Good division template for Mass Assault

  • We have updated our Community Code of Conduct. Please read through the new rules for the forum that are an integral part of Paradox Interactive’s User Agreement.
Status
Not open for further replies.

Akasi

Sergeant
2 Badges
Dec 15, 2016
97
0
  • Magicka 2
  • Crusader Kings II
I found myself in an awkward position when I went Mass Assault and researched the -0.4 width for infantry division. Sudendly my defensive division went 16 width from 20. I tried to add more infantry but I cant get 20 width again.
For those who enjoy/tried Mass Assault...can you help me with this width issue? How to create a good 20width division after that tech? What to add? Tanks?
 
What are those? Hspart, hspaa....?

Heavy Self-Propelled Artillery/ Anti-Air. You can research them along the Heavy Armor line, and if you can afford them, they are very good when mixed in with infantry (extra protection plus massive soft damage, more than almost any other unit type). But I play Superior Firepower, and I'm not sure how well they work in Mass Assault.
 
If you insist on getting up to 20 again the cheap ways to do so are 1 artillery, 1 aa or 2 anti tank. If we're talking about single player though, 10 infantry battalion divisions being 16 width is not a big deal. You can fit 5 into 80 combat width, and since the non organization stats of such a division are low anyway, the stacking penalty/lack of width used in 60/120 combat width is not that bad. Just make sure you have air support and you can hold forever.
 
Last edited:
People say that "10 wide" infantry are the best for defensive purposes because "you can make more of them . . . and you just cycle them in and out of combat."

That sounds legit in theory but in practice it sounds like micro-management hell. Not to mention the issues of timing. The computer has "perfect timing" and "he" can see and assess everything 60+ times per second and decide when you are pulling back a worn out div/set of divs and sending in a new set, so I'd say that eventually his more acute attention and accuracy are going to catch the human player and cause loss of a hex.

Unless I'm missing something about how hexes under attack cause EVERY unit in that hex to suffer organization loss/attrition?
 
People say that "10 wide" infantry are the best for defensive purposes because "you can make more of them . . . and you just cycle them in and out of combat."

That sounds legit in theory but in practice it sounds like micro-management hell. Not to mention the issues of timing. The computer has "perfect timing" and "he" can see and assess everything 60+ times per second and decide when you are pulling back a worn out div/set of divs and sending in a new set, so I'd say that eventually his more acute attention and accuracy are going to catch the human player and cause loss of a hex.

Unless I'm missing something about how hexes under attack cause EVERY unit in that hex to suffer organization loss/attrition?
It depends how you define "hell." If you don't want to have to touch your units at all, sure, it's annoying, but it's a lot less work than it seems since the computer is not very good at concentrating attacks efficiently. You just keep the battle tab open and click on the individual low org units from that window and move them out, and then add in some new ones from the row of provinces behind, and that's only happening in one or two provinces at a time, even as the Soviets, because of the above.

In single player I don't bother with 10 width anymore, since 10inf+eng and planes is enough to not even need reinforcement.
 
People say that "10 wide" infantry are the best for defensive purposes because "you can make more of them . . . and you just cycle them in and out of combat."

That sounds legit in theory but in practice it sounds like micro-management hell. Not to mention the issues of timing. The computer has "perfect timing" and "he" can see and assess everything 60+ times per second and decide when you are pulling back a worn out div/set of divs and sending in a new set, so I'd say that eventually his more acute attention and accuracy are going to catch the human player and cause loss of a hex.

Unless I'm missing something about how hexes under attack cause EVERY unit in that hex to suffer organization loss/attrition?

I can tell you two things right now.

1) Unless you have a legitimate super computer you are not having an AI compute 60+ assessments of the front while playing a game.

Assumptions:
- assessing the front requires 1 function per tile per unit
- 30 ais are doing important things
- 2000 tiles are contested
- 2000 units exist
- a single ai function requires 200 simple floating point operations

Then we get 2000*2000*200*30*60 = 1.44 trillion floating point operations per second.

My pc is pretty good, and the cpu in it can run about 90 billion floating point calculations per second.

Hoi 4 runs less than 1600% of my cpu. (runs ~30-50%)

Thus we can infirm the absolute max number of ai assessments per second is:

0.50 * 90 billion / 2000 / 2000 / 200 / 30 = ~2.

At 2 assessments per second the game does nothing else but ai calculations.

Source: CS major and game developer

2. Ive actually used this strategy, front ai does half the work, the other half is typical busy work of playing the game. Basically its unnoticeably more time consuming then playing the game normally.

