97 replies to this topic

[Koniving]

Nesutizale, on 01 December 2019 - 03:23 AM, said:

You are thinking from a gamers perspective "How do I end the mission min-max-ing the best?" Answere, destroy the one target that wins me the mission.

Indeed.

Quote

Try thinking from a world building point of view. Think of a real battlefield with soldiers fighting to their death. What target is the most importend one? The one that is activly trying to kill you or the one that dosn't try to kill you?

Considering that ultimately all enemies are a threat, the priority should reflect the personality of the mechwarrior. Since we don't know what that is, they could have any priority.

The one that's the direct threat does make sense. But so does the easiest kill. When playing MWO sometimes the guy that's actually a threat to you becomes secondary if you've got enough health to spare and in making sure you get that extra credit for the easy kill.

Sometimes you'll prioritize the smallest threat that you know will tear your back open rather than the larger threat that needs to get through your armor first.

And yes, sometimes it's better to ignore the lack of threat from a target.

But at no point should every single enemy ignore the player that's got his feet kicked up in the cockpit reading about the real reason that pilots like to go to the Marian Hegemony while drinking mountain dew.

Quote

Also how often, in a real world scenario, does killing one person makes everything else stop and I mean literaly stop. One soldier gets killed and all bullets just drop to the ground and everyone just goes home. Does that sound realistic what would happen?
If the answere is no then you idea of "meta gameing" isn't helping in world building.

I think All Systems Nominal made a joke about that, where the entire team was converging on the last enemy that had no weapons, one leg and not much else... and suddenly everyone had to stop because the base timer counted down and he 'took the base'.

Quote

Hell even in MWO most of the time when you encouner an AFK, while there are other targets around that shoot at you, you let the one guy stand and kill him later instead of ignoring the guys that are firing on you right now.
Matter is different when you are a light that encounters an AFK far from the front lines, then surely you kill him quickly.

I do both when possible and use the afk as a shield in the process.

Quote

Long story short, yes the AI should priorities the active targets instead of the AFK. Remember that those AI Mechs run stock builds. From playing stock a lot during testing I know that firing a lot is realy bad. You want to shoot as little as possible as your single HS Mechs can store a lot of heat but the dissapation rate is terrible.

(This is a frontloaded damage/heat, rapid fire rating issue, many stock builds wouldn't actually have that issue until after losing some heatsinks to damage).

Quote

So I only shoot the targets of most importance.

Command unit would realistically have some priority. But at the moment even if you tell the others to hold fire, the other AI pilots get the target priority for enemies.

Quote

That is indeed a flaw that I have encountered too. Also sometimes the AI not reacting to your presents, like a Locust that didn't started firing at me or move at all before I got into close range. That happened rarely but it does from time to time.

Realistically he shouldn't target you until you're detected. They react to proximity of your radar tick.
If one enemy is looking at you it puts you on everyone's radar and if they're in proximity they go for you (and if you've entered the proximity before they'll start going toward you until you reach a specific distance).

If the Locust was alone and no one was looking at you, the reaction range is their longest weapon range. In the case of the Locust that was probably 270 meters. Anywhere in 360 degrees if you enter their longest weapon range, they'll react to you.
The exception is if you somehow disappear from the radar, which you can do by standing inside some buildings.

[Jyi]

Nesutizale, on 01 December 2019 - 03:23 AM, said:

You are thinking from a gamers perspective "How do I end the mission min-max-ing the best?" Answere, destroy the one target that wins me the mission.
Try thinking from a world building point of view. Think of a real battlefield with soldiers fighting to their death. What target is the most importend one? The one that is activly trying to kill you or the one that dosn't try to kill you?
Also how often, in a real world scenario, does killing one person makes everything else stop and I mean literaly stop. One soldier gets killed and all bullets just drop to the ground and everyone just goes home. Does that sound realistic what would happen?
If the answere is no then you idea of "meta gameing" isn't helping in world building.

