Efficient use of siege engines

[yanner39]

I learned a while back that siege engines are important when composing your stack(s) and going to war. I noble, although I was/still not good at it, warring is relatively easy if you bring siege. I would give my siege the CR promo and attack. Suiciding cats was no big deal. Also, I tend to produce alot more than I may need and therefore, delay going to war sooner.

I am wondering if there is a better way to use/manage my siege as I move up levels. Should a portion of my siege engines be for bombardment of the cities, and the others have the CR promo? Before I send cats/trebs in to attack city units, should I look to see if collateral damage is even needed, considering the other units in my stack?

I understand suiciding cats to cause collateral damage, but it seem like I wasting hammers (although I am getting a city(ies)). And also, replacing the suicide cats is difficult since they take so long to join my stack. Should I just be stating a new stack and move them to a different part of the enemy territory?

It seems my wars last longer than they should because I lose more units than I should. Maybe I am not protecting them properly?

 

[Jerrymander]

I'm no expert on warfare.

Generally, I try to keep a 1:1 ratio of siege engines to city assault troops and I use the engines to both bombard the city defenses and then attack, giving them all CR promotions. I find, personally, that the promotions that improve Collateral Damage and Bombardment are less helpful than the City Raider promotions. With CR, your siege units have a better chance to survive the attack and gain XP. I also attack with only the siege engines until the maximum damage has been dealt. This makes my city assault troops' lives easy. I'll lose at most two catapults to 60%-70% odds in the first few attacks, but the rest are often 93% and up.

An exception to this is later warfare with Trebuchets, where I'll often lose 3 or 4 because my enemy has about 50 Macemen in a city.

 

[Ray Patterson]

The number of siege engines you need depends on your tech level compared to your enemy, on how many units they have and whether they're protective. I try to make sure that my normal attackers get at least a 60% chance of surviving. If they can't, I need more siege. Exception: if you are attacking a civ with less than 4 or so cities, so that losing your attackers is not a problem as conquest is already almost finished.

The idea is to use the siege suiciders to make sure that most of your other troops survive. If you are outteching your enemy (very possible at noble) then suicide siege are unnecessary and you can use them only to bring down the defense bonus.

 

[TheMeInTeam]

Going up vs castles is hell pre-cannons/late renaissance.

Trebs are solid city attackers with CR II vs everything before CG muskets (and aren't terrible there), the problem is bombardment time and keeping AI stacks off of you if it's contemporary war.

"Enough" is something you feel out over time. However, 50% siege isn't a bad rule of thumb initially, just remember that if you're keeping cities you have to leave garrisons in them (usually), and to keep enough counter-units in your stack so that you don't get curb stomped by a set of knights, xbows, or maces and you'll be fine.

 

[mariogreymist]

Another way of conserving siege engines is to bring along some flanking mounted units in your stack. You can use them to attack before the siege, usually losing about half of them, but damaging the top defenders and greatly increasing the retreat odds for your siege before collateral. And you only need to damage the toughest defenders before beginning with the siege, holding the rest of your mounted as reserves and to mop up after collateral. Not infrequently you find the AI has left 1-2LBs with a bunch axes/maces and spears/pikes in a city...if you damage the LBs with some FlankingII HAs or Knights, then your cats will have a much higher survival rate. This is even more true when your enemy gets muskets, as they need to be built and cannot be upgrades, so you often see only one at the top of a defender's stack. Damage it with some Flanking troops, and your life gets much easier.

 

[drlake]

While CR seems a better promotion for assault Cats, keep in mind that it is worth having a few Cats with bombardment for defensive purposes, to attrit an attacking AI SoD.

 

[Grashopa]

If your main city attacker will get 90% odds after bombarding the walls to 0, then you want 2 promo bombard catapults. You only need enough to bombard to 0 in one turn, so you'll need more if they can get to engineering. Pre-engineering thats what 5-7 bombard cats? If you don't get 90% odds you need to sacrifice a catapult or two until you do so you'll need more.