Edit:
All estimate in 1 are gross underestimates.
 
I can tell you two things right now.

1) Unless you have a legitimate super computer you are not having an AI compute 60+ assessments of the front while playing a game.

Assumptions:
- assessing the front requires 1 function per tile per unit
- 30 ais are doing important things
- 2000 tiles are contested
- 2000 units exist
- a single ai function requires 200 simple floating point operations

Then we get 2000*2000*200*30*60 = 1.44 trillion floating point operations per second.

My pc is pretty good, and the cpu in it can run about 90 billion floating point calculations per second.

Hoi 4 runs less than 1600% of my cpu. (runs ~30-50%)

Thus we can infirm the absolute max number of ai assessments per second is:

0.50 * 90 billion / 2000 / 2000 / 200 / 30 = ~2.

At 2 assessments per second the game does nothing else but ai calculations.

Source: CS major and game developer

2. Ive actually used this strategy, front ai does half the work, the other half is typical busy work of playing the game. Basically its unnoticeably more time consuming then playing the game normally.

Edit:
All estimate in 1 are gross underestimates.

Why would the algorithm need to check "1 function per tile per unit?" and where are you coming up with "200 simple floating point operations" per function?

Would one operation per tile be enough: Is hexDefenseVal < myAttackVal ? If yes: attack . . . something along those lines.

Are you gonna try to tell me that the application does NOT cycle at least 60 times per second, because if so, how does it manage to play the slide-show fast enough that it hits the human flicker fusion threshold of ~60 Hz? That is the basis for my assumption on how often the algorithm would check combat odds and the like, but yeah, that might be a bit too often.

1 GHz is capable of 1 billion operations per second is the factoid I remember from computer science 101, so a 3.5 GHz can handle 3.5 billion operations per second no? You have a 90 GHz machine!?

I have never programmed anything nearly this complicated and I admit it is amazing that it works at all, but what other periodicity would you suggest if not along with each cycle of the graphics? One per second? One every other second?

ADDIT: anyway! my point about playing the game remains: the computer "thinks" faster than we do, even if he is only assessing all his units and all their frontage every other second, that is far more than any human can manage to assess every other second.
 
Why would the algorithm need to check "1 function per tile per unit?" and where are you coming up with "200 simple floating point operations" per function?

Would one operation per tile be enough: Is hexDefenseVal < myAttackVal ? If yes: attack . . . something along those lines.

Are you gonna try to tell me that the application does NOT cycle at least 60 times per second, because if so, how does it manage to play the slide-show fast enough that it hits the human flicker fusion threshold of ~60 Hz? That is the basis for my assumption on how often the algorithm would check combat odds and the like, but yeah, that might be a bit too often.

1 GHz is capable of 1 billion operations per second is the factoid I remember from computer science 101, so a 3.5 GHz can handle 3.5 billion operations per second no? You have a 90 GHz machine!?

I have never programmed anything nearly this complicated and I admit it is amazing that it works at all, but what other periodicity would you suggest if not along with each cycle of the graphics? One per second? One every other second?

ADDIT: anyway! my point about playing the game remains: the computer "thinks" faster than we do, even if he is only assessing all his units and all their frontage every other second, that is far more than any human can manage to assess every other second.

The ai doesnt always take the entire tile and attack another tile, so it must look at it on a unit per unit basis.

200 floating point operations per function are an estimate about the complexity of an AI function. Keep in mind for the AI to decide what unit it wants to move to b it has to look at each unit, calculate how valuable it is where it is then pathfind the path to b and compare it to other units. Thats a fair amount of operations, iirc in my game just pathfind from a to b is >200 flops (im not the path finding guy though so could be slightly off). Doing an attack operation is hard to estimate since hoi4s ai cant take alot into consideration or the game would grind to a halt. So Im assuming it just looks at the units equipment & org status vs how valuable it is to attack. In that case I think 200 flops is perhaps an overestimate but on average a fair estimate. However, if im wrong about the ai deciding stuff unit per unit it drastically changes equations.

Also the clockspeed of a computer =/= the number of floating point operations per second it can handle. Basically per clock tick (i.e. per hz) a cpu can conduct an operation per thread and per core*. So at absolute maximum utilization with every thread, core, and every clocktick doing something intel found my cpu can do 90 gflops. I trust that they no what they are talking about there lol.

*Im not a hardware guy so I could be slightly wrong here.

