Is Barrage broken?

[Roxlimn]

Accuracy doesn't do that, though, but its comparison with Barrage 2 is based on another function entirely, so it's not normally considered directly comparable.

 

[OTAKUjbski]

So the question becomes, is the siege unit's life more valuable than the damage I need done?

Also stated as ...

"Is the additional damage done worth the Siege unit's life?" (-- which is essentially at the crux of my newfound dislike of Barrage.)

@ CivCorpse: RE: Mixed stacks:

Are mixed stacks necessary?

I think they may serve only to unnecessarily complicate the test.

Regardless of what the composition of a defending stack is, the goal of any attack is to weaken the [top] defender(s) to give your redlining units better Combat Odds.

In some cases, this might be weakening Formation Pikes so your Knights can move in. In others, Shock XBows so your Macemen can move in.

So whether the Siege units are softening up for a Knight with 10% Combat Odds or a Maceman with 10% Combat Odds, the goal is still the same even though the attacking/defending units may vary.

----

The reason I ask is because I'd like to use a "simple" test which I can run manually at high speed.

What usually happens against mixed stacks is that it comes down to finding the attacking unit with the best odds. If the Pike is "too damaged", for instance, and a Maceman becomes the best defender, it might be to the attacking stack's advantage to use an XBow instead. (But doing that manually after every attack would significantly increase the time necessary to conduct a viable amount of tests.)

I also think introducing mixed stacks and more promotions increases the likelihood of watering down the test results -- yielding more inconclusive data.

----

So I also pose the counter-question, "Do congeneric stacks yield less valuable data than mixed stacks?

 

[CivCorpse]

Also stated as ...

"Is the additional damage done worth the Siege unit's life?" (-- which is essentially at the crux of my newfound dislike of Barrage.)

Well according to the tests I ran, the little bugger pretty much died regardless of whether he had C1 &C2 promotions or if he had Barrage 1 &2 promotions. Once you get past swords/axes/spears Cats pretty much die when they attack. They just don't have the base strength to to survive. Shock/Cover or even C2 with formation for the knights. i even tried with drill2. They die. So if you're going to lose a unit, get your money (or hammers) worth. Just understand going in....they're gonna die. Some survive, but such a low percentage that it's not something you plan on. More survive as you through more at the enemy and weaken the defenders, so maybe you want to start adding combat if you have lots of cats. usually when knights are in the game cats are on their way out. They cost too much to upgrade so I just save on some future units maint costs and get something back for my investment.

 

[Roxlimn]

It's actually even more complicated than that. In an actual stack-to-stack fight, you have to figure out which unit or units are liable to produce the greatest amount of damage to the offending stack in general and then build shock units to take out that unit's counter unit.

In CivCopse's presumptive stack, Knights are the best general all-purpose attackers because Shock promotions of Knights (with a few Formation Knights or Ballista Elephants) will at once take out the Pikemen and also dismember the stack on the whole.

This is probably due to the CR Macemen content, which could have been devoted to more stack defense, or it could just be that the stack is just too small to present a credible threat.

A crossbowman here would not be very good because it runs against longbows AND Knights, against whom its odds are not very good, whereas Shock Knights will generally have good odds overall against the stack, barring the Pikemen and Knights (the latter of which can also be dealt with using Knights).

Conversely, a Maceman defensive stack would work better against a stack with a greater amount of Pikemen (take out Crossbows first) and a Pikeman stack is devastating against a stack with high numbers of mounted units (taking out Macemen first).

In general, I find Catapults unsuitable to the purpose of weakening attack stacks because they generally involve too much unit death which then generates WW for you and XP for the AI. You can dismantle a stack with less unit death using specialized unit matchups and promos for the purpose of manipulating attack odds rather than suiciding Cats en masse to weaken the stack universally.

Mounteds, are, of course, the best units for weakening attack stacks because they have the flanking ability, after which you can present fortified units in various guises for the attack stack to vainly throw itself against.

 

[InFlux5]

a single CR3 Trebuchet's lvl1 collateral damage increases the chances of a CR3 Maceman surviving by +29.6% while a B3 Trebuchet's collateral damage increases the chances of a CR3 Maceman surving by +31.9% (only +2.3% better).

I almost never use Barrage because of the way collateral damage works. The fact that collateral damage can never kill a unit, only reduce it to a certain level, means that, given enough siege units, Barrage becomes useless. Because the collateral damage will eventually "bottom out", your concern then becomes trying to maximize the chances that your siege unit will actually kill the defending unit. Collateral damage becomes irrelevant.

