2870 replies to this topic

[mevitar]

Prioritizing isn't possible, and even if it would, making units do everything by themselves won't happen. You're supposed to manually select targets with those units, not wait until they do everything themselves.

Oxidizer's debuff stacks. Magnetron affects multiple units at once.

Edited by mevitar, 30 July 2015 - 08:52 PM.

[BlackAbsence]

Prioritizing isn't possible

Aawwwwwwwww.... :c

You're supposed to manually select targets with those units, not wait until they do everything themselves.

But I'm lazy :3 and can be overwhelmed. 

Oxidizer's debuff stacks. Magnetron affects multiple units at once.

I did not know that o.o I'll definitely be using Oxidizers more often with this knowledge. 

Well.. a Magnetron can only pin down one target at a time. I agree it affects others around it, but not to the same degree of it's soul target. I guess this does give more time for micro management, though

Edited by BlackAbsence, 30 July 2015 - 09:32 PM.

[lovalmidas]

The AI does a bit of prioritizing in their targeting, but not within the scope that you want. The relevant info can be found in Appendix B of the 8th post here: http://forums.renega...ead.php?tid=938

 

More concisely, the same thing can also be found here: http://modenc.renega...e_Threat_System

 

This part of the code controls how AI-controlled units attack other targets. Human-controlled units are not affected by this (unless the player house has an internal-IQ more than 0 or the [IQ] settings changed to let the game AI interfere with player commands).

 

In MO, some of the coefficients are negative. This makes the AI more efficient in eliminating targets. The idea is to concentrate fire on weaker targets

 

To put it in a more intuitive language, in MO, a Target has a greater "Threat" (and be prioritized to eliminate first) when they:

1. The Attacker's best weapon has a smaller Versus against the Target's armor (TargetEffectivenessCoefficientDefault is negative)
2. The Attacker is currently attacking the Target (TargetEffectivenessCoefficientDefault is negative)
3. The Target's best weapon has a greater Versus against the Attacker's armor (MyEffectivenessCoefficientDefault is positive)
4. The Target has a greater Special Threat (TargetSpecialThreatCoefficientDefault is positive)
5. The Target has a lower strength percentage (TargetStrengthCoefficientDefault is negative)
6. The Target has a smaller distance from the Attacker (TargetDistanceCoefficientDefault is negative and small) 

The AI calculates this when (a) retaliating after being attacked, and (b) is idle and able to acquire targets (e.g. guarding or patrolling).

 

In addition, having occupants and passengers inside the Target may increase its threat too.

 

 

We can appreciate that this also tells us a bit on what the current system cannot do.

 

1. AI does not calculate threat all the time (or it will lag). The Magnetron pretty much has to be attacked or its target destroyed before it begins to think about choosing a different target. This is so exploitable.
2. AI cannot differentiate among different instances of the same unit other than through factors (2), (5) and (6), and (6) has a very small contribution. For instance, once a Magnetron targets one out of ten Rhino tanks, it will most likely continue to target that Rhino Tank unless factor (5) overpowers it. See also the fact that AI has to be forced to re-calculate threat before it can think about taking another target.
3. The absence of MyDistanceCoefficient function in the game code means that it is not possible to have the Magnetron prioritise closer targets without having other units prioritizing closer targets.
4. The threat calculation does not consider many factors that can potentially be used to tailor the AI response. Things like:

• anything related to AttachEffect or crate effects (may need to confirm this)
• Speed - Instahit artillery may choose to prioritize speedy targets, and inaccurate artillery may choose to prioritize slower targets
• Target's best weapon Range - speedsters may prioritize artilleries, and mid-range units with some MinimumRange can prioritze melee objects and kite them. (AI microing huehuehue...)
• A NeighbouringThreat system, where the target's threat to neighbouring units (within a range) are considered as well. May increase group survivably and likely defeat any player's attempts to exploit the rock-paper-scissors effect in an AI army.
• and many more... (some can be completely covered by Special Threat values, like Sight)

 

The most important limitation is that these improvements/enhancements will only improve the game AI, in particular against player micro. Unless Speeder thinks having the AI take partial control from the player, the human player will never enjoy these benefits. 

 

As such, if such prioritization is possible, you may be facing a more challenging AI. To be honest, I'd like that.

 

AFAIK, Ares has not touched upon this section of the game code, and it is up to them if they want to.

 

I am for the ideal that the game AI should make your micromanagement easier. And we are quite a distance from that ideal.

[lovalmidas]

What you can do with the Magnetrons in the current situation is to space them out with some distance (about 2-4 cells wide) between them. Since Magnetrons have a significant range, as the enemies approach the Magnetrons, the closest Magnetron will fire first and stop one tank, then the second-closest Magnetron, being out-of-range of the stopped tank, will target a different tank.

