Reconnaissance Company Test

[Zauberelefant]

So it will only determine who will be able to reroll tactics? And lvl2 general won't be able to execute a tactic with lvl3 or 4 requirement? Welp, that makes recon even MORE worthless.

I don't get that Impression reading the wiki. Most tactics require either high skill or a trait, and most defensive tactics need a skill advantage >0.

So I would say it's the other way around: especially for bad Generals, recon gives a needed edge.

 

[Secret Master]

So, I've finally poked around in the defines and tactics file. My results are mixed.

It seems like FM with Aggressive Assaulter has some impact on tactics selection, but the impact is barely enough to register.

I ran a number of tests with large, non-urban front with around 16 individual battles per day. I tested with a field marshal that had no Aggressive Assaulter and then with the trait. Then I tested with the trait again, but modded the tactics file so that the weight for having the trait went from 2 to 20000. After that, I modded the file again to take away the reduction in chance to fire the tactic in open terrain.

The results were odd.

With no aggressive assaulter and no modding, there were no assault tactics pulled at all throughout four tests.

With aggressive assaulter applied and no modding, there were a few assault tactics pulled. Some tests produced 6.25% assault tactics selection. Other tests produced no assault tactics selected.

With aggressive assaulter applied and a weight of 20000 applied to tactics selection for having the trait, I pulled around 6.25% assault tactics in all tests.

With aggressive assaulter applied and a weight of 20000 applied to tactics selection for having the trait, and the terrain modifiers removed completely, I got wildly different results. In some tests, I had over 30% assault tactics applied. In others, it was closer to 19%. In one, it was only 6.25% of battles.

My initial tests seem like they have too small of a sample size, but I'm open to the possibility that I'm screwing up the RNG seed in the save games in some way, resulting in duplicate results across tests that are not RNG.

One of the things that makes getting large data samples difficult is that the moment you pull assault, the next day you swap to close combat phase. So, that combat on the map can't be used anymore to poll the effectiveness of FM traits. So, I basically have to reload or clear all attacks and restart them every game day.

 

[BloodAsp]

Sad part is people DESPERATELY want for recon to be good. Plenty of youtube guides state that "Recon is important", or "Recon is better now"

 

[bitmode]

One of the things that makes getting large data samples difficult is that the moment you pull assault, the next day you swap to close combat phase.

You could mod the phase transition out of Assault?

My initial tests seem like they have too small of a sample size, but I'm open to the possibility that I'm screwing up the RNG seed in the save games in some way, resulting in duplicate results across tests that are not RNG.

That might indeed poison your test results. Tactics selection uses a random generator which gets re-initialized to the same value on reload. Even the random_seed console command only advances the random generator by one element from there. So if you take the exact same actions you'll like have correlated results. Better just cancel and restart the attacks without reloading.

To double check my finding from recently without WtT I modded Assault to get boosted by Cautious instead of Aggressive Assaulter. I then traced the tactics weights with different cautious/non-cautious leader combinations. Here's the baseline when both the general and the FM are not cautious:

(The first tactic slot is a dummy tactic. The others have the same order as in combat_tactics.txt i.e. Basic Attack is 2nd, Assault is 5th, and Shock is 13th)
The weights are as expected and Assault should happen with ~5% probability. With a Cautious general the value of Assault gets doubled, as expected:

But when only the FM has the trait, the value reverts to 0.400:

So I'm all but certain the tactics part of the FM traits does not work.

The results were odd.

With no aggressive assaulter and no modding, there were no assault tactics pulled at all throughout four tests.

That does sound very implausible.

 

[sekelsenmat]

To double check my finding from recently without WtT I modded Assault to get boosted by Cautious instead of Aggressive Assaulter. I then traced the tactics weights with different cautious/non-cautious leader combinations. Here's the baseline when both the general and the FM are not cautious:
...
But when only the FM has the trait, the value reverts to 0.400:

So I'm all but certain the tactics part of the FM traits does not work.

But cautious is a "Personality trait", I think it could be so that for FMs the tactic selection bonus could only work for "Field Marshal Traits"

So your conclusion would be wrong that it doesn't work at all, but in reality it works but only for "Field Marshal" Traits?

Maybe Personality Traits have zero effects on field marshalls in general (not only on tactic selection).

I have WtT, if you post a guide somewhere on how to do your reverse engineering magic (unless it requires an expensive software), I could do it and post the results here. I'm a software developer, so I could do it with some guidance. My previous experience with reverse engineering was always simply decompiling and visually inspecting the resulting code, never ran something I didn't write on a debugger.

You use gdb or Visual Studio to debug? They included debug info into the released game? Without debug info AFAIK it is impossible to debug C++ software, so I guess they included it.

 

[bitmode]

But cautious is a "Personality trait", I think it could be so that for FMs the tactic selection bonus could only work for "Field Marshal Traits"

So your conclusion would be wrong that it doesn't work at all, but in reality it works but only for "Field Marshal" Traits?

Maybe Personality Traits have zero effects on field marshalls in general (not only on tactic selection).

It's a possibility but the has_trait implementation looked trivial enough to me that I think I would have noticed something like that.

