A Retinue Composition Efficiency Metric

[redtide]

I've taken a look at some of the tribal retinues. First, a reminder that all of the costs of "feudal" retinues (initial cost, retinue cap usage, monthly upkeep, and monthly reinforcement cost) are based on the "maintenance" (as defined in defines.lua and special_troops.txt) of the units in the retinue: each cost is just the maintenance multiplied by a constant. Tribal retinues have a separate "hire_cost" as well as "maintenance_multiplier" as defined in retinue_subunits.txt. If I had ever gotten around to including nomadic horde retinues in previous versions of this project (or if I had ever actually played nomads), then I might not have been so confused by tribal retinues. It seems that horde retinues have always used the same system. If you're already very familiar with nomadic horde retinues, then this won't come as news to you. But anyway, here are the prestige retinues available to the Norse:

retinue            first   second   prest   cap  monthly
                    unit     unit    cost   usg   upkeep

Tribal Warrior    150 LI              25    105    0.02
Trapper           100 LI    50 A      50    170    0.03
Hunting Party     100 LI    50 LC     75    220    0.04
Veteran Warrior   100 LI    50 HI     75    220    0.04

Each of these retinues has a maintenance multiplier of 0.75, but the retinue cap usage is equal to that of a feudal retinue with the same units. The maintenance multiplier seems to apply to the monthly upkeep, which is 3/4 of what you would expect of a feudal retinue with the same units. From the initial prestige cost, we can deduce the prestige cost of the units in the retinue, to which we can compare the "maintenance" of those same units:

unit    maint    prestige    factor

LI       0.7      0.1667      0.238
A        2        0.6667      0.333
LC       3        1.1667      0.389
HI       3        1.1667      0.389

Unlike feudal retinues, the factor that relates maintenance to the initial cost of tribal retinues is variable. So I think we should just treat the initial prestige cost as arbitrary. I haven't observed tribal retinues reinforcing, but I'm hoping that it will be safe to treat the reinforcement cost as just 3/4 of the reinforcement cost of a feudal retinue with the same units (which is what the monthly upkeep seems to be).

EDIT: If I'm interpreting all of this correctly, it means that (given the way I define my metrics) a tribal retinue will have an offensive efficiency equal to that of an identical feudal retinue but a higher defensive efficiency than that of an identical feudal retinue.

 

[Dragatus]

To me it looks as if the upkeep cost is equal to 0.02% of the retinue cap usage (and rounded) instead of being linked to initial hire cost.

 

[redtide]

For feudal retinues or tribal retinues? Here's what I've been assuming for the units in feudal retinues (per troop):

retinue cap usage = maintenance
initial cost = maintenance x RETINUE_HIRE_COST_MULTIPLIER
monthly upkeep = maintenance x RETINUE_CONSTANT_COST / 1000
reinforcement = maintenance x RETINUE_REINFORCE_COST / 1000

RETINUE_HIRE_COST_MULTIPLIER = 0.14
RETINUE_CONSTANT_COST = 0.25
RETINUE_REINFORCE_COST = 3

Ok, yeah, so that would mean that upkeep is 0.025% of "maintenance". Do you think it should be 0.02%? I did notice that when I looked at Housecarls in game (while looking up the tribal retinue costs), it showed an estimated monthly upkeep of 0.13 instead of the 0.15 that I calculate (and which the wiki shows). If I use 0.02%, it gives me 0.12.

For the tribal retinues, it looks something like this:

retinue cap usage = maintenance
initial cost = hire_cost (arbitrary and unrelated to everything else)
monthly upkeep = maintenance x 0.75 x RETINUE_CONSTANT_COST / 1000
reinforcement = maintenance x 0.75 x RETINUE_REINFORCE_COST / 1000 (???)

 

[Dragatus]

I meant the tribal retinues, though my coment was also partially based on misreading yourpost. Somehow I read the second charge as you trying to estimate monthly prestige costs.

The 0.02% is just based on observing the cap usage and monthly upkeep you listed for the tribal retinues.