So even without the math, Barrage doesn't make sense in most situations. Nonetheless, if I'm short on siege units, and I know that the few I have will almost certainly die, I may want that extra 2.3% collateral, just to maximize the chances for my infantry to survive. The higher the ratio of defending units to my siege units, the more attractive Barrage becomes.

 

[CivCorpse]

It's actually even more complicated than that. In an actual stack-to-stack fight, you have to figure out which unit or units are liable to produce the greatest amount of damage to the offending stack in general and then build shock units to take out that unit's counter unit.

In CivCopse's presumptive stack, Knights are the best general all-purpose attackers because Shock promotions of Knights (with a few Formation Knights or Ballista Elephants) will at once take out the Pikemen and also dismember the stack on the whole.

But they won't take out the pikes first. The stack will use the C2 knight to defend which is stronger than a c1 shock knight...the odds are against you.
What unit do you magically possess that circumvents the whole rock/paper/scissors combat formula and allows you to pick who defends against you?

Conversely, a Maceman defensive stack would work better against a stack with a greater amount of Pikemen (take out Crossbows first)

Again I ask what did you use to take out the x-bows? Knights? the pikes killed them. Shock knights? the c2 knights killed them. Cover maces? the c2 knights killed them

and a Pikeman stack is devastating against a stack with high numbers of mounted units (taking out Macemen first).

Who killed the maces? Knights? Pikes killed them. Shock knights? The c2 knights killed them.

Mounteds, are, of course, the best units for weakening attack stacks because they have the flanking ability, after which you can present fortified units in various guises for the attack stack to vainly throw itself against.

If you give them flanking promotions you have to give up all your specialized counters. They do not have an automatic retreat chance, The flanking ability they start with is for damaging siege units in a stack. That's it.

If you attack a well balanced stack the initial attackers are going to suffer some pretty heavy losses. If you have to lose units, let them be the cheapest ones you have. ones that are near obsolete.
You're talking about softening up stacks using the most expensive units available at the time. You're lucky the attacker doesn't have ivory. I'd love to see what you do against a stack with War elephants.

 

[Roxlimn]

CivCorpse:

But they won't take out the pikes first. The stack will use the C2 knight to defend which is stronger than a c1 shock knight...the odds are against you.
What unit do you magically possess that circumvents the whole rock/paper/scissors combat formula and allows you to pick who defends against you?

Well, Ballista Elephants, for one, but I'm quite sure you knew about that. There's nothing magical about a Ballista Elephant, though.

Generally speaking, in a stack of that composition and promotion, the Pikemen will defend against Knights first. You can ameliorate the damage with Shock or multiple Combat promotions, or you might choose to go flanking instead. It depends on the terrain.

A Knight doesn't have to win to render the Pikeman useless for stack defense. he just has to damage the Pikeman enough so that it's got bad odds against another Knight.

Again I ask what did you use to take out the x-bows? Knights? the pikes killed them. Shock knights? the c2 knights killed them. Cover maces? the c2 knights killed them.

It's a little hard to explain, especially if you persist in thinking about individual units rather than stacks.

If you give them flanking promotions you have to give up all your specialized counters. They do not have an automatic retreat chance, The flanking ability they start with is for damaging siege units in a stack. That's it.

The Horse Archer, the Numidian Cavalry, Camel Archer, and Cavalry, I believe, have inherent retreat odds. Knights don't, IIRC, and I'm not sure about Cuirassers.

I'm sure I missed a few, but as you surmised, I was actually talking about their effect on siege weaponry.

If you attack a well balanced stack the initial attackers are going to suffer some pretty heavy losses. If you have to lose units, let them be the cheapest ones you have. ones that are near obsolete.
You're talking about softening up stacks using the most expensive units available at the time. You're lucky the attacker doesn't have ivory. I'd love to see what you do against a stack with War elephants.

It's a zero sum game. A stack with War Elephants is going to have that much less Macemen. A Landsknecht with the Shock promotion ought to do quite well against that general composition stack, although to comprise a specific plan for dismantling it, I would have to know the specific quantities of the units and what their promotions are - details you can easily research on the map, of course.

More than hammers here is WW. The plan is to specifically overwhelm the "best defenders" with high numbers of their counter units with anti-them promotions or some other way of evening the odds, and then proceed to attack at advantage through the defensive gap. Of course, it's the AI so it's not like it brings a well-balanced stack all the time, and strategic resource advantages also open up defensive gaps.

It's no different from attacking with suicide Catapults, actually, except that you lose less units. The hammer output may be more, but I find this kind of defense lowers my WW and lowers the general level of the AI's units. These advantages are harder to quantify, but my war odds have increased using this method rather than using suicide Cats.