I have WtT, if you post a guide somewhere on how to do your reverse engineering magic (unless it requires an expensive software), I could do it and post the results here. I'm a software developer, so I could do it with some guidance. My previous experience with reverse engineering was always simply decompiling and visually inspecting the resulting code, never ran something I didn't write on a debugger.

Spoiler: boring programming stuff

There is not really much magic to it. If you are on Linux you can put the breakpoint mentioned above directly into gdb and run the game:

$ cd ~/.steam/steam/steamapps/common/Hearts\ of\ Iron\ IV/
$ gdb hoi4
(gdb) break *0x1762b83
(gdb) run

I don't know how to debug just a binary on Windows and the last Windows version of the game I decompiled was 1.7.1. The address there was 0x1405eaeeb. In any case it is the call to GetWeightedRandom(int, CFixedPoint const*, int) in CLandCombatant::SelectCombatTactic(int, int). First argument is the number of selection weights, second is a pointer to the array of 32-bit weights, and third is the die roll/random number to use.

I'm using Hopper as decompiler (Linux/OSX only) but the freeware version of IDA should work as well and is available for all platforms. I don't want to post a more extensive guide on decompiling the game as people might use it for nefarious purposes or it could at least be seen as such by Paradox.

 

[HugsAndSnuggles]

That might indeed poison your test results. Tactics selection uses a random generator which gets re-initialized to the same value on reload. Even the random_seed console command only advances the random generator by one element from there. So if you take the exact same actions you'll like have correlated results. Better just cancel and restart the attacks without reloading.

Having AI on also can take care of seed cycling, but it might interfere with some tests.

 

[Secret Master]

That might indeed poison your test results. Tactics selection uses a random generator which gets re-initialized to the same value on reload.

Yeah, I know. That's why I'd reload, turn off AI, and wait a few days before firing combat. But maybe starting and stopping combat would be better and yield more test results in general.

This is why I hate testing certain things in land combat. Between the RNG seed and tactics and planning bonus and general traits, it's always a pain to get enough data.

I think you're right about it, but I want to be sure before.

 

[ppccctu]

Ok, so if Recon Companies aren't worth the cost, what kind of companies should we using. I've always included engineers and field hospitals for combat units and military police for garrisons. What else should I focus on?

 

[Mousetick]

Ok, so if Recon Companies aren't worth the cost, what kind of companies should we using. I've always included engineers and field hospitals for combat units and military police for garrisons. What else should I focus on?

Please don't take it the wrong way, but this thread focuses on the Recon support company only, and it would be nice if it stayed that way.

There are dozens of threads in this forum about which support companies are comparatively worth using or not, a new one pops up about once a month or so.
The latest is this one, you may want to read it:

Support brigades in Single Player

Hi, I decided to play several games in La Resistance, and boy Italy changed:) My question is about support brigades for lower rank Major, or stronger minor (as Ger/SU/US/UK can afford anything really, and tru minors can't afford them anyway)...

forum.paradoxplaza.com

And this specific post by Reman will answer your question in a nutshell:
https://forum.paradoxplaza.com/forum/threads/support-brigades-in-single-player.1407638/post-26760905

Hope this helps and thank you in advance for keeping this thread on-topic

 

[Secret Master]

I don't bother with HOSP in most cases. It's expensive, and I don't usually need the manpower savings due to how small they are.

Even if tactics selection is borked, RECON is still invaluable for any kind of fast formation (panzers, MOT).

ART, ENG, AA, and AT are all valuable against human opponents. LOG is rarely a poor choice.

 

[ppccctu]

Please don't take it the wrong way, but this thread focuses on the Recon support company only, and it would be nice if it stayed that way.

There are dozens of threads in this forum about which support companies are comparatively worth using or not, a new one pops up about once a month or so.
The latest is this one, you may want to read it:

Support brigades in Single Player

Hi, I decided to play several games in La Resistance, and boy Italy changed:) My question is about support brigades for lower rank Major, or stronger minor (as Ger/SU/US/UK can afford anything really, and tru minors can't afford them anyway)...

forum.paradoxplaza.com

And this specific post by Reman will answer your question in a nutshell:
https://forum.paradoxplaza.com/forum/threads/support-brigades-in-single-player.1407638/post-26760905

Hope this helps and thank you in advance for keeping this thread on-topic

Ok. Thanks. I didn't intend any harm, but I understand your point.

 

[Cavalry]

Speed is probably the only acceptable defense for Recon since it's difficult to quantify its value.

Recon is dirt cheap for infantry, especially when you have open slot because you cannot afford other expensive supports.
Don't forget the small combat stats, especially the defense.

the +10% speed is quite big in difficutlt terrain and you will notice it when you send infantry to reinforce. For example if your infantry is walking as 2km/h due to snow forest, then add a recon will give +10% of 4 km/h, and the actual speed will increase +20% to 2.4. Saving 14 hours from a 3 days walk. 14 hours arriving earlier.

So the recon price and benefits is just right. Well, look at the price again.

Reinforcement and speed always matter in real scenarios. Often the defending side starts weaker but have reinforcement and the attacking side have not.

 

[Anaraxes]

you can put the breakpoint mentioned above directly into gdb

Spoiler: Boring programming stuff

So Paradox left the debug info in the executable? I'd been wondering. It'd be an even more huge effort to reverse-engineer the assembly.

Which decompiler do you use?
[/spoier]