105 x 0.02% = 0.021
170 x 0.02% = 0.034
220 x 0.02% = 0.044

The prestige costs are somewhat arbitrary, but not entirely so. The formula seems to be something along the lines of retinue cap usage divided by 3 and rounded to the nearest multiple of 25.

105 / 3 = 35.00
170 / 3 = 56.67
220 / 3 = 73.33

 

[redtide]

The 0.02% is just based on observing the cap usage and monthly upkeep you listed for the tribal retinues.

Right, ok. That makes sense because 0.025% x 0.75 = 0.01875%.

 

[redtide]

I am still working on an update for Holy Fury. I might wait until the first patch to post the files. But in the meantime, I have edited the second post in this thread to contain an updated prospective ranking of retinues based on my "economic" analysis. I will update it as needed.

 

[doktarr]

Well... one option is something like 3 Ethiopians to 2 Skirmish, which gives you 23.5% archers and gets you 63% feint

That's not what I calculate for the likelihood of Feint. With a good commander (Martial >= 16), the Heavy Infantry in the Skirmish retinue will hit two Shieldwall multipliers, resulting in 40.2% Feint and 40.2% Shieldwall. I assume Martial = 18 for all of the calculations in my rankings.

To get 63% Feint with a good commander, you need something like Ethiopian 3:7 LS

Sorry, I was suggesting light skirmish as you suggest - Heavy Infantry obviously makes no sense. No excuse on the math, but you correctly sussed out the ratio I was suggesting.

which--compared to Ethiopian: Def: Shock (22:1:1)--does have higher skirmish "offensive intensity" (113.5% vs 77.0%) and "offensive efficiency" (123.4% vs 100.1%) but lower skirmish "defensive strength" (106.4% vs 193.5%) and "defensive efficiency" (116% vs 310.1%) and thus a lower combined metric (1.193 vs 1.631). I suspect that the trade-off is not worth it, but it's the sort of thing that should be tested by battles. I'd like to see what the results of that are.

In practice, offense generally trumps defense for a skirmish-focussed army, as you've basically got until skirmish ends to rout your flank or you're getting crushed in melee. That said, the defensive gap there is pretty big.

At any rate, the ratio there is not terribly different than Pictish Raiders so you could really just see how Pictish Raiders do in combat compared to the 22:1:1.

Funny aside: If you have 12-15 Martial it's actually 68% feint instead of 63% because there's no additional multipliers for feint after 12 but volley gets one at 16. This is one of those cases where higher martial leads to poorer tactics. The most ridiculous case is where using a Scottish commander with 7 Martial can allow all-pike (or pike+archer) flanks to throw Schiltron nearly half the time, which is about the best result you can get. It's more often than the same flank would throw Schiltron with a M16 commander, as well as more often than a carefully constructed 69pike/31Heavy flank with a good commander throws Schiltron.

 

[redtide]

I've actually done some more morale testing and I think I may have found a formula for morale damage, but it only applied when I had an equal number of soldiers divided into an equal amount of equally sized units. The formula seems to be:

morale damage = soldiers killed * moraleloss_factor * (death_morale_damage + 1)

I have some questions about this. What exactly do you mean by "morale damage"? I am most interested in what the battle UI calls "Average morale in this flank" (which at 20% causes the flank to flee). I would like to know how to calculate that value and/or its daily decrease. There is also the incoming "morale damage" dealt by the enemy flank (which seems unconnected to death_morale_damage).

Here are a few thoughts that I have based on some very limited testing (two agreements and one addition):

moraleloss_factor does indeed seem to multiply the entire expression: setting it to 0 results in no decrease in "Average morale in this flank" regardless of whether death_morale_damage is 1 or 0.

death_morale_damage does not multiply the entire expression: setting it to 0 doesn't prevent "Average morale in this flank" from decreasing (when moraleloss_factor is nonzero).