I don't sacrifice Catapults on active defense because I prefer to use them for suicide attacks in offensive stacks. There are many reasons why I prefer to do this. Active defense advantage using defense stack manipulation is only one of them.

 

[CivCorpse]

CivCorpse:



Well, Ballista Elephants, for one, but I'm quite sure you knew about that. There's nothing magical about a Ballista Elephant, though.

You're talking UU's which shouldn't be used in an analasys of generic catapults.

Generally speaking, in a stack of that composition and promotion, the Pikemen will defend against Knights first. You can ameliorate the damage with Shock or multiple Combat promotions, or you might choose to go flanking instead. It depends on the terrain.

No, against a Shock knight, the C2knight will defend. Try it in WB. I just did. If you choose Flanking promotions the Pikeman will defend and then the knight dies or retreats. Which requires even more units and you're risking high cost units as opposed to cats which are past their prime.

A Knight doesn't have to win to render the Pikeman useless for stack defense. He just has to damage the Pikeman enough so that it's got bad odds against another Knight.

Again, That is why you use catapults. They do the same job but cost about 1/2 as much and they damage EVERYONE not just one pikeman

It's a little hard to explain, especially if you persist in thinking about individual units rather than stacks.

I am discussing stacks. You're then one that keeps trying to use specific counters for specific units when the comupter will pick the best defender rather than allow Shock units to attack Pikes or Pikes to attack mounted when the stack has a better defender.

The Horse Archer, the Numidian Cavalry, Camel Archer, and Cavalry, I believe, have inherent retreat odds. Knights don't, IIRC, and I'm not sure about Cuirassers.

A. You're talking UU's again in some cases.
B. HA's will barely scratch a pike
C. Cavalry/Cuirassers are from a later period and at that point you have muskets and rifles too.

A Landsknecht with the Shock promotion ought to do quite well against that general composition stack, although to comprise a specific plan for dismantling it, I would have to know the specific quantities of the units and what their promotions are - details you can easily research on the map, of course.

A. You're discussing UU's again
B. I told you the exact composition of the stack in my original test.

 

[OTAKUjbski]

@ CivCorpse: RE: Mixed stacks:

It appears my suspicions are well-founded ...

Are mixed stacks necessary?

I think they may serve only to unnecessarily complicate the test.

Regardless of what the composition of a defending stack is, the goal of any attack is to weaken the [top] defender(s) to give your redlining units better Combat Odds.

In some cases, this might be weakening Formation Pikes so your Knights can move in. In others, Shock XBows so your Macemen can move in.

So whether the Siege units are softening up for a Knight with 10% Combat Odds or a Maceman with 10% Combat Odds, the goal is still the same even though the attacking/defending units may vary.

----

The reason I ask is because I'd like to use a "simple" test which I can run manually at high speed.

What usually happens against mixed stacks is that it comes down to finding the attacking unit with the best odds. If the Pike is "too damaged", for instance, and a Maceman becomes the best defender, it might be to the attacking stack's advantage to use an XBow instead. (But doing that manually after every attack would significantly increase the time necessary to conduct a viable amount of tests.)

I also think introducing mixed stacks and more promotions increases the likelihood of watering down the test results -- yielding more inconclusive data.

----

So I also pose the counter-question, "Do congeneric stacks yield less valuable data than mixed stacks?"

So I'd like to re-pose the questions before I put in the manhours testing tomorrow:

• Are mixed stacks necessary?
  [*]Do congenerous stacks yield less valuable data than mixed stacks?

 

[CivCorpse]

After someone told me that collateral damage is determined before the combat round is initiated and is a constant # only modified by promotions like barrage and drill. Barrage incresing it and drill2+ decreasing it.
As to your question are mixed stacks needed? It depends on what you're trying to test.
If you are trying to determine if composition effects damage from collateral, then only in what each unit suffers taken individually. Pikes will take the same amount of damage regardless of the composition of the stack.
I am not sure what you are asking about. If you are trying to determine the effect it has on the rock/paper/scissors component of Civ warfare. It does in the fact that weaker base strength units seem to lose a larger percentage of their strength/hit points. In doing so, specialised counters such as pikes and x-bows lose more than straight troops that have inherant high attack strengths such as maces and knights. Pikes seem to suffer the worst because for every point they lose, it is a point that cannot be doubled.
example Pike loses 1 point of strength due to collateral or direct combat. It now has a base of 5. Or 10 with the formation promotion. A loss of 2 points for combat purposes. Or 16% of it's combat value vs mounted which is about the only time it will rotate to the top as best defender.
A knight loses 1 point which is 10%.

