Why are simulations rigged against high tier troops?

[Sawdomise]

Every time I try the simulation, I have to save right before since the game decides to kill off my heavily armored mounted troops before my cloth wearing footmen. I have to reload 5-6 times for a logical outcome to happen.

It's insane the number of times I have 120+ troops and decide to kill off a small bandit group (5-20) but yet the only troops to die are Khan Guards or Heavy Lancers, not the 20+ Nomads that have far less HP and armor. At first I thought "maybe it's because they're mounted, they get into melee range faster and take more damage", but no, it also happens with Heavy Horse Archers, so it's really not coming from a logical point of view.

Is there a reason why this design decision was made? Is there plans to change it or will we always have to rely on mods to fix this issue?

 

[AugustusBob]

maybe the simulator uses Athletics for each troop to determine how much time that unit spends actually fighting. faster troops should spend more time in combat than the much slower scrubs. so because they are in combat longer, they just naturally have more opportunity to die, even though they have mad skills and good armor.

 

[Talbrys]

WB seemed to be the same. If you think of troop upgrade system as a cycle (recruit-tier1-2-3-4-5-dead-recruit again...) the higher tier troops are at the end of their life cycle so prioritizing their deaths gives you a reason to keep cycling new recruits. Otherwise you would just end up with 70% or so top tier troops and only a small number of recruit/tier 1 troops that keep dying with no mid tiers. This way you are more likely to have a spread of the various tiers.

At least that's what I always assumed was their reasoning. It could also just be there is no bias and we just remember the high tiers who die better than the newbies, I haven't actually kept records to make a ratio of.

 

[froggyluv]

WB seemed to be the same. If you think of troop upgrade system as a cycle (recruit-tier1-2-3-4-5-dead-recruit again...) the higher tier troops are at the end of their life cycle so prioritizing their deaths gives you a reason to keep cycling new recruits. Otherwise you would just end up with 70% or so top tier troops and only a small number of recruit/tier 1 troops that keep dying with no mid tiers. This way you are more likely to have a spread of the various tiers.

At least that's what I always assumed was their reasoning. It could also just be there is no bias and we just remember the high tiers who die better than the newbies, I haven't actually kept records to make a ratio of.

I dont think im understanding you correctly -youre saying the auto calc will prefer to kill off higher tier so that will give a chance for low tier troops to become mid tier troops -thats insane! Why would a strategy game do that?

i find the whole auto-calc system disappointing as it is. This is where the Arma 3 engine destroys all others as in it actually calculates battles even ones not near the player, all of them, in real time with everything actually taking place even though you are not there to witness it. That way if you have 100 infantry men with no AT weapons encountering a tank -the result will be 100 dead men and 1 live tank. As it should be.

 

[Slithy]

It could also just be there is no bias and we just remember the high tiers who die better than the newbies, I haven't actually kept records to make a ratio of.

It could be this.

I'm somewhat skeptical that Taleworlds intends for higher tier troops to be the only deaths. Autocalc is definitely simplified so you lose some nuance of higher tier troops being harder to kill, but they shouldn't be dying more than everything else.

 

[mylittletantan]

It could also just be there is no bias and we just remember the high tiers who die better than the newbies

Had a look with dnSpy and it seems like simulated battles choose troops completely at random.
SelectRandomSimulationTroop() is called for each round to select units from both sides. This function calls MBRandom.RandomInt(this.NumRemainingSimulationTroops) to determine which troop is selected.

For an explanation on how simulated battles work check out the detailed description in Light Combat Mod Collection.

 

[Apocal]

Is there a reason why this design decision was made? Is there plans to change it or will we always have to rely on mods to fix this issue?

I can't speak to the design intentions, but in case of autocalc (and this was essentially the same in WB), it is two reasons:

1. Because, as mylittletantan pointed out, troops are drawn at complete random.
2. The actual power differential between a top-tier troop and a looter is not insurmountable.

Therefore, if you roll the dice enough times looters will kill a high-tier unit and battles are nothing but a bunch of dice rolls between individual troops. There is nothing deeper going on there. It is just dice luck combined with a (relatively) small gap in performance that doesn't reflect the real gap you'd encounter playing the battle out manually.

edit:
Hold up.

For an explanation on how simulated battles work check out the detailed description in Light Combat Mod Collection.

Detailed description:
"Vanilla auto battles are giving players a hard time. It is a "roll-dice-and-die" game, full of random factors that cannot be controlled. Not only that, but also the player is treated differently than AI: when the player is leading a much stronger party against a weak party, like when hunting bandits, the weaker party receives a HUGE buff so the player party would have much higher casualties than what you'd expect."

OK, wtf? Why would they do this?

 

[froggyluv]

^^Thats literally the opposite of a good idea. Havent the foggiest who thought that was good strategy design. They need Paradox's help on the strategy aspect of this game- it doesnt have to be extreme but at least well thought out.

Ill have to check out that mod above that looks pretty sweet but really shouldn't have to be a mod.