The "morale damage" dealt by the enemy flank has an effect on my flank's "Average morale in this flank" that is not tied to death_morale_damage or soldiers killed: I set death_morale_damage to 0, and the "Average morale in this flank" still decreased; when I gave the enemy commander the Cruel trait (which gives +10% morale damage), the "Average morale in this flank" decreased more quickly even though troop casualties remained the same. Keeping death_morale_damage at 0 and increasing troop defense to 1000 (such that there would be no deaths) resulted in the same decreases in the value of "Average morale in this flank".

 

[Dragatus]

Morale damage in the formula I posted is how much the total morale of a flank is reduced by the damage and asociated kills inflicted by the enemy beyond just losing the morale that the killed soldiers were contributing to the flank.

For example if you have a pure Heavy Infantry flank (morale 4) and on a given day of battle 100 od them are killed you'd lose 400 morale for no longer having the soldiers in your flank and another 100 x 3 x (2.5 + 1) = 1050 morale from taking enough damage to kill 100 of your men.

Seting death_morale_damage to 0 does not nullify that becuase of the offset value (the +1 part of the formula). what it does is reduce the formula to morale damage = moralelosss_factor * kills * offset value.

Based on what you said it seems that the commander's morale damage bonus multiplies the entire expression.

What you mentioned about setting defence to 1000 and it having no effect on morale suggests that perhaps it's based on raw damage output rather than kills.

On a side note, the 3.0.1 patch notes mention the addition of a new Arberian cultural building and retinue. Outremer culture is also supposed to get access to Knight retinue and Couched Lance Charge.

 

[dD_ShockTrooper]

Hey, can I get for a comparison between pure light cav culture retinues and their superior mixed order compositions you've got? I know they're going to be worse, but I'm interested to know how much throwing disorganised harass actually hurts these retinues. (Since in the case of camels, it surprisingly wasn't too bad a tradeoff)

 

[redtide]

Hey, can I get for a comparison between pure light cav culture retinues and their superior mixed order compositions you've got? I know they're going to be worse, but I'm interested to know how much throwing disorganised harass actually hurts these retinues. (Since in the case of camels, it surprisingly wasn't too bad a tradeoff)

I'll quote just the offense and defense values for Andalusian Cavalry. (I'm hoping to develop a more accurate characterization of morale than what is currently in the files.) Skirmish percentages are relative to the Light Skirmish retinue, and Melee percentages are relative to the Defense retinue.

skirmish
composition           Δ_trp    A_trp    D_eff    O_eff
Andalusian Cavalry    244.9%   174.4%   101.8%    61.3%
Andalusian 4:1 LS     173.7%   206.2%   107.0%    86.9%

melee
composition           Δ_trp    A_trp    D_eff    O_eff
Andalusian Cavalry    121.4%    88.9%    92.9%    71.6%
Andalusian 4:1 LS      67.1%    71.5%    85.1%    69.0%

So, it's pretty similar to Camels, in that the mixed composition is only slightly better in skirmish, and the LC by themselves are better in melee. I should probably include the pure LC flanks in the rankings. As others have noted, the mixed composition is risky because 11.5% of the time it's initial skirmish tactic will be Volley, which penalizes Light Cavalry offense by -150%.

 

[Dragatus]

I've been thinking and if morale damage is based on enemy damage output rather than kills that would both explain a reported battle outcome that i had previously thought impossible and completely change the ranking of the various combat bonuses.

The battle in question is from a year or two ago. Someone tested Italian Pike vs Scottish Schiltron in a 1 vs 1 battle and reported that the Pike retinue won while using Force Back tactic and the Schiltron retinue lost despite triggering the Schiltron tactic. According to the old thinking that should be impossible becuase the Scots should have a ridiculous defense and higher defense meant fewer deaths and that in turn should lead to less morale lost too. The person indeed also reported that despite losing the battle the Scots lost far fewer men. I meant to run a test of my own to confirm, but never quite got around to it.

But if morale loss is based on raw enemy damage rather than kills that changes everything. It means that defense actually has a rather minor effect on the outcome of a battle compared to morale and especially attack. it still helps and a unit with a defense bonus will beat an equivalent unit without one because of how the remaining morale percentage is the percentage of the full morale of surviving soldiers and not the full morale of all the soldiers in the flank at the beginning of the battle.

