Attack-ground: include in A26 or not?

[wowgetoffyourcellphone]

16 minutes ago, Freagarach said:

When the enemies ranged units are in formation, targetting the formation will lead to each of your units targetting the closest member of that formation, which sounds more or less what you want? So the problem might also be addressed by quoting @wowgetoffyourcellphone: "Battalions".

Hmm, I would think one of the benefits of having battalions is that you could remove a bunch of range queries (or maybe not, I have no idea what I'm talking about, lol). You could definitely remove a bunch of vision range queries if vision range was tied to the battalion instead of the individual units in it. 

[a 0ad player]

When I imagine a line of archers in a battle, they fire arrows at an area. If the other ranged fighters are sighted, an area with as many people as possible is bombed with arrows. The area corresponds to the seen or assumed position of the enemy ranged fighters.

 

Translated I think an area is selected. The arrows are shot at individual ranged fighters with an appropriately increased inaccuracy. Most arrows land in the middle of the area (most enemy ranged fighters) and the intensity decreases towards the edge (distribution image). Lumped groups are hit more often, making battle orders more important. When most of the people in the area have died (less than 10 people remain in the area, for example) and the situation becomes confusing or new orders are given, the attack ground is over.

I think attackground for archers is a good idea. With spear throwers, I envision direct visible targets and not an area attack.

[real_tabasco_sauce]

Hi everyone, I appreciate the interest, and i'm glad we are having a fruitful discussion.

I suppose i wasn't clear about the distinction @chrstgtr:

attack-ground is a player-controlled attack blanketing an area as seen in @Freagarach's video.

however, far from aimless, it may do more damage to tightly packed armies because it avoids overkill. My imagination is that it could be used to the degree of onagers and mangonels in AoE2, where skilled players can anticipate movements and score effective damage against groups of weaker units.

On 31/12/2021 at 9:14 AM, chrstgtr said:

what I don’t want is a feature where units will just aimless shoot at an empty area (or stand idle) because they were “told” to while enemy units walk right in from of them

I see what you mean. Maybe something like what @BreakfastBurrito_007 mentioned would help, with the ranged units disengaging when no enemies are present. Another alternative would be to make the attack ground order non-repetitive, where a player is responsible for individual volleys if they want to take advantage of the benefits of attack ground.

When I mentioned attack-group, i was referring to your preference of attacking the units within a user-specified area.

 

maybe for future discussion we should define the two terms, so everyone is clear (let me know if i get something wrong here):

Attack-ground (see video): either a single or repeated attack on a circular player-specified area, where projectiles are likely to evenly spread their damage throughout the area.

 

Attack-group: Ranged units behavior is to attack enemies within a circular player-specified area, regardless of proximity within the area (similarly to how towers evenly spread damage)

^ feel free to change either. For Attack-group I think calling it a behavior is accurate.

 

Both of these sound appealing to me, and perhaps they are not mutually exclusive.

[chrstgtr]

1 hour ago, real_tabasco_sauce said:

Attack-ground (see video): either a single or repeated attack on a circular player-specified area, where projectiles are likely to evenly spread their damage throughout the area.

This doesn't make sense. Why would you ever ask your units to attack at random? In every way, this is less preferable than any of the options I describe below. 

 

I think there are only four real options here:

1. Tower-like attacks for a specified area that targets units within the area. This has at least two options as I explain below
   • With each volley, a new target is selected. This is the same as a tower.
     • Pro: This will probably encourage healers because units injured units won't just die right away like they do now while they are under attack from several units all at once.
     • Con: It will lead to a slow, sudden death of large armies. This will without a doubt lead to snowballing in 1v1s because large enemy armies will die almost all at once while the attacking army will lose basically no units. In team games, this will lead to a teammate's army being able to quickly wipe enemies even if the enemy army is large because units will be uniformly very injured. This will also make this type of attack
2. Targeting closest units within an area. This is the same way units are now except they will only attack units in a specified area.
   • Pro: It works. 
   • Con: It will still probably lead to some units being overly targeted. This will eventually lead to melee units being targeted as they walk through the ranged army
3. Repeated attacking of randomly selected units until they are dead.
   • Pro: this gets rid of the overly targeted problem that we have now
   • Con:  Coding? Nothing works like this now. This also leads to less control over attacking unit types here. This also make this type of attack not very workable with melee units. Additionally, this may lead to situations where your attacking units have to walk some distance before engaging in a fight, which is pretty undesirable, but I suppose that could either be fixed with code or itself could be seen as a feature because better micro would avoid this problem. 
4. Some combination of the above while selecting targets based on unit types
   • Pro: you won't randomly target melee walking through.
   • Con: Coding? Nothing works this way. Overly complicated?  

