Make One Division Training Great Again! (With the Algorithm)

[Darknessunder]

So I started my experiments on division training XP.

I tried to figure out that under the new training XP system, will two 10 battalion division(20 widths for all infantry) give more XP or a 20 battalion(later referred to as 40 widths) division give more XP.

From my observations, The clear and definite winner is the 40 width division.

Before I dive into my researches, I have to point out to you that under certain situations, two 20 widths will have the same XP gain as a 40 width, but that can be explained by the algorithm.

So my research started off in 1936 start U.S. regular setting. I first consolidate the starting divisions into 10 fully equipped 20 width infantry. (For these you do not know, you receive XP at a reduced rate for training divisions without equipment) The template was created through command console XP.

After they all gathered up in D.C., I checked the XP I would be gaining if I would be training them and then I switched these divisions to 40 widths and merged these divisions and looked that the XP I would be getting from training. That transition give me 5 fully equipped divisions as a result. The XP I would be gaining was the same for both scenarios.

Then I kept training divisions to see if I was getting a glitch. An interesting thing happened: the training XP keeps rising. So I started to watch it. Eventually, the XP gain level stopped at the point which you would normally get if you are doing One-Division-Training for a 40 width division.

Then I started to delete one of the five divisions. After initially dropping down the XP level to 4/5 of the maximum, it quickly backed up to the maximum level very quickly this time. I kept deleting one division at a time and the XP level recovered faster as there are fewer divisions.

So I realized something about the new algorithm that is about previous division count. My next target was France and it failed miserably. It took all the way to late 1937 for France to back to one division-training level. Thus, I took South Africa, which only starts with 6 divisions, to be my next experiment target. With South Africa, I learned that the XP level grows even if I am not training these divisions.

Thus I went back to where I started - the U.S.- to find the mathematical algorithm behind the XP level. I had Desmo and Excel open to plot the graph as I expected the algorithm to be logarithmic.

And then the final experiment begins, with the help of EasyBuff, I quickly changed all of my division to be level 3 30 battalion-sized(25+5) division to be notified about the smallest change in XP level.

I started to delete one division at the time and recorded the time it back to the maximum level. After 5 recordings, I realized that each recorded time has about a 10-day interval. To test out my observation, I started to delete 3 divisions and 6 divisions at the time( thus about one or two month). The result was very close. I kept testing until there's only 1 division left and the result was very assuring: it takes 10 days for the xp level to reach back to the maximum for each "missing" division.

Hence, I present it to you to the final algorithm (If a term is in all caps, it's quoted from the game documents) :

UNIT_EXPERIENCE_PER_TRAINING_DAY(0.0015) * TRAINING_EXPERIENCE_SCALE(62.0) % * equipment ratio * # of training division / record of division # in the army. If record of division # in the army is greater than the current # of all divisions in the army, then it will be reduced by 1/10 each day.



So, for a country like France who has 78 divisions in 36, One Division Traning is a No-no. For country like U.S. or U.K., it's still possible and you can still do it from late 1936(which would be about 3 years before historical WWII). And for minor nations, One division training is still a good option if you are not rushing wars(which is usually not the case for SP minors but is true for historical games).

----------------------------------------------------------------------------------------------------------------------------------------------------------------

So the point of this is informing you about the new training XP mechanics and not necessary to suggest you to do 1-division training.
My simple suggestion is to reduce your army size until they are fully-equipped and start your training from there.
If you are a minor nation need that extra edge of saving a few hundred guns by one division training, do so as you please.
If you think that 1-division training reduces your ability to deal with situations, not expanding your army while training is my advice to you.

Again, this is not advocating for one division training but rather indicating who to maximize your XP gain from training. If Paradox decides to actually give soldiers guns in the beginning of 1936, I advise you to do full army training. I mean, we all know it's ridiculous it is for most of the starting units to be 1/3 equipped.(and we know it's so that Germany actually stand a chance against Allies)

 

[ecpgieicg]

Nice experiments and nice discoveries!

 

[Zeprion]

Congratulations on the experiment! I thought the one division training was gone for good.

But I would have still preferred that the one division training was gone for good. It's not realistic at all, seems more like an exploit than a wanted feature.

 

[Darknessunder]

Congratulations on the experiment! I thought the one division training was gone for good.

But I would have still preferred that the one division training was gone for good. It's not realistic at all, seems more like an exploit than a wanted feature.

I would agree this is an exploit if most starting divisions are fully equipped. But Since Paradox made it impossible to train your entire army at the start, I would say one division training is fair.

 

[FindFloppies]

Congratulations on the experiment! I thought the one division training was gone for good.

But I would have still preferred that the one division training was gone for good. It's not realistic at all, seems more like an exploit than a wanted feature.

Agreed. Just one of those very game-y solutions that is completely without logic, and breaks immersion completely.
I have no plans to use one-division training, and I didn't do it before. Having said that, what you do in your single-player game is your affair, but make sure that's ok with other people you play MP with.