To clarify that with an example, if you have a flank of 1000 soldiers with 4 morale each then at the beginning of the battle 4000 morale is considered to be 100%. But if 200 of them die and 800 survive then 3200 morale will be considered to be 100% morale. And if 400 of them die and 600 survive then 2400 morale will be considered 100%.

Let's say we have two otherwise identical retinues of 1000 of these soldiers but one gets a 100% defense bonus. And let's say that they take enough damage to lose 200 and 400 soldiers respectively and take 800 morale damage on top of that. The one without the defense bonus will be at 1600 out of 2400 morale (67%) while the one with the defense bonus will be at 2400 out of 3200 morale (75%).

So that's how defense still helps in battles, but now let's say that the unit without the defense bonus instead gets a 100% morale bonus. That would put it at 4000 out of 4800 morale (83%). And an offense bonus would be even more powerful. The retinue would be at 1600 out of 2400 morale (67%), but the defensive one's loses would've been doulbled and so would the morale damage and it would be at 800 out of 2400 morale (33%).

Essentially attack would be the most valuable attribute becuase it both inflicts kills and reduced the morale of the survivors. It is as valuable as defense and morale combined becuase it counters both their functions. And out of morale and defense morale would be the stronger one as far as winning a given battle is concerned, though defense would still be useful for prolonged campains due to reduced erosion of troop numbers.

Of course I'll still need to run test battles to confirm that's actually the case, but it seems to be so based on what you reported and it has several exicitng implications.

 

[redtide]

if morale damage is based on enemy damage output rather than kills

I suspect that the decrease in the "Average morale in this flank" is based on both the "morale damage" inflicted by the enemy commander and the troop losses in a given subunit (via the death_morale_damage multiplier). I have been isolating and trying to deduce the contribution of the "morale damage" inflicted by the enemy commander by setting death_morale_damage to 0. I plan to do a similar thing for death_morale_damage by trying to set the "morale damage" inflicted by the enemy commander to 0 by modding the Cruel trait.

out of morale and defense morale would be the stronger one as far as winning a given battle is concerned, though defense would still be useful for prolonged campains due to reduced erosion of troop numbers

Yeah, I think this is exactly the right way to think about it.
higher morale + lower defense = more likely to remain in battle and suffer more losses
lower morale + higher defense = more likely to flee battle and suffer fewer losses

attack would be the most valuable attribute becuase it both inflicts kills and reduced the morale of the survivors

And I think that anything that increases a commander's morale damage (like the Cruel trait) should be considered very good.

 

[Eicna]

God. This stuff is too much for my small brain.

You mentioned that mixed in LI will yield better tactics(for example Andalusian 4:1 LS). So how do I compose them? do I put one LI with 4 Andalusian in one flank and leave the center and the other flank empty? or just spread them equally on all side(center, left flank, right flank)?

 

[jwalche]

For those who want a easy and all around composition, LS+Archers for Faint Tactic makes life much easier;

The practice is simple. After starting as or adopting feudalism, fill your entire retinue with only Light Skirmish, which has 400 LS and 150 Archers. Each man costs less than half of other types so it is by far the best for both having large military presence and siege.

It is weaker than other types, but you have twice more soldiers and can usually defeat enemy with just 1.5x more soldiers by just making sure that you have at least 20% archers on each flank, and nothing but LS and Archers to focus on firing only and more of Feint Tactic.

Feint Tactic 12 Defensive

• Skirmish
• Light Infantry 20%
• Archers 20%

• 3 – Martial 12, Archers 30% of all skirmish units
• 3 – Martial 12, Light Infantry 40% of the entire flank

• Light Infantry Offensive +100%
• Archers Offensive +150%
• Light Infantry Defensive +50%

For rare major battles, you can enhance its efficiency by hiring a strong defensive mercenaries for your two flanks with defensive commander, and put all your LA / Archers on one flank to kill off enemy even before they come to melee range.