Hell even in MWO most of the time when you encouner an AFK, while there are other targets around that shoot at you, you let the one guy stand and kill him later instead of ignoring the guys that are firing on you right now.
Matter is different when you are a light that encounters an AFK far from the front lines, then surely you kill him quickly.

On a real battlefield, would you rather 1) eliminate a tank that was shooting at you just a minute ago but then went completely silent or 2) a single soldier running around in circles, shooting at the sky? Because that's pretty much what's happening in a video clip where B33f goes AFK.

Also, in MWO, would you rather take out a stationary assault that was shooting at you just a moment ago or try to hit a Piranha that's swirling around, shooting machineguns in every direction?

If we try to think about the world of Battletech realistically, I think you ALWAYS prioritize taking out an assault mech if the opportunity presents itself. You don't chase the squirrel, because that gets you killed.

Quote

Long story short, yes the AI should priorities the active targets instead of the AFK. Remember that those AI Mechs run stock builds. From playing stock a lot during testing I know that firing a lot is realy bad. You want to shoot as little as possible as your single HS Mechs can store a lot of heat but the dissapation rate is terrible. So I only shoot the targets of most importance.

No, this is stupid. The AI should have some rudimentary threat assessment capabilities. If the AI just shoots at the target that shot them last, it can be abused so easily and the AI will just keep switching targets and getting nothing done. Like it does right now.

Quote

That is indeed a flaw that I have encountered too. Also sometimes the AI not reacting to your presents, like a Locust that didn't started firing at me or move at all before I got into close range. That happened rarely but it does from time to time.

This is an understatement in my experience.

Edited by Jyi, 01 December 2019 - 03:09 PM.

[Ilfi]

A bit late on the follow-up, but:
In my initial post, I was hinting at dropping solo so the AI wouldn't target your lancemates. That + tickling them with medium lasers so they look at you may give you a better read on just how dumb the AI is.

[Khobai]

Quote

Try thinking from a world building point of view. Think of a real battlefield with soldiers fighting to their death. What target is the most importend one? The one that is activly trying to kill you or the one that dosn't try to kill you?

But in a real battlefield youre fighting an intelligent enemy. Since the AI is super dumb its already not anything like a real battlefield. worldbuilding fail.

the AI can never win if an intelligent player can metagame the AI while the AI cant do the same in return. This has already been demonstrated to be true.

assuming the AI will always be bad (and it will always be bad because PGI), the only way the AI can ever win is if it prioritizes the player over lancemates. Because for the AI killing the player is the easiest way to make them fail the mission.

worldbuilding is great when its possible. But if your AI is horrendously stupid youve already failed at worldbuilding. So you might as well not have an easymode game too.

Edited by Khobai, 02 December 2019 - 12:36 PM.

[Nightbird]

Well, you can make a game be about fighting a few enemies that are skilled and scary, or a lot of enemies that are just fodder.

MW5 is certainly not the former, there is no sense of urgency or thrill at winning, the best you can expect is a feeling of grinding moving targets a little more efficiently.

[Jyi]

Nightbird, on 02 December 2019 - 01:21 PM, said:

Well, you can make a game be about fighting a few enemies that are skilled and scary, or a lot of enemies that are just fodder.

This isn't actually true.

For example, you can make a map that has multiple lances of enemy mechs that are EACH capable of putting up a challenge.

Then the game becomes an excercise in outplaying the enemy by employing strategies. Like having a spotter and using LRM's. Or feint attacks and ambushes. Or flanking and surrounding the enemy. And so on.

All of these are strategies that have been available (to some extent) in earlier Mechwarrior -games and other mecha-games. But then again, those games have had handcrafted maps and handplaced enemies with triggers and checkpoints, so the AI didn't need to be so smart.

MW5 is what you get when you concentrate on procedural maps without developing the AI to take advantage of them.

Edited by Jyi, 02 December 2019 - 01:47 PM.

[Nesutizale]

Well it is said that the first mech, the Mackie "utterly obliterated his test targets, four remote-control Merkava heavy tanks."
So when it comes to tanks and VTOLs you need a good number of them. Its part of the "Mechs are better" part of what makes BattleTech BattleTech.