 

What you can also do is to group 1 or 2 Magnetrons with tank killers (Lashers, Archers, Maruaders, Brutes). With 10 Magnetrons spread into 5 groups, you can target and stop 5 tanks in just 5 keyboard key presses and 5 clicks. The Epsilon army is best used when grouped according function rather than by the same unit. 

[Malekron]

The AI does a bit of prioritizing in their targeting, but not within the scope that you want. The relevant info can be found in Appendix B of the 8th post here: http://forums.renega...ead.php?tid=938

 

More concisely, the same thing can also be found here: http://modenc.renega...e_Threat_System

 

This part of the code controls how AI-controlled units attack other targets. Human-controlled units are not affected by this (unless the player house has an internal-IQ more than 0 or the [IQ] settings changed to let the game AI interfere with player commands).

 

In MO, some of the coefficients are negative. This makes the AI more efficient in eliminating targets. The idea is to concentrate fire on weaker targets

 

To put it in a more intuitive language, in MO, a Target has a greater "Threat" (and be prioritized to eliminate first) when they:

1. The Attacker's best weapon has a smaller Versus against the Target's armor (TargetEffectivenessCoefficientDefault is negative)
2. The Attacker is currently attacking the Target (TargetEffectivenessCoefficientDefault is negative)
3. The Target's best weapon has a greater Versus against the Attacker's armor (MyEffectivenessCoefficientDefault is positive)
4. The Target has a greater Special Threat (TargetSpecialThreatCoefficientDefault is positive)
5. The Target has a lower strength percentage (TargetStrengthCoefficientDefault is negative)
6. The Target has a smaller distance from the Attacker (TargetDistanceCoefficientDefault is negative and small) 

The AI calculates this when (a) retaliating after being attacked, and (b) is idle and able to acquire targets (e.g. guarding or patrolling).

 

In addition, having occupants and passengers inside the Target may increase its threat too.

 

 

We can appreciate that this also tells us a bit on what the current system cannot do.

 

1. AI does not calculate threat all the time (or it will lag). The Magnetron pretty much has to be attacked or its target destroyed before it begins to think about choosing a different target. This is so exploitable.
2. AI cannot differentiate among different instances of the same unit other than through factors (2), (5) and (6), and (6) has a very small contribution. For instance, once a Magnetron targets one out of ten Rhino tanks, it will most likely continue to target that Rhino Tank unless factor (5) overpowers it. See also the fact that AI has to be forced to re-calculate threat before it can think about taking another target.
3. The absence of MyDistanceCoefficient function in the game code means that it is not possible to have the Magnetron prioritise closer targets without having other units prioritizing closer targets.
4. The threat calculation does not consider many factors that can potentially be used to tailor the AI response. Things like:

• anything related to AttachEffect or crate effects (may need to confirm this)
• Speed - Instahit artillery may choose to prioritize speedy targets, and inaccurate artillery may choose to prioritize slower targets
• Target's best weapon Range - speedsters may prioritize artilleries, and mid-range units with some MinimumRange can prioritze melee objects and kite them. (AI microing huehuehue...)
• A NeighbouringThreat system, where the target's threat to neighbouring units (within a range) are considered as well. May increase group survivably and likely defeat any player's attempts to exploit the rock-paper-scissors effect in an AI army.
• and many more... (some can be completely covered by Special Threat values, like Sight)

 

The most important limitation is that these improvements/enhancements will only improve the game AI, in particular against player micro. Unless Speeder thinks having the AI take partial control from the player, the human player will never enjoy these benefits. 

 

As such, if such prioritization is possible, you may be facing a more challenging AI. To be honest, I'd like that.

 

AFAIK, Ares has not touched upon this section of the game code, and it is up to them if they want to.

 

I am for the ideal that the game AI should make your micromanagement easier. And we are quite a distance from that ideal.

Wait the AI can mirco?

[lovalmidas]

Computers can be terribly efficient at micromanagement AND marcomanagement if you code them properly, or at least defeat most human attempts to pull a fast one on it. Did you know that your short-term memory can only handle about 4-7 bits of information at any one time? Take a calculator with a ~10-digit display, key in random() or eqivalent command, recite the displayed numbers once (or twice), and then cover the display. How many numbers can you recite?

 

In almost all aspects, computers have greater processing capacity than the human brain.

 

The fact that the game AI does so poorly is entirely due to either (1) limitations of the program, or (2) AI design dumbing the AI down. There is an overall lack of memory learning (the AI is unable to collect patterns from previous games), no self-writing (AI cannot write its own teamtypes, the ini scripter must write everything), and godawful legacy code (Hardcoding IsCow. AITriggerType having very limited conditions. BLITTING. Really...). There are opensource games out there with some great AI.