I learned it from here;
https://www.reddit.com/r/CrusaderKings/comments/4t8zb4/guide_how_to_fight_with_light_infantry/

 

[redtide]

You mentioned that mixed in LI will yield better tactics(for example Andalusian 4:1 LS). So how do I compose them? do I put one LI with 4 Andalusian in one flank and leave the center and the other flank empty? or just spread them equally on all side(center, left flank, right flank)?

Each flank fires its own tactics. The examples that I have listed are flank compositions. So you would need to put the 4 Andalusian and 1 Light Skirmish in the same flank. You could have any number of them that reduces to a ratio of 4:1 (i.e 8:2 or 12:3, but 11:3 would also work). Don't leave the other flanks empty; put some other troops in them. They could be levies, mercenaries, a holy order, another Andalusian 4:1 LS retinue flank, or some other mix of retinues. Unfortunately, this means you'll have to do a fair amount of army micromanaging, especially when you merge armies because the game likes to mess up your finely tuned flanks.

 

[suokko]

Is this realy for 2.8+? I'm surprised that pikes still look that strong after nerf. I didn't analyse your math carefully but quick peek looked like sound idea. I'm considering that math seems to not to take into account that defending pikes have huge advantage during transition period because they get often switch same day to melee tactic with huge affinity bonus against potentially bad melee charge tactic for the attacker (unless that has been fixed but I don't think it has been).

Maybe the issue is that just looking at offense and defense numbers and trying to combine them doesn't really tell how battle goes. It depends a lot what composition enemy army has. Basic idea would be that skirmish oriented retinues should nearly only look to their attack values because they are often just trying to win the battle before switch happens. Difficulty level has fairly big impact on skirmish optimised units because AI gets morale boost. But even at very hard difficulty LS is effective retinue type against most AI armies.
Melee optimised setups often should try to minimize time spent in skirmish and then hope for favorable affinity bonus because melee affinity bonus will have much larger impact than skirmish affinity.

I was just about to check new tactic changes but I figured out a very strong composition for camels. It should be fairly easy to keep primary arabic culture but maintain a few potential tutors with ethiopian culture. Then you could compose something like 4 four camels with a ES. That has a small tolerance compared to optimised 74% camels target to allow harass fire even after some loses because light units have higher loses than camels. I have never tries it but it should be scary strong flank for skirmish attack. Probably easiest to archive if starting as Ethiopian character and later switching to arabic when you have enough retinue cap and income to maintain large standing armies. Early game with ES only should work well because they combine well with levies (like noted in this thread already).

Basically I like LS+CW flanks if I have high enough retinue cap to have two or three flanks. I don't like that composition for a retinue flank because unlikely volley triggering will mean you lose the battle. My idea is that battle is automatic win if any LS+CW flank chooses harass with flanking then causing a lot free damage soon. Two flanks both throwing volley is only about 1% chance while two avoiding harass completely is only about 20%.

ES would improve skirmish tactics a lot because it would eliminate bad tactics volley, shieldwall and volley harass. (Volley harass is pretty good tactic but horrible compared to harass) If you avoid inspired defense from commander then weights would be 3 for generic and 16 for harass. About 84% harass would be scary powerful without any horrible alternatives weakening it. Melee should still be raid only which would leave transition tactics only weakness for CW+ES flank.

Another potential missing powerful combination is pure nubian archers with english or welsh commander. MLVT has 3.5 modified attack per maintenance (better than CW harass) that is enough for a likely skirmish phase victory against any similar cost flank. I haven't ever even tested british archers in game to see how badly their low morale affects the battle. But I would assume it would be pretty similar to pre 2.8 horse archer where battle must be completely won in skirmish or you will lose it.