Edit:
Im not sure when/how the ai makes decisions, but id assume it happens per game tick so per hour which tends to be longer than 1 sec.

Edit2:
Also I was unclear last post 2 ai decisions per second was for all 30 ais in the game. So if thats slower and set to per engine tick and perhaps less detailed then im assuming ai functions then it could easily handle the once per tick updates.
 
The ai doesnt always take the entire tile and attack another tile, so it must look at it on a unit per unit basis.

200 floating point operations per function are an estimate about the complexity of an AI function. Keep in mind for the AI to decide what unit it wants to move to b it has to look at each unit, calculate how valuable it is where it is then pathfind the path to b and compare it to other units. Thats a fair amount of operations, iirc in my game just pathfind from a to b is >200 flops (im not the path finding guy though so could be slightly off). Doing an attack operation is hard to estimate since hoi4s ai cant take alot into consideration or the game would grind to a halt. So Im assuming it just looks at the units equipment & org status vs how valuable it is to attack. In that case I think 200 flops is perhaps an overestimate but on average a fair estimate. However, if im wrong about the ai deciding stuff unit per unit it drastically changes equations.

Also the clockspeed of a computer =/= the number of floating point operations per second it can handle. Basically per clock tick (i.e. per hz) a cpu can conduct an operation per thread and per core*. So at absolute maximum utilization with every thread, core, and every clocktick doing something intel found my cpu can do 90 gflops. I trust that they no what they are talking about there lol.

*Im not a hardware guy so I could be slightly wrong here.

Edit:
Im not sure when/how the ai makes decisions, but id assume it happens per game tick so per hour which tends to be longer than 1 sec.

Edit2:
Also I was unclear last post 2 ai decisions per second was for all 30 ais in the game. So if thats slower and set to per engine tick and perhaps less detailed then im assuming ai functions then it could easily handle the once per tick updates.

Well since we're all assuming and estimating here...you can assume not all calculations are made "from scratch" but cached and queued somewhere. Also the decisions are rather easy. It's basically "Value > threshold" while Threshold might be modified (e.g. encircled units will attack if they can).
I don't know about complexity of AI functions because I really skipped that (more or less getting pissed because people talking about AI today a just talking "Big Data" and real AI is far out of reach :) ) but the combat-decisions in HoI4 aren't complex. The result and the changes maybe, but those are only updated once every 24h I think.
 
While interesting to some, computer calculations are slightly off the OP's topic :rolleyes:

@Akasi
Using MA and targeting 20w, go for 12 Infantry. It will be 19.2W. Due to rounding you cant get to 20 once you get the width reduction, but you still get to cram 2 more batallions into the width for free.

After that, eng support is a must for infantry just for the defense buffs you get. That should do for your defensive infantry template using MA.
 
When you look at the overall bonuses of the left tree of Mass Assault (Deep Battle) you see that the bonuses compared to the other trees are lackluster as far as organization, recovery rate, and attack and defense stats. The main benefit is all the supply reduction for a total of -30% which is huge. Therefore, Deep Battle is meant to be an unstoppable late game powerhouse. This is because it allows the player to fit more divisions in a supply zone than the enemy and with the added reinforcement rate bonuses the added divisions have a much higher chance to come in from reserves.

As far as division builds go I do the following. I make infantry 40W by using 25 infantry brigades. For support, I use engineer company, anti-tank (keeping it up to date us a must), logistics company, and signal company (if you have resources). This unit is overkill on defensive stats so the enemy has a hard time breaking through. It also has very high organization which is the most important stat in the game. With this high organization, your units will be able over time push the enemy units for all but the most fortified places. By late game with upgrade logistics company combined with Deep Battle bonuses you get -70% supply consumptions meaning you will always outnumber the enemy by a very large margin.

As for armor for Deep Battle, I also go 40W with ten medium tanks brigades and 10 motorized brigades. For support, I use same as infantry. These units are used to push the front where the enemy line is denser. For motor, I use 40W with 14 motor brigades and 4 rocket artillery brigades. These units are used the same as armor.

I use heavy armor for breaking fortifications. These are 20W and I don’t make that many. They consist of 5 heavy armor brigades, and 5 mechanized brigades, and with the same support as infantry.