Attackers you need only 9 or 10. You simply need more than the AI has in any one stack and enough to leave 1 in each city you take. Expect to lose less than 1 per city in a good war.

If you are up against a protective civ with longbows then your problem becomes damaging the CG2 longbow as your siege will lose to that longbow without damaging it though it will do collateral to others. There having city raider helps some, trebs are better and you'll just have to suck up losing a city attacker taking him.

 

[Baldyr]

Although City Raider isn't a bad choice for siege units, its not flexible at all. You'd wanna use siege units on defense also - and for counter-attacks. There the Barrage promotions are more effective.

The way I usually do it is that I don't actually choose promotions for units until they are used for attack or placed (more or less permanently) on defensive positions (or assigned as escort units). This way I can still use my siege units in the field, if I need to.

Once my stack reaches an enemy city (on the offense) I can decide whether or not I need to give the Accuracy promotion to one or a few siege units in order to get those defenses down on the first turn. Only once the actual suicide run begins will I assign City Raider promotions to the units concerned. Note that the City Rainder and Barrage promotions do nothing for defense, and its also good to save up the healing power of promotion - in case your units get damaged en route.

Chances are, however, that the city defenders will be battered beyond the limits of my siege units, and not having set promotions to all siege units still preserves some flexibility for my stack. If the enemy counter-attacks and in turn besieges the city I just captured, I can promote those last unpromoted and full strength siege units to Barrange and deal collateral damage in order to protect my own weakened stack that now occupies the city.

So not setting all promotions at once gets you some flexibility and gives you more tactical options.

 

[drlake]

While CR seems a better promotion for assault Cats, keep in mind that it is worth having a few Cats with bombardment for defensive purposes, to attrit an attacking AI SoD.

Grr, I meant Barrage. Brain fart got me...

 

[Jerrymander]

Although City Raider isn't a bad choice for siege units, its not flexible at all. You'd wanna use siege units on defense also - and for counter-attacks. There the Barrage promotions are more effective.

This is a good point.

 

[Nkat]

Good information!

 

[yanner39]

Thanks everyone. I guess alot of what I was doing was ok. I was looking at a walkthrough recently (mind you, the player was about a 100 times better than me) and his early war without military overkill seemed to have gone very smoothly with minimal losses. Alot of good info.

@ Baldyr: Your point about sitting on the promos until you reach the target city is very well taken. I usually promote the unit once it's produced. When I reach the target city, I could sometimes use the +8% damage the Accuracy promo would give me.

There are times where I waste waay too many turns bombarding the damn city.

@ Grashopa: Yeah, I really should check the % of my attackers more often. I bombard to %0, ALWAYS suicide a few cats, then send in my CR attackers. I never bother to check if I even need to cause collateral damage.

My wars are so inefficient and long and my use of siege engines is a big reason.

 

[vicawoo]

You only need about as many non-siege units as there are defenders, plus some spares/defenders. Common wisdom is to go city raider over barrage for attacking cities. Stronger attacking catapults means less catapult casualties means hammers saved.

But I've been doubting this common wisdom since they changed barrage. Ideally someone should do a test against n defenders (say longbows and some percentage of siege) with as many level 2 units as they need, and see what approach minimizes hammer losses/injured siege.

The combat breakdown:
bombardment. injured siege perform just as well as full health siege, they just won't heal. This is why accuracy siege aren't necessary, just used your injured siege, unless they have lots of flanking.
Collateral damage. Either with your cr or barrage siege. Due to how withdrawal works, most of your siege will die or be left with almost no hp.
Further siege attacks. Your siege cannot do anymore collateral, but you will can damage units to 25%. Yes it gets experience, but it's probably an unnecessary risk of your hammers.
Cleanup. Your elephants/swords/whatever kill off the weakened units, hopefully taking minimal losses.