I noticed there was some talk about how morale lose is calculated. I haven't tried to figure it out completely but basic idea seems to be attack causing morale damage with formula unit numbers*attack value*0.01*commander modifiers and then death morale lose is added to that. Death morale lose seem to be in most cases much higher than direct morale damage shown in tooltips which makes unit loses often dictate how quickly morale is lost. For retinue optimisation understanding the morale lose from deaths correctly would be important. But for practical purposes you could pick any at least descent retinue and use them with composition which gives descent battle tactics.
I have once seen AI army with insanely high morale defensive modifier for a flank which had only withdraw when I managed to kill about 60-80% troops from the flank. The battle was a while ago (maybe 2.6) but back then I was surprised losing a "clear" winning battle. Then I analysed my second battle with my full army against same army I noticed single flank surviving insanely long and nearly winning against my massively larger army.

Personal preference vs AI: But in empire building game I prefer pure LS retinues unless cultural retinues offer LI+A units. LS is very cheap to build and quickly grows to about 10k-20k armies. which can be then distributed all over the empire. This allows quick opportunistic wars at any border while distributed armies provide defense against defensive pacts. AI is so bad at playing the game that you don't need more powerful troops. Then I have a few specialist assault armies either heavy retinues or holy orders or mercenaries (vassalized obviously). I only raise levies in begin and to later to defend my core lands (eg. huge faction).

 

[redtide]

Is this realy for 2.8+? I'm surprised that pikes still look that strong after nerf.

Yeah, the files that are posted right now are based on the post-2.8 unit stats. They don't have any of the retinues or tactics that were added with v 3.0, but I don't think any of the unit stats have been changed. The prospective rankings in the second post are based on an updated spreadsheet that I haven't posted yet.

The v 2.8 changes didn't hit Pikemen too hard (especially compared to past changes to unit stats):
"Maintenance" increased from 2 to 2.5
Skirmish defense increased from 3.5 to 4
Melee attack decreased from 5 to 4.5

That's a 28% decrease in Pikemen's melee offensive usage-effectiveness. After 2.8, Pikemen have 1.8 melee offense per retinue cap usage compared to Heavy Infantry's 2, but the Pikemen offense and defense bonuses imparted by the Defense retinue are greater than the Heavy Infantry offense and defense bonuses imparted by the Shock retinue, and the Defense retinue has a lower percentage of Archers. The Scottish and Italian tactics remain very powerful.

I'm considering that math seems to not to take into account that defending pikes have huge advantage during transition period because they get often switch same day to melee tactic with huge affinity bonus against potentially bad melee charge tactic for the attacker (unless that has been fixed but I don't think it has been).

Yeah, my approach treats the skirmish and melee phases in isolation, which is unrealistic, so it neglects any peculiarities of the transition from skirmish to melee.

Maybe the issue is that just looking at offense and defense numbers and trying to combine them doesn't really tell how battle goes. It depends a lot what composition enemy army has. Basic idea would be that skirmish oriented retinues should nearly only look to their attack values because they are often just trying to win the battle before switch happens.

ARCEM doesn't necessarily predict which compositions are more likely to win their battles, and since it considers a flank in isolation (treating incoming damage as an unknown quantity) it neglects the composition of the opposing flank and any affinity bonuses that it might get. It also considers offense and defense to be of equal importance regardless of whether the composition is skirmish-oriented, melee-oriented, or balanced. I would argue that defense is still important for skirmish-oriented flanks because a troop that survives can continue to deal damage. But for a melee-oriented flank, it probably makes more sense to emphasize skirmish defense than a combination of skirmish offense and defense.

Melee optimised setups often should try to minimize time spent in skirmish and then hope for favorable affinity bonus because melee affinity bonus will have much larger impact than skirmish affinity.

For a melee-oriented flank, it probably makes sense to aim for a balance between high skirmish defense and a high probability of charging into melee, but I haven't accounted for the latter.

I figured out a very strong composition for camels. It should be fairly easy to keep primary arabic culture but maintain a few potential tutors with ethiopian culture. Then you could compose something like 4 four camels with a ES.

I haven't included any "exotic" combinations of cultural retinues in the rankings because I thought they seem a little too difficult to assemble and maintain and because I didn't build a way to toggle between cultural buildings into the spreadsheet. I have added it to the updated spreadsheet for the purpose of correcting an error (the omission of cultural building bonus to the units in the generic retinues), and it wouldn't be too hard to apply it to the cultural building bonuses of the cultural retinues (right now they are all set to receive cultural building bonuses, so if you use the existing spreadsheet to devise an "exotic" combination it won't have realistic values).