Personally, I like (3) and would like the option of adding (4) somewhere down the road. 

[real_tabasco_sauce]

1 hour ago, chrstgtr said:

This doesn't make sense. Why would you ever ask your units to attack at random? In every way, this is less preferable than any of the options I describe below. 

 

@chrstgtr I think you might be misunderstanding me. It's not random, it is an area that is shot at. Units in the area receive damage if an arrow hits them, which is more likely with a smaller radius (higher arrow density).

The biggest advantage of Attack-ground is that 50 archers for example, that would often times all shoot 1 unit at a time now deal more overall damage to a group of units, especially if they are tightly packed. This feature would be controlled by the player in response to micro in the battle. There's nothing random about it, other than the distribution of arrows within the circle.

In addition, there seems to already be working code for this and all that is required is testing by some means.

 

 

1 hour ago, chrstgtr said:

Repeated attacking of randomly selected units until they are dead.

• Pro: this gets rid of the overly targeted problem that we have now
• Con:  Coding? Nothing works like this now. This also leads to less control over attacking unit types here. This also make this type of attack not very workable with melee units. Additionally, this may lead to situations where your attacking units have to walk some distance before engaging in a fight, which is pretty undesirable, but I suppose that could either be fixed with code or itself could be seen as a feature because better micro would avoid this problem. 

 

This sounds like what I described for Attack-group above, and I agree that seems like a good option. Coding may be difficult as you said.

 

 

What I am saying is we have an existing option, which will be more easily implemented and tested. I'd like to know if people want to test it or not. Wouldn't it be better to test existing ideas before new stuff is developed?

If attack-ground should fall short of what we are looking for, maybe the next step is to try something along the lines of 3.

[Nullus]

23 minutes ago, real_tabasco_sauce said:

Units in the area receive damage if an arrow hits them, which is more likely with a smaller radius (higher arrow density).

 

I think that this would be the main usability problem with a pure attack-ground, where the units fire at a point on the ground. Pyrogenesis doesn't actually simulate the physical trajectories of projectiles, so it has trouble detecting if a projectile hits a unit if the unit wasn't specifically targeted. That would make most of the arrows miss harmlessly, even if they landed in the middle of a group. The main area in which I can see this being useful is with ranged units with splash damage, such as bolt shooters.

2 hours ago, chrstgtr said:

Con:  Coding? Nothing works like this now. This also leads to less control over attacking unit types here. This also make this type of attack not very workable with melee units. Additionally, this may lead to situations where your attacking units have to walk some distance before engaging in a fight, which is pretty undesirable, but I suppose that could either be fixed with code or itself could be seen as a feature because better micro would avoid this problem. 

The issue with units walking to targets could be fixed if the ranged units randomly selected targets within a player-defined area. That would also reduce the necessity of micromanagement.

[chrstgtr]

1 hour ago, real_tabasco_sauce said:

@chrstgtr I think you might be misunderstanding me. It's not random, it is an area that is shot at. Units in the area receive damage if an arrow hits them, which is more likely with a smaller radius (higher arrow density).

The biggest advantage of Attack-ground is that 50 archers for example, that would often times all shoot 1 unit at a time now deal more overall damage to a group of units, especially if they are tightly packed. This feature would be controlled by the player in response to micro in the battle. There's nothing random about it, other than the distribution of arrows within the circle.

In addition, there seems to already be working code for this and all that is required is testing by some means

This is either my (1) as described above or random shots. They are either targeting units or not. Not targeting units is useless. 
 