 

The YR combat system is quite an achievement a few decades ago, however. As one of the main products of Tiberian Sun, this threat calculation and evaluation logic has a lot of potential to give the AI some 'intelligent activity' in picking up their targets. It is a tremendous improvement from RA1 logic. But-

 

https://www.youtube....h?v=TXeSBfcEW_0

Well, ini coders in WW did not know how to use the threat calculation system properly. The AI subsequently does poor micro. Mental Omega improves the values somewhat, so in the very least, you get to observe the AI focusing on your damaged units first

 

When players exploit this (be it against the AI or another player, it is not wrong, but it places the AI at such a disadvantage that other solutions were relied upon to provide a challenge (yes, "AI obviously cheats" and other tricks).

 

------------------------------------------------------------------------

The overall execution is far from intelligent though. The threat calculation is recalculated too rarely. See Speeder's dev playthrough of S04: Side Effect. At 4:16, observe how Speeder retreats a single damaged Rhino tank behind his Refinery, to draw the Zephyrs to the Battle Bunkers. To resolve that requires some rewrite in the game engine.

 

Yes, the AI can micro in the sense that it can select who to priorize to kill. The AI is also known to autocrush you. And then there are also the chrono-miners that quickly teleport back to refinery when it is hit below 50% health regardless of whether it has mined its full load. Teamtypes set to aggressive will automatically retailiate against an attacker. Yes, the AI has always been microing all these while.

 

Just not good enough.

 

------------------------------------------------------------------------

If you are interested to know more about the engine, looking at what Ares has achieved will open a whole new world. There are other projects working on the YR engine, but Ares is one of the more accessible ones. OpenRA is another place where you can appreciate how much potential lies in the original game. Ares tries to build on the existing engine, OpenRA tries to rewrite it from a blank slate. It will be quite sometime before OpenRA starts to work on RA2, but RA2 engine inherits many things from TS, so it's a start.

 

And you will also appreciate a bit why many mod developers blame Westwood for their misfortunes.

Edited by lovalmidas, 02 August 2015 - 04:39 AM.

[Petya]

With some units, it is redundant for more than one of them to be attacking the same target.

This is the case with units such as; Magnetrons, Dragonflys, Oxidizers, and Stalkers (Only when stunning infantry), for; they only supply a supporting effect which cannot be applied further.

i.e: 10 Magnetrons pining down 1 tank in a group of 10 tanks while letting the other 9 of those tanks reach the 10 Magnetrons is just dumb AI 

 

So I was wondering; Is it possible for some units to prioritise targets which aren't already being effected by the effect they apply?

Because if so; it would alleviate all the tedious micro management you'd have to do, in order to maximize their effectiveness in larger groups.

i.e: You'd end up with 10 Magnetrons pinning down each tank in a group of 10 tanks without having to individually order them to do so.

 

I am also wondering; is it possible for your units to prioritize targets that are debuffed with an effect, mainly; the "armour down" effect?

Becuase if so; it would approve upon their effectiveness, too, and again, get rid of some micro management.

i.e: There's a swarm of Wolfhounds above your base. All you have is Oxidizers and Gatling Tanks and it is hard to select the Wolfhounds that have the bebuff applied to them via the Oxidizers due to them being swarmed in close to each other, but fortunately; your units prioritize the debuffed targets for you (which otherwise would have been near impossible), maximizing their (Oxidizers & Gatling Tanks) effectiveness.

 

 

If any of these prioritization things are possible; that would be sweet

If not; Aawwwwwwwww.... :c

Sorry to disappoint you, but this a RTS game, which means it will involve micromanagement and macromanagement. All RTS games have micro intensive units, missions etc.. Just look at StarCraft 2. It involves lots of micro. As for the units you mentioned... Magnetron is a micro intensive unit so in order to make it useful, each Magnetron should attack another target, not the same one. I don't have a problem with it, it compensates its power. As for Gatling Tank + Oxidizer: just group them and then order the groups to attack Wolfhounds one by one.

[Martinoz]

Did you know that your short-term memory can only handle about 4-7 bits of information at any one time?

4 to 7 bits it's still a lot for us, 128 tasks at a time .

[BlackAbsence]

Sorry to disappoint you, but this a RTS game, which means it will involve micromanagement and macromanagement.

No disappointment here.

Magnetron is a micro intensive unit. I don't have a problem with it, it compensates its power.

I'd think their cost and time constructing them would be enough compensation for their power. Valuble.. valuble time.. o.o

As for Gatling Tank + Oxidizer: just group them and then order the groups to attack Wolfhounds one by one.

