficulty, 3 prone heavy gunners vs a 8-squad. The squad wasn't impressed and finally won.

I can' imagine the bullets sent ratio, but far in favor of the static squad (no waypoint in test) .

[Jvin 0]

Thanks. I'll try that script in the link you mentioned see if it changes anything. I've seen mods that increase the suppression effect on ai in arma 3, I wonder if those mods increase/correct the lack of suppression effect for heavy gunners and machinegunners added by the DLC.

 

If you do that same scenario but replace the 3 heavy gunners with autoriflemen, the autoriflemen actually have a decent chance of winning that fight due to their suppression effect and as long as none of the riflemen escape and flank them, effectively removing them from the area affected by the suppression. I sometimes feel that fire from heavy gunners doesn't seem to have any suppression effect at all.

[Janez 531]

Is AI even "using" bipods? If not, then their accuracy shown in pierremgi's videos is quite representative of undeployed MMG accuracy that player can achieve. I remember one of the marksman trial courses, where you had to use SPMG .338. First shot was on target, if your patient, and the rest goes to the sky.

 

Now, I haven't done any recent proper testing with different global AI difficulty settings, but I'm all for AI having as close capabilities to player as possible, rather then T-800's.

[mmm 35]

The "suppression" is really just artificial accuracy penalty for AI when bullets fly by, not because they adopted more protective posture at cost of lethality. It's all the more ironic the least effective weapon somehow scares the bot more.

 

MGs at least need to be reasonably lethal under some circumstances to be credible "suppression". Having larger AI dispersion when standing(but maybe not 20 times), or poor close combat snapshot accuracy similar to that shown in the video are all reasonable downsides. But some "simulated resting" in prone like before is sorely needed, it doesn't have to take effect the instant AI hits the deck or be nearly as powerful as it was before. Also AI's urge to spray bullets all the time needs to be kept in check. If they can't use the "simulated resting" then they need to fire in much shorter burst or single fire. If BI can take a step further to detect objects in front of AI's weapon for resting bonus other than prone(even if AI can't consciously make use of cover), then great.

[pierremgi 4890]

On 04/01/2018 at 9:53 PM, Janez said:

 I haven't done any recent proper testing with different global AI difficulty settings,

You should.

[mmm 35]

I'll necro it myself since I'm in all the sudden feeling I still can't come to term with it. Can someone more familiar with the inner working of ARMA tell me: Is there anything governs the prone firing accuracy of AI in the cfg files? I remember in the earlier days of ARMA 3 AI machiengunners gone prone on their belly were rather lethal, you'd be very wary with their volleys since every shot actually count. While it could have been in somepart due to the "AIdispersion" with LMG(save for a few that were treated as assault rifle), if my memory servers me they were no more accurate than average assault rifle when not in prone. In fact I recall even assault rifle gained accuracy in AI's hand when prone. Does not seem stance has any noteworthy effect on AI accuracy now.

 

So anyone has better idea what happened exactly? And when was that?

[SwissArmy1984 8]

Not only is dispersion a problem.  The distance at which LMGs, DMRs, and sniper rifles engage targets is too short.  There seems to be a hard stop at 750 meters for everything.  This makes zero sense.