Regardless of what you attack with, the computer will select the best avialable defender. In the case you describe attacking with a x-bow will usually pop the defending x-bow. The purpose of collateral is to reduce the stregnth of all units en masse so that when you begin attacking the best defender is still weaker than your attacking unit. Knights will still draw pikes as defenders but they are significantly weaker to the point that the knight has the advantage. What Roxlimn fails to gras[p is that by assaulting immediately with specialized counters. You originally face other units until the unit you are trying to get at is the best defender. That can lead to a large amount of losses of expensive specialty troops. Unless you have a UU that gives an advantage that is just poor tactics. I would rather sacrifice slow moving weak catapults that are the cheapest unit available at the time than throwing highly mobile, expensive strong knights into near certain death.

Back to the question of whether mixed stacks matter. They do matter in that it takes more collateral to damage all the units to the point that your best attacker has better strength than the best defender against it no matter what the computer picks. Using low base strength units that gain high bonuses vs certain classes is not the best for assaulting. Because most often they will face a different combat class. And now they are relying on a low base strength. After collateral damage. it is best to just begin with whatever your strongest base strength unit of the time is and use brute force.

 

[DanF5771]

Because that's how the code does it ... from CvUnit::collateralCombat:

int iCollateralStrength;

...

iCollateralStrength = ((((getDomainType() == DOMAIN_AIR) ? airBaseCombatStr() : baseCombatStr()) * collateralDamage()) / 100);

If the developers would drop the silly "/ 100" and reserve that until the end of the formula, then Barrage wouldn't be losing as much damage.

interesting - so this means Arathorn's example in his strategy articel is wrong!?

Example: A catapult (with Barrage I and Barrage II) is targeting a longbow with 71 hps in a city on a hill for collateral damage. The catapult’s adjusted strength is 7.5 (5 * 1.5) and the longbow’s adjusted strength is 6 – the hill defense, the city defense, the hps just don’t matter. The catapult does 11 damage to the longbow.

 

[vicawoo]

If barrage can save you from having to use a (or n) catapult(s), then to do better, city raider will have to average saving an extra catapult(s). So maybe use it if you get a +20% win rate (if you need 5 attackers) from city raider/combat, otherwise use barrage.

You might have to add in the odds or the replacement swordsmen dying, though.

From another perspective, this means cannons are somewhat over rated for collateral. You would get much more collateral out of 2 unupgraded catapults. So for collateral purposes, think of barrage catapults as slightly weaker cannons.

 

[Diamondeye]

Yargh! This looks advanced and I died in the spreadsheets. I personally never use Barrage and I do not intend to. Now go play civ, silly OTAKU.

No, seriously, I think it's nice you care to do all this experimenting and maths and spreadsheets and everything. I thought of something that might help Barrage a bit, if a condition was something like this:

The defender has only a minimal edge against the nonsiege attackers

This could sway the statistics for Barrage units, since those few extra collateral dmg will be able to push the defender below the attacker - since the difference in odds is most critical near equality. This would work even better if there was too few siege units - since the CR units will have a tendancy of pushing the statistics by reducing several (in your examples 6v6, sometimes all) of the defenders to minimum.

Try:

A city with 0% culture defended by CG I Rifles without fortificaiton.
Use numbers like:
40 attacker rifles (unpromoted)
5-8 attacker cannons (Barrage/CR)
40 defender rifles (CG 1)

this might yield some interesting results. That said, I am sleepy. Go ahead and try it out And Siggboy, nice idea about the simulator.

 

[CivCorpse]

Yargh! This looks advanced and I died in the spreadsheets. I personally never use Barrage and I do not intend to. Now go play civ, silly OTAKU.

No, seriously, I think it's nice you care to do all this experimenting and maths and spreadsheets and everything. I thought of something that might help Barrage a bit, if a condition was something like this:

The defender has only a minimal edge against the nonsiege attackers

This could sway the statistics for Barrage units, since those few extra collateral dmg will be able to push the defender below the attacker - since the difference in odds is most critical near equality. This would work even better if there was too few siege units - since the CR units will have a tendancy of pushing the statistics by reducing several (in your examples 6v6, sometimes all) of the defenders to minimum.

Try:

A city with 0% culture defended by CG I Rifles without fortificaiton.
Use numbers like:
40 attacker rifles (unpromoted)
5-8 attacker cannons (Barrage/CR)
40 defender rifles (CG 1)

this might yield some interesting results. That said, I am sleepy. Go ahead and try it out And Siggboy, nice idea about the simulator.