 

[bitmode]

Hence, I present it to you to the final algorithm (If a term is in all caps, it's quoted from the game documents) :

UNIT_EXPERIENCE_PER_TRAINING_DAY(0.0015) * TRAINING_EXPERIENCE_SCALE(62.0) % * equipment ratio * # of training division / record of division # in the army. If record of division # in the army is greater than the current # of all divisions in the army, then it will be reduced by 1/10 each day.

At least when I checked in March, the formula looked slightly different: https://forum.paradoxplaza.com/foru...n-from-exercise-on-1-9.1355821/#post-26359717

 

[bitmode]

So, for a country like France who has 78 divisions in 36, One Division Traning is a No-no. For country like U.S. or U.K., it's still possible and you can still do it from late 1936(which would be about 3 years before historical WWII). And for minor nations, One division training is still a good option if you are not rushing wars(which is usually not the case for SP minors but is true for historical games).

I was just thinking about a mixed approach:
For example when trainingexercising the USA's 36 starting divisions continuously, we'd get 1*X*N experience and lose 36*L*N equipment across N days (where X and N is whatever a division normally loses and generates per day). By directly going down to one division we only lose N*L equipment but only get C*X*N experience with C much lower than 1 for several months.

If experience gain is relatively more important than equipment loss, it might be better to track the interpolation. I.e. remove one division every 10 days. This way C stays very close to 1 throughout but equipment cost goes down linearly. For example after 340 days (when the division count reaches 1), the total equipment loss would only be half as much as training all starting divisions continuously.
At this point one-division-training from the start lags 235 training days behind in terms of XP, while ramping down the division count only lost 13 days.

This assumes fully equipped starting divisions. With underequipped starting divisions the fighting strength penalty is just as bad as reducing division count, so it is probably best to remove more divisions immediately to get close to full strength and then wait a while with further reductions until the interpolated division count catches up.

 

[Vlad123]

In short world, the exploit is train full equipment division and every 10 day remove 1 division from training?

 

[bitmode]

In short world, the exploit is train full equipment division and every 10 day remove 1 division from training?

Sorry, by "training" I meant exercising.
I deleted all starting divisions but 14 national guards. In July '36, the penalty from reducing divisions is gone and the divisions get fully equipped slightly earlier. From then on I deleted one division every 10 days, reaching one division in November. From then on it is normal one-division-training.

 

[GSP Jr]

The only problem with the 'train-1' scenario is when somebody attacks you and all your divisions (save the golden 1) are quickly destroyed without experience.
How do you work that into the algorithm?

I'll continue to train all my divisions, thanks.

 

[Darknessunder]

I was just thinking about a mixed approach:
For example when trainingexercising the USA's 36 starting divisions continuously, we'd get 1*X*N experience and lose 36*L*N equipment across N days (where X and N is whatever a division normally loses and generates per day). By directly going down to one division we only lose N*L equipment but only get C*X*N experience with C much lower than 1 for several months.

If experience gain is relatively more important than equipment loss, it might be better to track the interpolation. I.e. remove one division every 10 days. This way C stays very close to 1 throughout but equipment cost goes down linearly. For example after 340 days (when the division count reaches 1), the total equipment loss would only be half as much as training all starting divisions continuously.
At this point one-division-training from the start lags 235 training days behind in terms of XP, while ramping down the division count only lost 13 days.

This assumes fully equipped starting divisions. With underequipped starting divisions the fighting strength penalty is just as bad as reducing division count, so it is probably best to remove more divisions immediately to get close to full strength and then wait a while with further reductions until the interpolated division count catches up.

I also have a different mixed approach. I personally would just train the few fully equipped divisions to ensure a usable standing army and not worry about deleting one division every ten days.(Just too much micro to do). Maybe deleting one division or two to reduce pressure on equipment production when increasing tamplets.

 

[Darknessunder]

The only problem with the 'train-1' scenario is when somebody attacks you and all your divisions (save the golden 1) are quickly destroyed without experience.
How do you work that into the algorithm?

I'll continue to train all my divisions, thanks.

I would say do not train all of your divisions for sure(unless you are Germany). The bare minimum is to reduce your army to a point that all of them are fully equipped, which is only applicable for 1936 Germany.

The 25% experience debuff is not that bad considering you will end up deploying untrained divisions in proper wars anyways, and they are decent troops giving some entrenchment.

And the spirit of my post is not necessary to suggest people do one division training, but rather help people to understand their training strategy. For instance, you do not need to train new units if you want more XP, all you need to do is increase the size of your template.

 

[Darknessunder]

Sorry, by "training" I meant exercising.
I deleted all starting divisions but 14 national guards. In July '36, the penalty from reducing divisions is gone and the divisions get fully equipped slightly earlier. From then on I deleted one division every 10 days, reaching one division in November. From then on it is normal one-division-training.

I would suggest to reduce your starting unit to 10 fully equipped national guard and keep training all of them from that point. For countries like the U.S., XP is more important than IC. So training early and increase your templet size early is more useful than saving some equipment from one division training.