Afterwards, you usually move into the city and heal your units. However, even with a medic 3 unit, your siege can take 3 to 4 turns to heal, which coupled with the 1 tile stack movement of medieval warfare, and walls/castles, is unacceptably slow. Your cleanup units can heal, with the help of promotions, while your siege bombards those walls.

 

[VoiceOfUnreason]

But I've been doubting this common wisdom since they changed barrage. Ideally someone should do a test against n defenders (say longbows and some percentage of siege) with as many level 2 units as they need, and see what approach minimizes hammer losses/injured siege.

Funny how there's always more people suggesting that more experiments are necessary than there are people willing to do the work. I'll just point out that most of us who dig into collateral have already done more experiments than are fun.

Those of you who missed the earlier excitement may want to review the Collateral Thread of Death.

Collateral damage. Either with your cr or barrage siege. Due to how withdrawal works, most of your siege will die or be left with almost no hp.

Note that here too, injured siege units deliver as much collateral damage as their healthy counterparts.

...

I usually find that there are a few additional stages
1) Drop defenses
2) Collateral Damage
3) Can opening (gotta kill that top defender)
4) Siege killing - as a rule, mounted units produce better results against siege units than other siege units do. Horses for courses.

At this point, the defense is pretty well shredded, so you have a bit of choice about how you want the remaining XP divided. You can continue to use siege as attackers (not scoring any collateral damage), and then use your other units for the kill, or you can kill immediately.

(Side thought: one of your options in here is to attack with an unpromoted unit, and then use a promotion to heal afterwards).


If we consider the measure of a siege unit to be the amount of collateral damage it delivers during its lifetime, we may be able to make some interesting estimates. Consider a simple case where we don't heal. This implies that your improved survival rate is your improved collateral damage (math not shown).

As Barrage 1 improves your collateral by 20%, this would imply that you would need the CR promotion to improve your survival odds by more than 20% to be effective.

What's wrong with this guideline? Clearly it suffers from the fact that Barrage isn't necessarily 20% better than without, but you can go back and work out the right number there if you care to.

It also suffers from the fact that improved survival odds implies improved expected damage against the primary defender. So we might want to work that out too. This can be complicated by the problem that the improved damage will be a function of how badly wounded top defender already is.

If you think that siege healing is more appropriate to model, then you need a 16% better survival rate to match a 20% collateral promotion, with appropriate adjustments for the other factors.


My guess for an easy mnemonic is: as long as the top defender is undamaged, use barrage.

 

[vicawoo]

Funny how there's always more people suggesting that more experiments are necessary than there are people willing to do the work. I'll just point out that most of us who dig into collateral have already done more experiments than are fun.

Those of you who missed the earlier excitement may want to review the Collateral Thread of Death.

Funny how I think I posted in that thread, and that every suggestion I've ever made I've ended up doing on my own and posting in my own uniquely incomprehensible way. It's also funny that you list a thread were almost all of the tests were pre 3.17, and the post-3.17 posts were predominantly testing how barrage had changed as opposed to it's effectiveness.
I've been meaning to do this for a few months now, but I'm more comfortable with code and numbers than worldbuilder.

The predominant difference in the analysis is accounting for non-siege hammers lost and the different strategies involved. Maybe you're fine withe leaving a strong stack defender and sacrificing a few nutcrackers as long as the rest of the stack is weakened.

Good suggestion on using damaged siege for the first few attacks. They're going to die anyway, right?

 

[Jhoosier]

I'd like to reiterate what one person said upthread, and doesn't seem to be mentioned otherwise. Mounted units with flanking promotions are a very efficient extender of siege engine life. It's even possible they can be your entire siege stack, if you're doing an early chariot rush or nobody's made it to rifling yet and you have cuirrassiers. (and if the AI was foolish enough to not build sufficient spears/pikes).