True. Joined focus fire would be best, but what if their your allies Oxidizers? lol (Which would probably never happen)

Edited by BlackAbsence, 02 August 2015 - 05:08 PM.

[lovalmidas]

 

Did you know that your short-term memory can only handle about 4-7 bits of information at any one time?

4 to 7 bits it's still a lot for us, 128 tasks at a time .

 

 

I agree, but it still pales in comparison to 32-bit and 64-bit processors. Our long-term memory is still a mystery (in Windows OS we can use FileSystem as an analogy, somewhat, but we know quite little about the human brain.)

[Malekron]

@lovalmidas

I agree, but it still pales in comparison to 32-bit and 64-bit processors. Our long-term memory is still a mystery (in Windows OS we can use FileSystem as an analogy, somewhat, but we know quite little about the human brain.)

 

It's evolution baby.

[MrFakeSmile]

"Do the evolution" should be the name of a mission.     

[Meyerm]

Why do you keep calling them Foehn?

[Solais]

Also the image of their logo is called logo_foehn.

[BlackAbsence]

 

Why do you keep calling them Foehn?

 

Because in the side 4 reveal trailer says the wind is changing, the foehn or phoehn is something related to the winds. Or. Something...

 

http://www.merriam-w...ictionary/foehn

Also the image of their logo is called logo_foehn.

Perhaps this is hinting towards them being a very aerial oriented Side.

We've got Soviets as Unit oriented.

Epsilon as Infantry oriented.

Allies is a mix of everything.

so Foehn is aerial oriented?

Just a thought.

Edited by BlackAbsence, 06 August 2015 - 09:44 PM.

[Divine]

I think it would be realy nice if the Soviets (or at least one of their fractions) got some sort of a flying aircraft carrier. I'm thinking of some sort of an airship, either a modified Kirov, or something like this:

Obviously, you can see no take-off ramp on this thing, and it's a gunship... but I really like this double-baloon design. About the aircraft: some sort of a rocket povered, point-defense aircraft, loud, fragile, with good firepower. Soviet technologies in RA2 are often inspired by experimental nazi weapons, so I think it would be really nice to stick to this tradition:

Ladies and gentlemen, this is a Natter. This aircraft was designed by nazi Germany to intercept Allied bombers. The aircraft's frame was made of wood, and it was powered by a cheap and primitive rocket engine. The Natter took off vertically, with fuel enough for 5-7 minutes of flight. The pilot used the rocket pod in the nose to engage enemies, firing all rockets in a barrage, then jumped out of the plane with a parachute. This was an experimental aircraft, and while some were sent to combat units, it never took part in engagements. This should be pretty much the aircraft launched by the Soviet carriers, with the difference that they should also try to return to the carrier once their payload is dropped, and also probably should have bomblets instead of rockets, or make attack runs with machine guns, so it would not be too much like the Hornet drone.

Edited by Divine, 09 August 2015 - 11:25 PM.

[Solais]

I think if there should be a flying carrier, it should be either belong to the Allies or the 4th side, based on this.

[Meyerm]

The natter would be useful as is. But for balancing reasons, maybe make it a tech structure. It launches  AA drones that hunt aircraft in range (will pursue if aircraft leaves building's range), causing insane damage but only firing once. The catch is it can be shot down.

[BlackAbsence]

You know how infiltrating a Barracks grants veteran Infantry and infiltrating a War factory grants veteran Units.

Well, what if there was a way to acquire "veteran" (upgraded) Defences?

Personally; I think this would be cool if infiltrating a Construction Yard granted this.

I know by already infiltrating a Construction Yard it grants the Stolen MCV variants, but what if that was removed to make way for this "veteran" (upgraded) Defences idea, and instead; to acquire Stolen MCVs you must infiltrate both a Construction Yard + a Battle Lab for an added challenge?

Also, could you make it so if a Soviet infiltrates a Soviet Construction Yard, it doesn't grant a Stolen Soviet MVC (as well for the other sides), for that's a bit redundant. Just the Apoc would be nice.

Also, seeing that the Soviets and Epsilon are gaining Air Fields, personally; I think it would be nice to have a way to acquire veteran jets and whatnot via infiltrating an Airbase.

Also, is it possible to acquire a specific faction's MCV? i.e: Acquiring a ConYard belonging to PF will allow you to build PF stuff (and only PF stuff), instead of EA stuff by default, and so forth.

 

Thoughts?

Edited by BlackAbsence, 13 August 2015 - 07:44 AM.

[CLAlstar]

As far as i know, new tech structure will allow its owner to place veteran defense buildings.