A Quantitative Analysis of Resistance and Compliance

[Reman]

I recently did a deep dive into HoI4's resistance and compliance system. My main results are presented on YT here, but I thought sharing some of my findings on the forum would be appreciated for those who prefer text-based content.

First off, this analysis only covers 6 of the occupation laws (Civilian Oversight, Local Police, Forced Labor, Harsh Quotas, Liberated Workers, and Local Autonomy), because the other 5 are inferior in practically every scenario (the video goes into why they're bad, but for most it's pretty obvious).

Here's how compliance grows over time. Not shown here are Forced Labor and Harsh Quotas, which will typically remain at 0 compliance indefinitely.
The assumptions for most of these charts are as follows:

• The occupier starts with no compliance in the state, so they didn't run a collaboration mission, nor are they occupying areas where compliance had already ticked up (e.g. Africa)
• The occupier is at war, so there's no +10% bonus to compliance gain from being at peace
• The nation being occupied is annexed, so Government In Exile doesn't affect things
• No other modifiers to compliance gain other than for the occupation law being tested

Next, here's how the amount of resources you gain from occupied areas changes over time. There are a few things to note:

• The special laws for Communists and Democracies are much better than Civilian Oversight
• The typical breakeven point between Civilian Oversight and Forced Labor is around 2.5 years
• The jumps you see represent the bonuses from Reorganized Workforce coming online at 40% compliance

Next, here's how factories change over time:

But this chart for factories doesn't paint the whole picture, because what we're actually concerned with is the total amount of equipment produced. So here's a chart that gives a proxy for that. The Y axis is a cumulative sum of the percent of factories available each day, which I use to dodge around issues of things like production modifiers, conscription laws, etc. The result is that Civilian Oversight only produces more equipment than Harsh Quotas after 3 years:

Next, here's how things look when Government in Exile is factored in, which dampens compliance growth. Now, Civilian Oversight takes over 4 years to produce more equipment than Harsh Quotas:

Finally, here are some charts that show the relative amount of casualties taken from different occupation laws. The constraints for the experiment are:

The results for manpower are shown here. These mostly follow the theoretical equations pretty closely:

The results for equipment are shown here. These results do NOT follow the theoretical equations at all, so it might be helpful for further investigations to figure out what's going on (my guess is something finnicky is happening with rounding):

If you want to take a look at specific numbers, I've attached the excel sheet where I made these charts. While the charts were made in Excel, I verified that the results were accurate 2-3 times, so if you find a discrepancy, make sure you're following the same constraints that I've laid out here. If some of the numbers still look wrong, let me know and I'll make adjustments as necessary.

If you have any questions, feel free to ask.

 

[Secret Master]

That's really interesting. A few of my assumptions about factory output were partially wrong it seems.

Question:

Is there any way to compare factory days from Harsh Quota to equipment lost with Harsh Quotas with CAV and Harsh Quotas with AC and Harsh Quotas with plain INF?

Put another way, how do I compare factory days gained from Harsh Quotas with the equipment lost via Harsh Quotas with various set ups?

I'm asking, because Harsh Quotas seems really good for short term (as in, 2 years), and I could easily win the war in that time frame. But if I'm bleeding rifles and ACs, then it's not the bargain it appears to be.

 

[HugsAndSnuggles]

The results for equipment are shown here. These results do NOT follow the theoretical equations at all, so it might be helpful for further investigations to figure out what's going on (my guess is something finnicky is happening with rounding):

Too bad these are suspect. "IC gained vs IC lost" chart would, actually, be interesting.

 

[Reman]

Is there any way to compare factory days from Harsh Quota to equipment lost with Harsh Quotas with CAV and Harsh Quotas with AC and Harsh Quotas with plain INF?

Put another way, how do I compare factory days gained from Harsh Quotas with the equipment lost via Harsh Quotas with various set ups?

It's theoretically possible, although it sounds really difficult. One could take a typical setup in terms of economy law, tech, etc., then multiply it by the number of factories one of the nations had to get an IC value, then compare that against the losses sustained in my practical test of garrison combat.