Mech to mech combat, that is another topic. Here I agree that it should be harder. Still I have seen my lancemates getting killed by the Ai and I have lost an arm or leg during combat too. So it can't be said the Ai dosn't do any damage, its just rare to do seriouse damage. I still hope that this will change.

Some "Hero" or "Nemesis" enemys with a particular good Ai would be nice too.

[Kubernetes]

Some of these complaints are silly.

"For example, you can make a map that has multiple lances of enemy mechs that are EACH capable of putting up a challenge.

Then the game becomes an excercise in outplaying the enemy by employing strategies. Like having a spotter and using LRM's. Or feint attacks and ambushes. Or flanking and surrounding the enemy. And so on."

And 90% of players will get smashed without completing the mission. There is no healing or mid-mission repair, so multiple engagements against quality AI opponents will break you down. If the AI were even half as good as a T5 MWO pilot, you couldn't get past 10 enemy mechs, much less all the tanks and aerospace fighters coming at you.

[Prototelis]

As much as grognards hated it; the MFB in MW3 was AWESOME. Especially on longer missions.

[Nesutizale]

The MBF where pritty much the only thing that made longer missions even possible I think.

Also from looking at the MWO playerbase, to all those complain about the game not beeing difficulte enough... I would say you are not the target audiance as even in MWO you make up what? The top 10% of players? Thats just by going with all the "there where again 11 stupid people in my team" comments.
Its good for you to be so good, be happy but you are most likely the smallest amount of players, even more so I guess from the people that will play MW5.

Pesonaly I think PGI should patch in an extra hard mode for you guys later on, kay?

[Prototelis]

Good AI != Difficult AI

It isn't about the game being difficult as much as it is about not having AI equivalent to a 90's game

[YueFei]

Prototelis, on 02 December 2019 - 04:40 PM, said:

Good AI != Difficult AI

It isn't about the game being difficult as much as it is about not having AI equivalent to a 90's game

Correct. You could trivially make an AI that's got perfect aim and headshots you every time, whilst having it behave stupidly in every other way, and it would be "difficult" but not immersive and provide no challenge because it doesn't give the player any possibility of counter-play.

As others have already mentioned, this is already a solved problem. Look at FEAR's AI. There are even academic papers written about it that detail how that system works. And that's a 14-year-old game.

[Ilfi]

A bit of a tangent, but Doom has phenomenally good AI and enemy design that holds up even today.

I hope PGI can tune up the current system, because clueless AI that ignores you until you headshot them won't suffice in the slightest. They need to be more aware of their general surroundings than that.

[Jyi]

Kubernetes, on 02 December 2019 - 03:23 PM, said:

Some of these complaints are silly.

"For example, you can make a map that has multiple lances of enemy mechs that are EACH capable of putting up a challenge.

Then the game becomes an excercise in outplaying the enemy by employing strategies. Like having a spotter and using LRM's. Or feint attacks and ambushes. Or flanking and surrounding the enemy. And so on."

And 90% of players will get smashed without completing the mission. There is no healing or mid-mission repair, so multiple engagements against quality AI opponents will break you down. If the AI were even half as good as a T5 MWO pilot, you couldn't get past 10 enemy mechs, much less all the tanks and aerospace fighters coming at you.

Not if there's an easy difficulty setting where they 1) take more damage 2) do less damage 3) have worse accuracy or any other easily adjustable variable.

This is still the dumbest argument against having a smart AI. Stop thinking that making games for the lowest common denominator is a good idea.

[Nesutizale]

Maybe its a bug but yesterday beeing shot by allies suddenly didn't seam to cause as much damage as it should or none at all.
Can someone confirm that? Have they patched in a "friendly fire detection" so you don't get killed by your own lancemates or was that just a bug?

[Jyi]

Nesutizale, on 03 December 2019 - 01:40 AM, said:

Maybe its a bug but yesterday beeing shot by allies suddenly didn't seam to cause as much damage as it should or none at all.
Can someone confirm that? Have they patched in a "friendly fire detection" so you don't get killed by your own lancemates or was that just a bug?