If shots aren’t being targeted and are just shooting at an area (as opposed to units within that area) then I don’t see why anyone would ever use that

 

[chrstgtr]

27 minutes ago, Nullus said:

The issue with units walking to targets could be fixed if the ranged units randomly selected targets within a player-defined area. That would also reduce the necessity of micromanagement.

That would be (1) or (3) as I describe above. I believe (1) is bad for the reasons k describe 

[real_tabasco_sauce]

25 minutes ago, chrstgtr said:

If shots aren’t being targeted and are just shooting at an area (as opposed to units within that area) then I don’t see why anyone would ever use that

yes attack-ground means the units are shooting an area, which is assigned by the player. The archers do not target, but instead the player does the targeting. For example: enemy has 30 pike and 30 skirm, I have 30 pike 30 sling. I would use attack ground to target the skirmishers because i know they will kill my pikes first. This should kill the skirmishers faster than individually clicking each one (because of avoiding overkill and not defaulting to shooting pikes).

The reasons to use this are:

-give the player some control over where arrows go, as opposed to the default of shooting the closest unit. This allows ranged units to shoot past melee for example.

-when attacking a lot of units that are close together, damage can be dealt to units in the area simultaneously. This means more damage is done as opposed to when all of 50 slingers snipe one or two enemies, with most of the projectiles being wasted.

 

I don't think it would have to be repetitive like in @Freagarach's video, perhaps individual volleys.

I may be wrong about this but maybe if the diff(https://code.wildfiregames.com/D1971) could be made into a mod for A25, we could all get together and test it!

The other proposals (especially (3)) sound good, but would likely require more development time.

[Nullus]

16 minutes ago, chrstgtr said:

That would be (1) or (3) as I describe above. I believe (1) is bad for the reasons k describe 

Yes, what I had in mind was a combination of the two. However, I disagree about the cons of (1). If anything, I think it would reduce snowballing. In general, during an attack, the attacking army is closely gathered into a smaller group, while the defenders are coming from all directions. The defenders could use this attack-area to more effectively destroy the attacking army, while it would be harder for the attackers to use, since the defenders aren't gathered into a group.

[wowgetoffyourcellphone]

4 hours ago, chrstgtr said:

• •  

Con: It will lead to a slow, sudden death of large armies. This will without a doubt lead to snowballing in 1v1s because large enemy armies will die almost all at once while the attacking army will lose basically no units. In team games, this will lead to a teammate's army being able to quickly wipe enemies even if the enemy army is large because units will be uniformly very injured. This will also make this type of attack

I don't think we can be absolutely sure of the cons until something is testable. 

[chrstgtr]

23 minutes ago, real_tabasco_sauce said:

yes attack-ground means the units are shooting an area, which is assigned by the player. The archers do not target, but instead the player does the targeting. For example: enemy has 30 pike and 30 skirm, I have 30 pike 30 sling. I would use attack ground to target the skirmishers because i know they will kill my pikes first. This should kill the skirmishers faster than individually clicking each one (because of avoiding overkill and not defaulting to shooting pikes).

The reasons to use this are:

-give the player some control over where arrows go, as opposed to the default of shooting the closest unit. This allows ranged units to shoot past melee for example.

-when attacking a lot of units that are close together, damage can be dealt to units in the area simultaneously. This means more damage is done as opposed to when all of 50 slingers snipe one or two enemies, with most of the projectiles being wasted.

 

I don't think it would have to be repetitive like in @Freagarach's video, perhaps individual volleys.

I may be wrong about this but maybe if the diff(https://code.wildfiregames.com/D1971) could be made into a mod for A25, we could all get together and test it!

The other proposals (especially (3)) sound good, but would likely require more development time.

I don’t think you are understanding what I wrote. Either units are targeted, which may or may not lead to many units being targeted all at once (see above), or individual geographies are being targeted—regardless of whether or not units are in those aimed at locations. Shooting at random geographies is pointless. Shooting at units is useful. How you target units is a different discussion, which I lay out above

[chrstgtr]

13 minutes ago, wowgetoffyourcellphone said:

I don't think we can be absolutely sure of the cons until something is testable. 