With all this added together the way Deep Battle plays is different from other trees. You don’t have good breakthrough bonuses and move speed so encirclements are not the best. You don’t get huge bonuses for sitting around and planning. Deep Battle is designed for the player to push across a front and doesn’t stop so the enemy can’t recover. With the finisher from Deep Battle (Continuous Offensive) and the Offensive Doctrine field marshal trait, you get a massive -35% organization loss when moving. This is hard for any enemy to stop especially the later the game goes on. Be warned though, because the attack stats like soft attack, hard attack, and breakthrough are worse than the other trees you will take a lot of casualties compared to the enemy on a push across a front (especially against superior firepower). In the end, though you will triumph because it is all about who loses organization first regardless of manpower losses.

As for the right tree of Mass Assault (Mass Mobilization) is very situation. It is used if you are resource poor, but what you have is bodies to slow down the enemy which ANY country does with the bonuses for Mass Mobilization and mobilization laws. I find it works best with nothing but 10W infantry consisting of no support. This means the only thing being produced is guns. You can make defensive lines and fall backlines with bodies that will take the enemy a long time to chew through if they even can. Don’t worry when the -0.4 width kicks in from Human Wave Offensive. The new combat width of 8 still divides evenly with the battle combat widths of 80, 120, 160, 200 so you will incur no penalties.

I use Mass Mobilization for Italy which may surprise many people. This is because I build my industry which civilian factories and don’t want to trade for resources for the minuscule amount you start with for a major. I throw bodies at the allies if they attack me in Europe or Africa. Once the axis gets territory and pushes into the Soviet Union I ask for states with resources I need and when the resource problem is gone I switch to another doctrine (I usually build up a navy to compete with the British as top priority).

I hope this helps anyone who wants to play with Mass Assault. It is loads of fun!
 
Last edited:
  • 1Like
Reactions:
Ever tried making a 40-width doomstack? It's actually hilarious!
hoi4_87.png

One downside tho...:rolleyes:
hoi4_88.png
 
The ai doesnt always take the entire tile and attack another tile, so it must look at it on a unit per unit basis.

200 floating point operations per function are an estimate about the complexity of an AI function. Keep in mind for the AI to decide what unit it wants to move to b it has to look at each unit, calculate how valuable it is where it is then pathfind the path to b and compare it to other units. Thats a fair amount of operations, iirc in my game just pathfind from a to b is >200 flops (im not the path finding guy though so could be slightly off). Doing an attack operation is hard to estimate since hoi4s ai cant take alot into consideration or the game would grind to a halt. So Im assuming it just looks at the units equipment & org status vs how valuable it is to attack. In that case I think 200 flops is perhaps an overestimate but on average a fair estimate. However, if im wrong about the ai deciding stuff unit per unit it drastically changes equations.

Also the clockspeed of a computer =/= the number of floating point operations per second it can handle. Basically per clock tick (i.e. per hz) a cpu can conduct an operation per thread and per core*. So at absolute maximum utilization with every thread, core, and every clocktick doing something intel found my cpu can do 90 gflops. I trust that they no what they are talking about there lol.

*Im not a hardware guy so I could be slightly wrong here.

Edit:
Im not sure when/how the ai makes decisions, but id assume it happens per game tick so per hour which tends to be longer than 1 sec.

Edit2:
Also I was unclear last post 2 ai decisions per second was for all 30 ais in the game. So if thats slower and set to per engine tick and perhaps less detailed then im assuming ai functions then it could easily handle the once per tick updates.

Interesting stuff! Thanks for sharing!

I'll leave it at that so we don't keep the off-topic trend going :)
 
When you look at the overall bonuses of the left tree of Mass Assault (Deep Battle) you see that the bonuses compared to the other trees are lackluster as far as organization, recovery rate, and attack and defense stats. The main benefit is all the supply reduction for a total of -30% which is huge. Therefore, Deep Battle is meant to be an unstoppable late game powerhouse. This is because it allows the player to fit more divisions in a supply zone than the enemy and with the added reinforcement rate bonuses the added divisions have a much higher chance to come in from reserves.

As far as division builds go I do the following. I make infantry 40W by using 25 infantry brigades. For support, I use engineer company, anti-tank (keeping it up to date us a must), logistics company, and signal company (if you have resources). This unit is overkill on defensive stats so the enemy has a hard time breaking through. It also has very high organization which is the most important stat in the game. With this high organization, your units will be able over time push the enemy units for all but the most fortified places. By late game with upgrade logistics company combined with Deep Battle bonuses you get -70% supply consumptions meaning you will always outnumber the enemy by a very large margin.