I think it was just a bug.

There hasn't been a game version update on the demo.

And I don't think there should be, either. They are concentrating on release, and hopefully tackling these issues for that version.

[Koniving]

Nesutizale, on 02 December 2019 - 01:39 PM, said:

Well it is said that the first mech, the Mackie "utterly obliterated his test targets, four remote-control Merkava heavy tanks."
So when it comes to tanks and VTOLs you need a good number of them. Its part of the "Mechs are better" part of what makes BattleTech BattleTech.

Mech to mech combat, that is another topic. Here I agree that it should be harder. Still I have seen my lancemates getting killed by the Ai and I have lost an arm or leg during combat too. So it can't be said the Ai dosn't do any damage, its just rare to do seriouse damage. I still hope that this will change.

Some "Hero" or "Nemesis" enemys with a particular good Ai would be nice too.

I've read that short story in the past few months.

One of the big issues is that the Merkava shells would literally bounce off the armor without doing any damage.
The tanks were also sporting BAR 7 armor and only 4 and a half tons of it, each.
So 72 points (144 in MWO/MW5 terms) of armor, where being hit by military grade weaponry would get 3 points extra damage per weapon rating (Merkava Mk IV tanks.)
So the AC/10 the Mackie sports would deliver 13 damage per cassette, the PPC would deliver 13 damage per shot, the medium lasers would deliver 8 damage each...
Meanwhile, each AC/5 shot fired at the Mackie, lacking the explosive component, would bounce off doing little if any damage at all. After that, it became standard to have an explosive element to autocannon rounds (so they explode on contact).

This said if you set a pair of Scorpion tanks (25 tons each) against a Shadowhawk 2D or 2H in tabletop, then you've got a battle of wits or there's a chance you won't come out alive, considering that Scorpion tanks have 64 points of BAR 10 armor and 15 structure. Granted that's 16 front, 11 per side, 10 rear and 16 on the turret.

Course, that's assuming you found two in perfect condition. Quikscell being the manufacturer, many tanks arrived damaged or missing components.

In Mw5, many of the tanks are already damaged before encountering them for some strange reason, often heavily damaged to the point where an AC/5 can kill them in a single blow.

In terms of modern vehicles, it's actually said the principal advantage mechs have over vehicles is maneuverability and reaction time. (Mechs can react to and deal with multiple threats at once or in rapid succession, most vehicles cannot do so nearly as efficiently.)

Edited by Koniving, 03 December 2019 - 03:48 AM.

[Jyi]

Koniving, on 03 December 2019 - 03:47 AM, said:

In Mw5, many of the tanks are already damaged before encountering them for some strange reason, often heavily damaged to the point where an AC/5 can kill them in a single blow.

That's because the game is so buggy, and perhaps because the AI keeps shooting at each other.

In my video of the assassination mission I show how there's a helicopter (not 100% certain if it's a helicopter, but what else could it be?) that keeps spawning in the sky and instantly exploding. After that I have also realized that the bouncy tank you can see in the video happens 100% of the time, but you can only see it from that direction.

I bet there are even more bugged spawns in the map that we just don't see.

Also, the 1st Locust is on yellow armor 100% of the time before it even gets to you - if it doesn't decide to bug out and just stand still. Because the enemies just don't care.

I'm 100% certain this also applies to tanks, and I'm 100% certain they occasionally outright die because a mech stomped on them.

Edited by Jyi, 03 December 2019 - 08:16 PM.

[FRAGTAST1C]

Enemy units that are already damaged isn't anything new if anyone has played the HBS's Battletech. You often encounter enemies there with less than 50% of the armour 'cause they're considered to be in "bad maintenance" and the mission plays out like you're sent to teach some annoying elements a lesson.

[HellJumper]

i dont own the game so i cant comment on how the AI is


on the other hand, if the games AI was hard, as in first hitting the player, killing him, damaging him etc

we will be seeing another topic here complaining about how hard the AI in this game is...