But yeah, a combination of Camel Warrior and Ethiopian Skirmisher retinues would be very powerful. I made a little mod that changes the compositions of all of the cultural Light/Camel Cavalry retinues to 70% LC/CC and 30% LI so that they can get Harass instead of Disorganized Harass without also getting Volley.

I noticed there was some talk about how morale lose is calculated. I haven't tried to figure it out completely but basic idea seems to be attack causing morale damage with formula unit numbers*attack value*0.01*commander modifiers and then death morale lose is added to that. Death morale lose seem to be in most cases much higher than direct morale damage shown in tooltips which makes unit loses often dictate how quickly morale is lost. For retinue optimisation understanding the morale lose from deaths correctly would be important.

I'm hoping to come up with a more accurate account of morale. I haven't been able to do any tests recently, but it seems to depend on the incoming morale damage, troop deaths, DEATH_MORALE_DAMAGE, and MORALELOSS_FACTOR.

 

[suokko]

ARCEM doesn't necessarily predict which compositions are more likely to win their battles, and since it considers a flank in isolation (treating incoming damage as an unknown quantity) it neglects the composition of the opposing flank and any affinity bonuses that it might get. It also considers offense and defense to be of equal importance regardless of whether the composition is skirmish-oriented, melee-oriented, or balanced. I would argue that defense is still important for skirmish-oriented flanks because a troop that survives can continue to deal damage. But for a melee-oriented flank, it probably makes more sense to emphasize skirmish defense than a combination of skirmish offense and defense.

True. It is hard to predict exact effectiveness. But for skirmish oriented troops defense is much less important than balanced and melee oriented. (At least against AI) The difference is that offence works as defense when battle is over in a very short time (preferable before or soon after first tactic reroll). Basic target is to have skirmish oriented troops generating about four percent point morale lose per day which often means opponent is only having fraction of your offence.

For a melee-oriented flank, it probably makes sense to aim for a balance between high skirmish defense and a high probability of charging into melee, but I haven't accounted for the latter.

True. But if opponent has also melee oriented flank against you then you would want as attacker just to buff your shield wall high and wait defender to trigger transition. That is much safer than charging because melee transition tactics are often bad for your flank.

But yeah, a combination of Camel Warrior and Ethiopian Skirmisher retinues would be very powerful. I made a little mod that changes the compositions of all of the cultural Light/Camel Cavalry retinues to 70% LC/CC and 30% LI so that they can get Harass instead of Disorganized Harass without also getting Volley.

Those retinues sound like you can just setup all three flanks with retinues and watch them blitz enemies with harass.

I'm hoping to come up with a more accurate account of morale. I haven't been able to do any tests recently, but it seems to depend on the incoming morale damage, troop deaths, DEATH_MORALE_DAMAGE, and MORALELOSS_FACTOR.

Actually I was wondering same because numbers that game show and how fast morale goes down didn't really make sense to me. Potential base formulation that I game up just now was actually pretty simple morale damage from flank composition view times MORALESLOSS_FACTOR plus deaths times DEATH_MORALE_DAMAGE. Then I just apply commander morale lose defense multiplier and I get pretty close predictions to what game shows as morale lose percent point change. I didn't actually try to modify those values to validate if my formula keeps giving correct predictions with different values.

Same time I also noticed that subunit commanders seem to be additive to flank commander when calculating damage, defense and morale damage. Basically if you have more than three good commanders then subunit commanders on most important levies can have minor improvement to your chances of winning. Specially in early game when number of subunits are still low. But retinues are split to so many subunits that only flank commanders have any impact to results.

 

[Beyond Disbelief]

(everything)

*sees hardcore statistical math*

Sir, are you sure you're not formerly a Korean pro StarCraft player?

Joking aside, definitely bookmarking this (or whatever equivalent we got for these forums) for future studies.​