Combat testing results

[odyseusz79]

Here are the results of some test battles I fought; I tried to explore the significance of using guard mode and depth (number of ranks). Hopefully this will be interesting and lead to some good debate on how to best use units. I decided to do this testing because I was amazed at how effective Halberdiers are in this game, even against more "elite" infantry.

Player: Kislev, 8 x Halberdiers (no upgrades or xp)
AI: Khorne, 8 x Khorne Chosen (no upgrades or xp)

Difficulty: Medium

TEST 1, SW, G, 4 deep

Kislev units line up in a single line, spear wall (SW) enabled, guard mode (G) enabled, the ranks are about 4 deep. Battle description: Khorne attacks along the entire line, battles are roughly 1 unit to 1 unit. Result: Kislev victory; losses: Kislev 512, Khorne 997

TEST 2, SW, G, 8 deep

Kislev units line up in a single line, SW enabled, G enabled, the ranks are about 8 deep. Battle description: Khorne attacks only along the narrower frontage presented by the Kislev line. No attempt at flanking. Result: Kislev victory; losses: Kislev 521, Khorne 990

TEST 3, SW, nG, 4 deep

Kislev units line up in a single line, SW enabled, G disabled, the ranks are about 4 deep. Battle description: Khorne attacks along the entire line, battles are roughly 1 unit to 1 unit. Result: Kislev victory; losses: Kislev 515, Khorne 990

TEST 4, SW, nG, 8 deep

Kislev units line up in a single line, SW enabled, G disabled, the ranks are about 8 deep. Battle description: Khorne attacks only along the narrower frontage presented by the Kislev line. No attempt at flanking. Result: Kislev victory; losses: Kislev 371, Khorne 990

So, based on the above, it looks as though the best way to fight here, by some margin, is to have a shorter line with deep ranks, with Guard mode disabled. I guess this is because a longer line of Halberdiers means that more units are actually charged into by the heavy Khorne infantrymen leading to more casualties. A shorter frontage means only a few units absorb the charge.




Finally, a test where a short line, about 10-12 deep, was ordered to march through the enemy line. That is, I selected my entire line and gave it an order to move to a location behind the enemy line that resulted in my troops walking into the enemy as it charged into them.

TEST 5

Khorne again fought only along the frontage presented to it. No attempt at flanking was made, even thought the formation was even narrower than in previous testing. Kislev won again, with losses of 486, so slightly better than standing still and accepting a Khorne charge. Interestingly, combat cancels the marching order. That is, once my units got into a fight, and then finished it, they reverted to "ready" status - they did not continue the march to the position I initially ordered them to.

My next tests will probably involve figuring out the difference b/w using spear wall, on offense (e.g. attacking or marching through) as well as on defense. Tentative testing indicates that rank depth is more important that shield wall/no shield wall, but I'll see.

[The Holy Pilgrim]

Very nice! I look forward to more tests. This will help in my balancing for Real Warhammer.

+rep

[Reiksfart]

This just shows how much the official stats need a vast improvement and the AI is not very good.

I just tried your last test with a 12 deep rank, using Littledwarf's EDU and ReallyBadAI. Three chosen flanked, one of them hit in the rear and I was routed with 830 men lost to 258 chosen dead. Two chosen were held in reserve.

[LittleDwarf]

OMG, there are like 2 people using my EDU! I am so proud! Hope you are having though fights with it. I am not really sure about balance in single player though - it was all meant for multiplayer mostly and only then for custom battles. In multi it tends to work very well i think, i have rather even fights with my friend most of the time, using varying factions.

[Reiksfart]

I am not really sure about balance in single player though - it was all meant for multiplayer mostly and only then for custom battles. In multi it tends to work very well i think, i have rather even fights with my friend most of the time, using varying factions.

There is that, but I have been using my own EDU for a long time which was always similar to yours. I didn't update after the 1.43 patch, not really playing so much while I've been working on my submod, so been using yours.

Coupled with the ReallyBadAI I lose a lot in the campaign on medium/medium difficulty. Last campaign playing as nurgle I lost my first six battles against Tzeentch when usually they would be a walk in the park, outnumbered or not. That's even with the slightly tougher nurgle units from my submod.
I'm not new to total war either, been playing since Shogun 1 which was 11? years ago.

As for the AI expect it to be released for CoW sometime soon. Which reminds me, I must send Germanicu5 what he needs today.