I'm excited about this because it worked well in my last game as Russia, cossacks are good fun, even against riflemen. Though my chariot rush against Sitting Bull failed miserably using this tactic, though so did a mixed force approach (protective and UU at that era means it gets ugly).

It pays to conserve CR-promoted siege by using CR3-promoted attackers or flanking mounted if you have a particularly hard nut to crack. I currently play on Prince, so this may not be valid at Monarch (since I never live long enough to find out).

 

[yanner39]

Another way of conserving siege engines is to bring along some flanking mounted units in your stack. You can use them to attack before the siege, usually losing about half of them, but damaging the top defenders and greatly increasing the retreat odds for your siege before collateral. And you only need to damage the toughest defenders before beginning with the siege, holding the rest of your mounted as reserves and to mop up after collateral. Not infrequently you find the AI has left 1-2LBs with a bunch axes/maces and spears/pikes in a city...if you damage the LBs with some FlankingII HAs or Knights, then your cats will have a much higher survival rate. This is even more true when your enemy gets muskets, as they need to be built and cannot be upgrades, so you often see only one at the top of a defender's stack. Damage it with some Flanking troops, and your life gets much easier.

So I would use the cats to reduce defenses to 0%, send in my mounted flanking units to soften up city defenders, then attack the city with siege for collateral damage to further soften up the enemy?

 

[mariogreymist]

So I would use the cats to reduce defenses to 0%, send in my mounted flanking units to soften up city defenders, then attack the city with siege for collateral damage to further soften up the enemy?

Yep. If you have a few top defenders, say CGIII LBs/muskets against a protective leader on a hill, your trebs will have some difficulty damaging the top defender. A single flanking II knight or Curassier (HAs vs LBs on flat land can work) will be able to damage the top defender and have a reasonable chance of survival where a CRIII treb would likely inflict collateral but not damage the top guy. After those first 1-3 tough defenders are wounded, the siege can go to work with collateral. I find this to be the best method (in terms of efficiency) of "can opening."

 

[VoiceOfUnreason]

I lied: I did a little experimenting with this.

This test, shows a little bit about swords vs longbows.

The long bows are in a city, on a hill, surrounded by a moat.
The swords are carrying a CRI promotion.

If we just attack with swords, we find that we need a little bit more than 3:1.

(What we find is that in the first wave, we'll kill maybe one or two longbows (1.5%). The second wave kills 15 or so (15%). The third wave kills 60 or so (75%). The 4th wave clears the mess.

If we have 1:1 siege, about 25% will survive, depending on the promotions we use. From there, we need about a 1.25 : 1 ratio of swords (we get a 75-80 percent survival rate on the first wave, and the second cleans the mess).

For this siege volume, the CRI catapults had about 5% better survival, and that translated to saving two or three of the first wave swords as well.

I don't advise drawing any strong conclusions from these results - this is really just a first pass at experiment design - trying to figure out how to keep track of things in the file, getting a sense for what kind of scoring we might want to use, learning how to build a moat, and so on.

Edit: v2 of the save has smaller groups of siege units, so that you can mix and match strategies more easily.
Edit: v3 of the save turns on the random seed.

 

[VoiceOfUnreason]

After poking a bit further, I decided that I needed to establish some statistical boundaries.

So I ran 20 trials for the following
(a) Slam the city with 100 Barrage I catapults
(b) Slam the city with 100 CRI Swords

Catapult survival rate was 19.55%, with a standard deviation of 3.94%. Swordsman survival rate was 75.3%, with a standard deviation of 3.69%. (note: only 20 trials, so don't take those decimal places too seriously).

Correlation between catapult survival rate and swordsman survival rate was about .03

My interpretation of this: in this kind of sample space, we need at least 5% better survival rates before we should believe that we have an improved sequence of attacks.

If I use the CRI catapults, instead of Barrage I, I record a catapult survival rate of 22.7% with a deviation of 4.62%, and a swordsman survival rate of 76.3% with a deviation of 3.68. Correlation is about .39