 

[Talbrys]

I'm curious if that's true or they just pulled it out of their butt to promote their mod ?

 

[mylittletantan]

I'm curious if that's true or they just pulled it out of their butt to promote their mod ?

I haven't read all the code related to simulated battles but it looks mostly right, found nothing that contradicts his statements. If he was lying to promote his mod some other modder would probably call him out. Anyone can download a decompiler and look through his mod and or the vanilla dll.

If you have doubts, just use a decompiler and look through MapEvent in TaleWorlds.CampaignSystem.dll to see how vanilla handles sim battle and the mods dlls to see what it does/changes.

 

[Skatan]

I stopped autocalcing mid-game for the risk of loosing a few tier 5-6 troops to looters. It's a bit frustrating but can be circumvented by going into battle and press F6, though of course that takes more time than a regular autocalc battle.

I wouldn't mind a more nuanced autocalc feature that tries to emulate the outcome of an actual battle in a better way, ie if you have a massive archer advantage vs looters, they would rout before they even reached your infantry thus you won't have any losses. But let's see what happens in the future.

 

[markp27]

Why is the OP simulating battles if they have nomads in their army?

Personally I use looters to train the lower tier troops. Put them into their own group and then have them attack the looters, whilst the rest of the army sits back and watches.

 

[HalfMetalJacket]

Why is the OP simulating battles if they have nomads in their army?

Personally I use looters to train the lower tier troops. Put them into their own group and then have them attack the looters, whilst the rest of the army sits back and watches.

There are greater implications on the rest of the game beyond just training. Like exactly what kingdoms are losing after every battle.

 

[Ananda_The_Destroyer]

I wish they would add something else for simulation, like a card game.
I never use it as it is, useless. Was useful when looters couldn't kill. Now they can kill and you recruits somehow get less exp then live battle. ****.

 

[markp27]

There are greater implications on the rest of the game beyond just training. Like exactly what kingdoms are losing after every battle.

Well I would imagine that they are using the exact same simulator as you are. Therefore it is a level playing in that respect as they would be losing their best units first too. Then again it actually gives the advantage to the player if they fight the battles themselves. Basically players lose out if they are lazy and simulate, but gain something if they put in the effort and fight the fight themselves. Let's be honest here most battles against 20 looters takes about 1 min plus load times, so to not fight them is pure laziness upon the player. Therefore they lose any sympathy from me.

 

[HalfMetalJacket]

Well I would imagine that they are using the exact same simulator as you are. Therefore it is a level playing in that respect as they would be losing their best units first too. Then again it actually gives the advantage to the player if they fight the battles themselves. Basically players lose out if they are lazy and simulate, but gain something if they put in the effort and fight the fight themselves. Let's be honest here most battles against 20 looters takes about 1 min plus load times, so to not fight them is pure laziness upon the player. Therefore they lose any sympathy from me.

Its more that an army can just end up losing their best units against an army full of recruits, just because their elites end up dying more in current autocalc. That's not great at all.

 

[markp27]

Its more that an army can just end up losing their best units against an army full of recruits, just because their elites end up dying more in current autocalc. That's not great at all.

All the AI will do is set everyone to charge. Those with higher athletics will arrive first and if outnumbered will be overwhelmed. You also have to take into account injuries from previous battles. You will notice that if you continuously auto battle taking out a load of looters in multiple battles you will lose more troops when you simulate. Better to herd looters into one area and fight one battle rather than a load of battles.

 

[HalfMetalJacket]

All the AI will do is set everyone to charge. Those with higher athletics will arrive first and if outnumbered will be overwhelmed. You also have to take into account injuries from previous battles. You will notice that if you continuously auto battle taking out a load of looters in multiple battles you will lose more troops when you simulate. Better to herd looters into one area and fight one battle rather than a load of battles.

And I think that's some nonsense too. I'd like if AI were a bit more intelligent and used tactics other than charge or wait on their side of the map. Also as far as I'm aware, lost HP is restored on troops for the next battle, it only really matters whether said troop is 'wounded' or not.

Again, I'm more concerned with the effects of AI vs AI, not me with my troops. I don't even autocalc in my battles.

 

[markp27]

And I think that's some nonsense too. I'd like if AI were a bit more intelligent and used tactics other than charge or wait on their side of the map. Also as far as I'm aware, lost HP is restored on troops for the next battle, it only really matters whether said troop is 'wounded' or not.

Again, I'm more concerned with the effects of AI vs AI, not me with my troops. I don't even autocalc in my battles.

I would rather they work on battles full stop. The tactics are nearly non-existent, it does not matter if you are playing manually or simulating,

 

[HalfMetalJacket]

I would rather they work on battles full stop. The tactics are nearly non-existent, it does not matter if you are playing manually or simulating,

That we can agree on, although to give the AI credit, they do know where to really park themselves. Dislodging mob from a hill can be a fun little exercise.

I can only pray that TW can get things going once they sort out their code refractor business.