[Pietrak]

Reiksfart - an AI where the AI flanks?! Where can I get it?

[odyseusz79]

Reiksfart,

AI is indeed a problem when you use 8 units per side, as I did. However, when I tried 1 vs. 1 testing, the results were skewed by the fact that both units were general units; the death of a general potentially had too much of an impact on the result - and I wanted to test formations, units, abilities etc.

[naq]

One question here: is it better to have unit charging into the enemy than standing still or marching into them?

[odyseusz79]

naq,

That's exactly the sort of question I want to help answer with these tests.

So far, it looks that if you stand still, and present a wide front to the enemy (i.e. you have a wide and shallow formation), then most of the guys in your first line will absorb a charge from the enemy, resulting in higher losses. So, to reduce the casualties from a charge, reduce the width of your front (roughly speaking).

Marching into the enemy with a narrow front (i.e. a narrow and deep formation) produced fewer casualties than standing still, but this was just one test.

I haven't done a test on charging into the enemy - if you want, do some testing, and report back here.

[Toho]

oh Reiksfart so you are already working with Germanicu5? then i guess i dont need to get myself involved

[naq]

@odyseusz: you should do the test in larger scale. I see the difference is not too much now.

[Toho]

alright i just did a test on my own EDU( still in progress) and here are the results, and there were numerous flanks on both sides if i didnt have 2 knights charging them from behind and the low orc moral i would have most probably lost.

Spoiler Alert, click show to read: 

Spoiler Alert, click show to read: 

Spoiler Alert, click show to read: 

so why is it that this AI flanks a lot(the entire freaking game) and yours don't even acknowledge it?

[naq]

They will try to attack your missile troops if you put them in the flanks.

[Toho]

they were covered by dismounted raiksguard and greatswords. their charge completely decimated the reiksguard but the halberds and greats swords held them off, do you see that long line of killed soldiers under the black orcs? that was suppose to be the defense XD

[Pietrak]

@ Toho - that is not proper flanking. A proper flanking includes a reaction to counter-flanking maneuvers, which the AI neglects to do. When their troops pile up on my flanks (like they did on your tests) I just send my archers/crossbows/handguns around and shoot freely at their rear, they never react to that. In one test battle I managed to pull that stunt off with a cannon - grapeshot at point blank range <3

[odyseusz79]

All right, here's another round of testing. Testing is done on version 1.4.4 with no EDU or AI modifications.

Culprits are the same as before:

Player: Kislev, 8 x Halberdiers
AI: Khorne, 8 x Khorne Chosen

No xp or weapon/shield upgrades.

Difficulty: Medium

TEST 1, nSW, nG, 4 ranks

Kislev units line up in a single line, spear wall (SW) disabled, guard mode (G) disabled, with a depth of about 4 ranks. Battle description: Khorne attacks along the entire line, battles are roughly 1 unit to 1 unit. Results: Kislev victory both times; losses: Kislev 477, Khorne 997; and Kislev 489, Khorne 1009. *** This is strange. Halberdiers with spear war DISABLED did better than Halberdiers with the wall enabled. I tested this twice, got the same result.

TEST 2, nSW, nG, ~11 ranks

Kislev units line up in a single line, SW disabled, G disabled, with a depth of about 11 ranks (each unit is thus roughly square). Battle description: ***Khorne attempted flanking, and even attacking from behind. However, he still lost. Note the low Kislev losses. Result: Kislev victory; losses: Kislev 415, Khorne 990 Yep, I finally got some flanking going, but these square formations responded very well to it. I couldn't believe it, so I re-run the test with 11 ranks and WITH spear wall. This time, for some inexplicable reason, Khorne did NOT flank. That's right. He flanked a non-spear-wall formation, and refused to flank a spear-wall one. My hypothesis as to why this should be is below.

TEST 3, nSW, G, 4 ranks

Kislev units line up in a single line, SW disabled, G enabled, with a depth of 4 ranks. Battle description: Khorne attacks along the entire line, battles are roughly 1 unit to 1 unit. Result: Kislev victory; losses: Kislev 464, Khorne ~990 Yet again, Kislev Halberdiers with spear wall turned off perform BETTER than with spear wall turned on, assuming that Guard is also turned on.