As for armor for Deep Battle, I also go 40W with ten medium tanks brigades and 10 motorized brigades. For support, I use same as infantry. These units are used to push the front where the enemy line is denser. For motor, I use 40W with 14 motor brigades and 4 rocket artillery brigades. These units are used the same as armor.

I use heavy armor for breaking fortifications. These are 20W and I don’t make that many. They consist of 5 heavy armor brigades, and 5 mechanized brigades, and with the same support as infantry.

With all this added together the way Deep Battle plays is different from other trees. You don’t have good breakthrough bonuses and move speed so encirclements are not the best. You don’t get huge bonuses for sitting around and planning. Deep Battle is designed for the player to push across a front and doesn’t stop so the enemy can’t recover. With the finisher from Deep Battle (Continuous Offensive) and the Offensive Doctrine field marshal trait, you get a massive -35% organization loss when moving. This is hard for any enemy to stop especially the later the game goes on. Be warned though, because the attack stats like soft attack, hard attack, and breakthrough are worse than the other trees you will take a lot of casualties compared to the enemy on a push across a front (especially against superior firepower). In the end, though you will triumph because it is all about who loses organization first regardless of manpower losses.

As for the right tree of Mass Assault (Mass Mobilization) is very situation. It is used if you are resource poor, but what you have is bodies to slow down the enemy which ANY country does with the bonuses for Mass Mobilization and mobilization laws. I find it works best with nothing but 10W infantry consisting of no support. This means the only thing being produced is guns. You can make defensive lines and fall backlines with bodies that will take the enemy a long time to chew through if they even can. Don’t worry when the -0.4 width kicks in from Human Wave Offensive. The new combat width of 8 still divides evenly with the battle combat widths of 80, 120, 160, 200 so you will incur no penalties.

I use Mass Mobilization for Italy which may surprise many people. This is because I build my industry which civilian factories and don’t want to trade for resources for the minuscule amount you start with for a major. I throw bodies at the allies if they attack me in Europe or Africa. Once the axis gets territory and pushes into the Soviet Union I ask for states with resources I need and when the resource problem is gone I switch to another doctrine (I usually build up a navy to compete with the British as top priority).

I hope this helps anyone who wants to play with Mass Assault. It is loads of fun!
Does this still work ?
 
If you insist on getting up to 20 again the cheap ways to do so are 1 artillery, 1 aa or 2 anti tank. If we're talking about single player though, 10 infantry battalion divisions being 16 width is not a big deal. You can fit 5 into 80 combat width, and since the non organization stats of such a division are low anyway, the stacking penalty/lack of width used in 60/120 combat width is not that bad. Just make sure you have air support and you can hold forever.
Going over width by 10 is bad enough in penalty that you're much better off adding 2 more infantry battalions and being just under 80.

As far as division builds go I do the following. I make infantry 40W by using 25 infantry brigades. For support, I use engineer company, anti-tank (keeping it up to date us a must), logistics company, and signal company (if you have resources). This unit is overkill on defensive stats so the enemy has a hard time breaking through. It also has very high organization which is the most important stat in the game. With this high organization, your units will be able over time push the enemy units for all but the most fortified places. By late game with upgrade logistics company combined with Deep Battle bonuses you get -70% supply consumptions meaning you will always outnumber the enemy by a very large margin.
One thing I suspect many players don't realize is that SF doesn't just do more damage with pure infantry than mass assault...it's almost strictly better in most ways, with one of the notable exceptions being recovery rate for mas mob.

SF infantry does more damage per IC, more damage per width used (yes, even with the -0.4), and more damage per supply consumption (even with deep battle).
As for the right tree of Mass Assault (Mass Mobilization) is very situation. It is used if you are resource poor, but what you have is bodies to slow down the enemy
This is a misconception about doctrine choices. Mass mobilization uses more equipment than alternative doctrine choices, and is more expensive to use for very poor countries than either SF or GB.

Also, org isn't everything. If a division doesn't do damage, having a ton of org won't do anything but buy a little time, because it won't de-org the opponent's armies either. Org is one of many important stats, but it's unconvincing to say it's the "most important". 300 org and 0 damage will get you strength-deleted divisions. 1000 damage and 0 org won't fight. Units moving at 1 km/h will struggle much to use either stat, etc. There is a place in the game for high-org divisions that take a while to stop fighting. Those are great candidates for pre-nerf CAS, pinning even now, and as damage sponges to help get better divisions in place in time. But it's still a mistake to under OR over value org.
 
  • 1Like
Reactions:
Status
Not open for further replies.