An easier way would probably be to just try it out in a test game where as many counfounders are held constant as possible, e.g. use factories from occupations to produce equipment that isn't used in combat, then compare that to IC losses from the garrison combat tab.

 

[Secret Master]

It's theoretically possible, although it sounds really difficult.

I see. I was kind of hoping the answer would be "Oh, yeah, here's the data points I took 2 minutes extrapolating from my initial math."

One could take a typical setup in terms of economy law, tech, etc., then multiply it by the number of factories one of the nations had to get an IC value, then compare that against the losses sustained in my practical test of garrison combat.

I could actually do the math for that if I knew how to compute factory-days on your chart into production output as modified by IC techs, export laws, and stability. But math is not my strong suit, and MIC has production efficiency gain.

An easier way would probably be to just try it out in a test game where as many counfounders are held constant as possible, e.g. use factories from occupations to produce equipment that isn't used in combat, then compare that to IC losses from the garrison combat tab.

I've done something similar with other tests, but what muddies the waters here is that the captured CIC from Harsh Quotas can be used to build MIC that produces more equipment, too. Because CIC is assigned automatically, and all possible CIC is assigned in blocks of 15 (that is, you can't tell the CIC screen to just build with the 11 captured CIC, because it will fill in all remaining CIC), I'm not sure I could only count the extra MIC produced by captured CIC.

Actually, I could hypothetically delete all German CIC before starting any wars and just use captured CIC to build factories. That would make it less murky.

 

[bitmode]

The results for equipment are shown here. These results do NOT follow the theoretical equations at all, so it might be helpful for further investigations to figure out what's going on (my guess is something finnicky is happening with rounding):

This is very likely due to truncation. When the garrison takes X% damage, their current equipment pool gets scaled by (100%-X%) and then truncated to whole numbers for each item. So garrison equipment losses are always rounded up, which could be significant for support equipment.
With 25CAV+MP, most states will only have a small fraction of a division, i.e. less than a handful of Support Equipment, leading to disproportionate losses.

 

[bitmode]

The resistance level only influences the frequency of attacks on the garrison, but not the attack value (except Emboldened Resistance at >50% resistance, +100% attack).

The attack value is 1, scaled by resistance_damage_to_garrison and resistance_damage_to_garrison_on_our_occupied_states.

Assuming the latter is zero, the occupation law mostly determines the attack value:

• 0.75 for civilian oversight
• 0.5 for local police force
• 1.5 for harsh quotas

For equipment loss, the attack value is scaled with 2% (NResistance.GARRISON_EQUIPMENT_LOST_BY_ATTACK) leading to

• 1.5% for civilian oversight
• 1% for local police force
• 3% for harsh quotas

of equipment lost per attack. So for anything but infantry equipment, the losses will be dominated by rounding errors. And even infantry equipment gets rounded up quite a bit. Especially in small states just using 1CAV may be preferable unless IC cost absolutely does not matter.

 

[HugsAndSnuggles]

the losses will be dominated by rounding errors.

Will they? Seems that threre is a relatively similar manpower/eqipment loss ratio within the same law, so whatever rounding errors are taking place, they should be minimal. Must be something else going on here.

 

[Secret Master]

So garrison equipment losses are always rounded up, which could be significant for support equipment.

Especially in small states just using 1CAV may be preferable unless IC cost absolutely does not matter.

Hmmm, that seems really odd from a design perspective. A major point of the garrison system is that it should be granular in a way the old one was not.

Would performance be that bad if the rounding was not done at all and instead fractional equipment losses were stored in some way until you lost a whole piece of equipment?

 

[bitmode]

Will they? Seems that threre is a relatively similar manpower/eqipment loss ratio within the same law

The ratios are not similar. Look for example at Civilian Oversight versus Harsh quotas. The manpower cost is about 1-to-2 but the IC cost is almost equal between them.