TEST 4, nSW, G, 8 ranks and 11 ranks

Kislev units line up in a single line, SW disabled, G enabled. I ran two tests here, one with 8 ranks and one with 11 ranks. Battle description: In both battles, Khorne flanks successfully and rolls up the Kislevite line. 8-rank result: KHORNE victory; losses: Kislev 748, Khorne 617 11-rank result: KHORNE victory; losses: Kislev 777, Khorne 711 The losses here are sort of random. As Khorne rolled up the flank, Kislevite units were routed. Some recovered and reformed, however, since they'd been running away, they reformed with their backs to the Khorne line, and were easily routed again. I followed my policy of not issuing any orders to Kislev units during battle - they were on their own. So, in this battle, we finally see Khorne victory, AND flanking.

So the crazy part is that Halberdiers perform best with both Spear Wall (SW) and Guard (G) mode disabled. Why this should be, I don't know. I honestly anticipated that units with SW disabled would get chewed up by Khorne Chosen.

It is therefore time to examine other factors that could account for the incredible toughness of the most basic Kislevite infantry (yep, Halberdiers are Kislev's most basic, most easily acquired unit). It could be that their class, which is Heavy Infantry, makes them tougher than if they were Spearmen. It could also be that their weapon attribute, short_pike, is making them tougher. Maybe their unit mass also plays a part.

It is worth noting that regardless of SW or G status, an immobile Halberdier unit that expects an attack from an enemy is described as "bracing" rather than "ready".

Having G mode on causes the front line of troops to kneel down, so in theory the unit (regardless of SW status) should have more halberds in contact with the enemy, but in practice units in G mode perform worse.

Finally, a bit of theory: I suspect that AI's flanking behavior depends on whether the Player's line of troops is already engaged from the front. This makes sense, since flanking a unit that isn't already in combat is tough.

However, this creates a potential paradox. Since it is, theoretically, much tougher to attack a unit in Spear Wall from the front, that means that Khorne units were continually attacking, and then breaking off their attack (because they realized the futility of attacking Spear Wall from the front). So, at no point, was my line fully engaged. Since engaging from the front is a prerequisite to flanking, Khorne never flanked a SW formation, even though it makes great sense to do so. The unintended benefit here was that Khorne units got to execute multiple CHARGES - I suspect that this, rather than some weakness of SW formation, is responsible for higher Kislevite losses when in SW.

With non-SW units, Khorne soldiers stayed engaged longer, so their spare units actually had time to flank and actually scored some victories, but only when Kislev was in G-mode.

The above is just my theory, though. If you have something that supports it or tears it down, by all means share it here.

[The_Gimp]

I think you touched an excellent subject. I only have my experience from numerous battles and didn't run any tests, but it explains some combats results. I mostly play without pausing battles, so sometimes things like halberdiers without SW happens. March them fast without SW and then get engaged with it still turned off (or I forgot to turn in it on) - things like that.

While your tests are very interesting and I think potentially revealing (like so many other things in this game concerning combat behaviour), it also depends how you use this units in a real combat situation.

I for example need guard mode to hold the line so I can shoot / charge from the side / rear without messing everything up and too many friendly casualties. Same goes for gate fights and boiling oil. While the formations may be a disatvantage on their on in such "clinical" situations like your tests, they actually reduce casualties and enable the execution of other tactics.

Of course all this doesn't change the fact that your findings are great

So thanks a lot for this test runs!
Edit, My first rep spread around!

[Pietrak]

Fun Fact, Odyseuszu
I've recreated your set up

Kislev units line up in a single line, SW disabled, G disabled, with a depth of about 11 ranks (each unit is thus roughly square).

Battle description: ***Khorne attempted flanking, and even attacking from behind. However, he still lost. Note the low Kislev losses.

Result: Kislev victory; losses: Kislev 415, Khorne 990

with the result: Khorne victory; losses: Kislev 695, Khorne 350. The AI did flank and even attacked from the rear! It seems that the AI has such a way of acting:
-charge up front
-if front is doing good, flank
-if flanking is doing good, attack rear.

About the outcome: I am using RedDwarfs Dwarven Forge mod - it uses Klierowski's EDU, which is more lore-wise.

[naq]

Do the difficult level affect AI behaviors?

[Pietrak]

Do the difficult level affect AI behaviors?

I think that difficulty affects only morale. It gives the AI +4 morale at Hard and +8 at Very Hard.