LOL, about 3 pages ago we all sort of agreed that for attacking cities CR wins hands down. Now we were trying to figure out which is better in the open field. I am a firm believer in Barrage in the open field

 

[TheMeInTeam]

For field battles, has this been tested yet? I get the feeling that combat is better to inflict actual damage when odds are bad (to the top defender), while barrage will win out when you already have winning odds and thus want to do more damage (like cannons vs knights or xbows outside a city or something)

 

[OTAKUjbski]

I'm testing Combat vs Barrage in the open field tonight and tomorrow ...

... I should have an even more confusing spreadsheet for you guys tomorrow afternoon.

 

[CivCorpse]

For field battles, has this been tested yet? I get the feeling that combat is better to inflict actual damage when odds are bad (to the top defender), while barrage will win out when you already have winning odds and thus want to do more damage (like cannons vs knights or xbows outside a city or something)

Already did a test. Combat does less collateral damage and they die. Barrage does more collateral and they die.

 

[siggboy]

Executive Summary:

• Barrage is a very bad promotion overall, and should not be used for city raiding in particular.
• City Raider is the promotion line of choice for siege weapons that are on the offense (capturing cities).
• Combat is the promotion line of choice for siege weapons that are on the defense.

There's some pretty good evidence that the best promotion to raid cities with is... city raider.

That's what line 2 of my exec summary claims. You quoted line 3 with your above reply. What gives?

But I haven't seen any proof that barrage is useless in the field.

It's not useless against cities either. Just so weak that increasing your survival figures by going CR instead is always the better choice.

Now since in the field we can't use the very strong CR promotion line, Barrage needs to be re-evaluated. It might turn out that picking Combat + X for field siege does not make a noticeable difference for overall success/survival rate, and in that case Barrage might be better -- at least you'd get some additional collateral damage inflicted.

Basic siege is so powerful that you could leave them unpromoted and barely even care about it. If they were more balanced it would be painfully obvious to anyone how worthless barrage is.

Of course you'd have to care about promoting your siege weapons because you can greatly enhance their survival rate that way. Regarding collateral damage done you wouldn't have to care, since Barrage does not increase it enough.

interesting - so this means Arathorn's example in his strategy articel is wrong!?

No it's not. It doesn't matter if you multiply by 150 and then divide by 100 or instead multiply by 1.5. The C++ code deals with percentage values and therefore they have to divide by 100. Arathorn uses decimal factors in his examples so he doesn't divide by 100.

And Siggboy, nice idea about the simulator.

Well it will probably be a few days longer until I come out with something, because in order for it to become more than a quick hack I need to parse the XML files and use that information so I get correct behavior for all the units.
It is not a "hard problem" from a technical perspective, just a lot more work than simulating a "cats vs. longbows" situation and nothing else.

However, a general-purpose simulator could also be used to answer questions unrelated to siege, for example how much hitpoints the Drill promotion line saves you on average etc. These are questions that have been posed in other threads but not answered satisfactorily so far.

 

[OTAKUjbski]

No it's not. It doesn't matter if you multiply by 150 and then divide by 100 or instead multiply by 1.5. The C++ code deals with percentage values and therefore they have to divide by 100. Arathorn uses decimal factors in his examples so he doesn't divide by 100.

Well ... actually ...

Yes, Arathorn's example is wrong!

Example: A catapult (with Barrage I and Barrage II) is targeting a longbow with 71 hps in a city on a hill for collateral damage. The catapult’s adjusted strength is 7.5 (5 * 1.5) and the longbow’s adjusted strength is 6 – the hill defense, the city defense, the hps just don’t matter.

This first part isn't inherently wrong, because the Catapult's adjusted iCollateralStrength should be 7.5. If we assume he simply omitted the "floor" thinking everybody knows to floor it here, then we would naturally just read it as "7" in that we knowingly automatically and subconsciously drop the ".5".

Unfortunately, either not at the time of the writing or unknown to Arathorn, the value IS floored at that moment from 7.5 to just 7 BEFORE being applied to any other part of the collateral damage calculation.

The catapult does 11 damage to the longbow.

So because Arathorn used the incorrect, unfloored value of 7.5, Arathorn's quoted value of "11" is wrong. The amount of damage done to a Longbow by a Barrage II Catapult is 10!

 

[r_rolo1]

Check the dates OTAKU..... when Arathorn published his article the SDK was not yet out, so I suppose that his article was based in a enourmous and tiresome reverse engineering... Not surprising that his work has some errors