Some things are just simple math. If all units in an area are randomly targeted then all units will die more or less at the same time. If I have 10% more units then my units will have ~10% health left when all the enemy dies, which will occur more or less all at the same time. An army of 100 (even if it just has 10% health) is going to roll over any enemy with an army of 0. Barracks won’t be able to spam quickly enough to even make a dent

[wowgetoffyourcellphone]

9 minutes ago, chrstgtr said:

Some things are just simple math. If all units in an area are randomly targeted then all units will die more or less at the same time. If I have 10% more units then my units will have ~10% health left when all the enemy dies, which will occur more or less all at the same time. An army of 100 (even if it just has 10% health) is going to roll over any enemy with an army of 0. Barracks won’t be able to spam quickly enough to even make a dent

That is naive math (sorry) that doesn't take into account the chaos of battle. On an infinite scale, yes, the math is absolutely correct, but in the ebb and flow of the battlefield it won't work out that way. Players won't constantly be using the feature and players who do will still have to manually target or retarget their troops throughout the battle as the conditions change. There will be all kinds of gaps in usage and enemy troop movements to confound the "simple" math.  

[chrstgtr]

7 minutes ago, wowgetoffyourcellphone said:

That is naive math (sorry) that doesn't take into account the chaos of battle. On an infinite scale, yes, the math is absolutely correct, but in the ebb and flow of the battlefield it won't work out that way. Players won't constantly be using the feature and players who do will still have to manually target or retarget their troops throughout the battle as the conditions change. There will be all kinds of gaps in usage and enemy troop movements to confound the "simple" math.  

Sure, there are confounding factors. But for every enemy change the attacking player can make an appropriate reaction. What I described is more or less optimal play that will lead to pretty wonky outcomes. But we’ll see

regardless, unless the dev team wants to undertake a lot of work at the front end, we are going to need to pick one preferred choice and then test that option.

My preference is known 

[real_tabasco_sauce]

43 minutes ago, chrstgtr said:

Shooting at random geographies is pointless.

yes this is what I'm talking about. Why is it pointless and how have you already come to that conclusion? How about we test it to find out?

 

firstly the areas are not random, they are player specified. Targeting is done by the player.

secondly it could literally do more damage in some cases, so i don't see how that is pointless.

Edited January 2, 2022 by real_tabasco_sauce

[chrstgtr]

3 minutes ago, real_tabasco_sauce said:

yes this is what I'm talking about. Why is it pointless and how have you already come to that conclusion? How about we test it to find out?

 

firstly the areas are not random, they are player specified. Targeting is done by the player.

secondly it could literally do more damage in some cases, so i don't see how that is pointless.

Because if you are shooting aimless then many (and likely most) of your units are guaranteed to be useless. It is like sending your men into battle and then setting 20% of them to passive. 
 

more to the point, how is random shooting at an area. Better than targeted shooting at an area? You can achieve the same thing that you want with random targeting of units with an area (1) above, but not have any of the downside. 

[real_tabasco_sauce]

@chrstgtr

3 minutes ago, chrstgtr said:

Because if you are shooting aimless then many (and likely most) of your units are guaranteed to be useless. It is like sending your men into battle and then setting 20% of them to passive. 

@chrstgtrif you send your units to battle and snipe one ranged unit at a time with all of them you are wasting at least 80 percent of their value. (80 percent do damage to a unit that is already killed). These are rough numbers obviously since it depends on the unit.

do you now get why its powerful to spread out damage a bit, imagine this for crossbows!

[chrstgtr]

3 minutes ago, real_tabasco_sauce said:

@chrstgtr

@chrstgtrif you send your units to battle and snipe one ranged unit at a time with all of them you are wasting at least 80 percent of their value. (80 percent do damage to a unit that is already killed). These are rough numbers obviously since it depends on the unit.

do you now get why its powerful to spread out damage a bit, imagine this for crossbows!