The ideal formulas predict for each 1.8% of manpower lost 2% of equipment is lost, regardless of occupation law.
25CAV+MP costs 25,500 manpower and 1,560 IC. So e.g. the loss of 44,500 MP in Poland with Civilian Oversight should be matched by 44,500/22,500/1.8%*2%*1,560 = 3,428 IC.
And Harsh Quotas: 99,500/22,500/1.8%*2%*1,560 = 7,665 IC.

Harsh Quotas is least affected by rounding errors because it increases both garrison size and damage to garrisons. If @Reman still has the save game, I'm pretty sure the loss statistics will be disproportionate for support equipment but match the expectations for infantry equipment.

Would performance be that bad if the rounding was not done at all and instead fractional equipment losses were stored in some way until you lost a whole piece of equipment?

The resistance activity events are not performance critical; there are just a few of them per day. That said, removing the truncation would probably slightly improve performance because there are no extra steps to be done. An equipment pool can hold fractions of an item no problem. Alternatively, as is done in many similar cases the amount could be rounded statistically (like the number of attack rolls in land combat).

 

[Riekopo]

Do we know what the historic Garrison deaths were for the areas under German occupation? I really doubt tens of thousands of German soldiers were killed by resistance fighters.

 

[Kryndude]

Amazing work!

 

[Haresus]

Very interesting graphs, great video too, thanks for sharing this here.

 

[Secret Master]

Do we know what the historic Garrison deaths were for the areas under German occupation? I really doubt tens of thousands of German soldiers were killed by resistance fighters.

I won't link the sources due to the presence of swastikas and the figures being mixed in with stats about things we are not allowed to discuss on this forum, but I've seen figures as high as 100,000 for Germans in Yugoslavia during the war, not counting Italians, Bulgarians, or collaborators from inside Yugoslavia. Official German records were much lower, but the "official" ones found were in the personal documents of Hermann Reinecke, but since he's a PR guy, I'm not sure those count as real.

On the Soviet front, up to 10% of German manpower/equipment was devoted to anti-partisan operations during the war. I've seen figures as high as 50,000 KIA just in Belarus. There were points in the war where more German divisions were being used in anti-partisan operations just against the Soviets than the entire DAK and Italian division count in North Africa.

There were also defections among the Axis on the Soviet front, including entire units in some cases. But I'm not sure how much manpower that accounts for.

What I find most interesting is the number of troops that changed sides more than once on the Soviet front, which really messes with the numbers.

 

[Simon_9732495]

Thank you for the great work!

The nation being occupied is annexed, so Government In Exile doesn't affect things

That might be a bigger issue on manpower and equipment losses.
Government in Exile gives up to +20% resistance. (Depending on legitimacy, and vs. AI IIRC you will see the 20%)
(And if you annex them you don't have the modifier "country has capitulated", that gives another 10% resistance.)
So in a real game you have 20% (or 30%) more Resistance in Poland, Romania and Yugo.

20% more resistance means very likely 1 tier of resistance more. (Steps are 25%, 50% and 75%)
It is likely that it makes you go over 25% and you get "Organized Resistance" or over 50% and you get "Emboldened Resistance".

"Organized Resistance" gives Garrison Penetration Chance: +50%
"Emboldened Resistance" gives Damage to Garrisons: +100%

That's a lot more losses...

(And gov. in exile also slows compliance growths in reality)

 

[PanzerMan7]

Liberated workers looks really really good.

 

[Bki]

Liberated workers looks really really good.

Right, I already used it as the USSR for areas I wouldn't hold for long like eastern Poland to maximize what I got from them, but it do seems like it's also better than civ oversight over the long run, so I might have to switch to using it for everything in the future.

 

[Morbus Bubbonicus]

The fact that GER get more resistance from Allies than from Cominter only shows how ignored and neglected situation with USSR and its focuses\features in the latest patch are.
I mean I bet if it was possible to strap -90% combat modifier to Spanish Civil War it is posible to strap +resistance modifier to Soviet lands. At least that would serve as a aband-aid solution until supposed Soviet DLC.
But no