You aren’t getting what I am saying: randomly targeting units within a defined area (the way a tower works) is always going to be better than randomly shooting projectiles within a defined area (imagine a tower, but half your projectiles go in the wrong direction)

 

I laid out four different ways units can be targeted and only one of those is what you are describing 

[real_tabasco_sauce]

12 minutes ago, chrstgtr said:

more to the point, how is random shooting at an area. Better than targeted shooting at an area? You can achieve the same thing that you want with random targeting of units with an area (1) above, but not have any of the downside. 

I am on board with this idea as I have already stated. My only issue is that it has not yet been developed, which is not the case for attack-ground.

[real_tabasco_sauce]

Just now, chrstgtr said:

I laid out four different ways units can be targeted and only one of those is what you are describing 

yes I already said random targeting in an area is fine too.

->

7 hours ago, real_tabasco_sauce said:

Both of these sound appealing to me, and perhaps they are not mutually exclusive.

 

[chrstgtr]

12 minutes ago, real_tabasco_sauce said:

yes I already said random targeting in an area is fine too.

->

 

I don’t get what you were disagreeing with then. 
 

but fine, the undeveloped option is (always) better than the option that is currently developed. My point is, we should strive for targeting of some type. If you want to implement some intermediate, transitionary option, fine. I don’t think it will be useful, but I’d be happy to be wrong I don’t have to use it if I am right. but development shouldn’t stop there, especially since we know a better option that just isn’t developed yet

Edited January 2, 2022 by chrstgtr

[real_tabasco_sauce]

1 minute ago, chrstgtr said:

I don’t get what you were disagreeing with then. 

No worries, i was just defending the possibilities attack-ground already brings.

I do think using attack-ground involves more opportunity cost (is it worth it to use or better to use the default behavior?) and will involve activity on behalf of the player (continually updating attack area and moving units).

It may impart a higher skill ceiling.

 

@Freagarach can what you wrote be used in A25, or would it need to be updated?

you and the other devs put a lot of work into this, so I would like to give it a try!

 

[LetswaveaBook]

3 hours ago, Nullus said:

Pyrogenesis doesn't actually simulate the physical trajectories of projectiles, so it has trouble detecting if a projectile hits a unit if the unit wasn't specifically targeted. That would make most of the arrows miss harmlessly, even if they landed in the middle of a group.

From what I have seen, this is indeed observable. From my experience it is that if javelins are shot at a dense group, it does not mean something is hit (which might seem counterintuitive).

 

3 hours ago, Nullus said:

The main area in which I can see this being useful is with ranged units with splash damage, such as bolt shooters.

Could this be the solution for the previous situation? If ranged units had a really tiny splash damage(that is so small it only hits the intended target), then they might have more realistic accuracy. Though I think this might become computationally heavy.

2 hours ago, Nullus said:

However, I disagree about the cons of (1).

I also disagree about the cons of (1) and I think it might be a good solution if it was toggable (player can chose between "current" targeting or tower targeting). Tower targeting means pikemen get hit as often as the ranged units, so the ranged units die earlier.

1 hour ago, chrstgtr said:

Some things are just simple math

Some things are complex math. If a garrisoned fortress distributes 23 arrows over a group of units, they die all simultaneously with current tower targeting. If 23 towers shoot 1 arrow and distribute the next volleys, then that is not the same. Also tower targeting only targets the units in range as far as I am aware.

41 minutes ago, chrstgtr said:

the undeveloped option is (always) better than the option that is currently developed.

Again, a merit of (1). @wowgetoffyourcellphone and @real_tabasco_sauce said something needs to be tested. I guess (1) is very testable.

[LetswaveaBook]

I made a mod, where I gave archers buildingAI or tower targetting. In these settings, I set up 15 archers+pikes vs 15 skrimishers+pikes. The side with the archers won(13 archers survived, but there appears significant randomness and sometimes the side with the javelins wins). I also set up 5 archers+pikes vs 5 skrimishers+pikes and this time the side with the skirmishers won narrowly (4 pikemen won with a total HP of 103, but that is only because the archers did not run away after the javelins died).

 

So (1) will shift the balance and it does enough to counter pikemen. Archers+pikes won't be OP, it is just that pikes lose some of their usefulness if they no longer absorb all damage.

 

TemplateModwitharchbuildingAI.zip