About plan valuation

[bitmode]

I was going to write a post about how the AI internally decides to start or stop executing battle plans and it is as wonky as you probably think it is. But one feature the AI partly relies on for that is player-visible and just as wonky, so I'll bring that up first: estimated plan value, i.e. this:

The value also decides whether there's a green tick, yellow dots or red exclamation mark next to the plan execution arrow. I guess most experienced players are not paying much attention to this value because it only makes sense in rare circumstances. Here's how it's calculated:

Spoiler: calculation of plan value

• value starts at 0%
• look at air superiority in relevant strategic regions
• enemy advantage yields -10%
• allied advantage +10%
• compare armies
• if enemy is at least twice as strong: -50%
• if own units at least 50% stronger: +40%, up to +80% if they are twice as strong
• if there are no enemies at all: +100% (currently broken)
• large river touching more than 8% of provinces: -10%
• strait touching more than 8% of provinces: -10%
• more than one level of fortification on average: -50%
• any own unit still planning: -20%
• any own unit moving: -50%

Where the "strength" of each division is (soft_attack + hard_attack + defense + breakthrough + hardness + armor + recon + max_entrenchment + suppression) * HP_percentage * Org_percentage. No, I'm not making this up.

This is a "risky" plan (i.e. neither very advantageous nor disadvantageous):

This plan is exactly as risky:

This plan is worse than both of them because of the river crossing:

Do you use the plan value in your games? How would you redefine it?

 

[elektrizikekswerk]

This plan is worth than both of them because of the river crossing:

In this case: What river?

For the rest: Yeah... I tend to ignore that information and just press "Go". The only parts I'm (sometimes) waiting for to disappear are the "division moving" and "division preparing" modifiers.

 

[bitmode]

In this case: What river?

For the rest: Yeah... I tend to ignore that information and just press "Go". The only parts I'm (sometimes) waiting for to disappear are the "division moving" and "division preparing" modifiers.

A river is touching the southern corner of the attack line

 

[bERt0r]

Do you use the plan value in your games? How would you redefine it?

No, the whole calculation should be thrown out as well as the way battleplans currently work like.
I made a thread about the advantage of "dumb" battleplans in comparison to the "smart" AI controlled we have now. If plans worked like the image below, a preset sequence of troop movements and battles, they would be much easier to approximate for the AI.

 

[krios41]

Do you use the plan value in your games?

Yes, but i generaly only care if it sais "superior enemy" or "inferior" enemy because it lets me know where i have to micro the most, but that's about it.

 

[bitmode]

No, the whole calculation should be thrown out as well as the way battleplans currently work like.
I made a thread about the advantage of "dumb" battleplans in comparison to the "smart" AI controlled we have now. If plans worked like the image below, a preset sequence of troop movements and battles, they would be much easier to approximate for the AI.

What you call dumb is much more advanced than what the AI does right now (edit: assuming you still account for fortifications, rivers and such)

 

[bERt0r]

What you call dumb is much more advanced than what the AI does right now (edit: assuming you still account for fortifications, rivers and such)

I disagree. The AI reevaluates the situation on the front every game hour and bases it's actions on that analysis. With a dumb plan that is evaluated once you have something to stick to so you don't need to reevaluate the frontline every hour. Any amounts of enemy divisions do not have to be part of a dumb plan, you base battleplans primarily on terrain because enemy units can move anytime, and should they move in a way to spoil your plan, you have to hand it to the ability of your opponent.

Now for the AI to decide whether a plan is feasable or not, it can now calculate the exact attack orders with the divisions that are attacking and the divisions it knows are defending the line.

 

[bitmode]

I disagree. The AI reevaluates the situation on the front every game hour and bases it's actions on that analysis. With a dumb plan that is evaluated once you have something to stick to so you don't need to reevaluate the frontline every hour.

The AI already evaluates plans only once per day. I don't think evaluating any less frequently would be a good idea regardless of the other aspects.

Now for the AI to decide whether a plan is feasable or not, it can now calculate the exact attack orders with the divisions that are attacking and the divisions it knows are defending the line.

Again this is not a dumber plan, because it has to look at each front province individually and decide what to do their (which it does not right now).

should they move in a way to spoil your plan, you have to hand it to the ability of your opponent.

Should they move in a way to spoil your plan, you should stop/change the plan.

 

[bERt0r]

-.- You didn't get it at all. The AI decides what division moves to or attacks what province on an hourly basis. That's part of what's slowing down the game in later years. The dumb plan does not do that. It makes the decisions what units go where once (when you draw the plan) and then executes the orders. And of course, when your enemy spoils your plan you have to react. You can immediatly send a popup to the player that the plan he executed failed instead of your units getting overrun while you're busy somewhere else.

The dumb plan has to look at each province that's within the proposed offensive line once. The current battleplan does so all the time, shuffling units and still not being able to pull off concerted attacks.

 

[Vohen]

No, the whole calculation should be thrown out as well as the way battleplans currently work like.
I made a thread about the advantage of "dumb" battleplans in comparison to the "smart" AI controlled we have now. If plans worked like the image below, a preset sequence of troop movements and battles, they would be much easier to approximate for the AI.

One thing that I miss quite often in my games is the concept of "phase".
Maybe not in the micro sense that is shown in the image, but I think it would help the AI as well to have short, mid and long term objectives with its frontlines, as to better concentrate their forces and mount more dangerous attacks more often.
That and I also think it would be more beneficial to have the battleplan assigned to a front instead of an army.
Those armies would have different roles in those fronts, like basic infantry army to hold the line, advanced inf and maybe heavy tank armies to attack and push the line, light/medium tank army to be concentrated in two specific favorable spots to attack and meet to attempt an encirclement, etc etc.
In particular, the encirclement tactic could have its scope adapted by the AI depending on the size of that army (like, choose points A and B as starting points and point X as the meeting point to close the pocket, the distances between A, B and X could increase or decrease depending on the size of that tank army for a bigger or smaller pocket).
These strategies, and army and division compositions could be all adaptable to fit the doctrine they have chosen, among possibly other things.
I think those changes would give a lot of fluidity to the AI, and I would expect them to become quite menacing as a result, maybe even obligating players to have reserves in fallback lines to hold potentially dangerous assaults, or pay attention to the front to figure out when and where they are coming (with encryption/decryption possibly playing a bigger role).

 

[bitmode]

Those armies would have different roles in those fronts, like basic infantry army to hold the line, advanced inf and maybe heavy tank armies to attack and push the line, light/medium tank army to be concentrated in two specific favorable spots to attack and meet to attempt an encirclement, etc etc.

Even just the ability to concentrate armor into an army and using the spearhead command (though that's only available with TfV) for a straight push would provide some of these benefits.
Outside of production and garrisons, I'm not so sure the AI can currently even tell difference between an infantry division and a tank division

 

[AEhere]

Got here from your other AI battleplan thread, thinking my jaw couldn't drop any lower.

I stand corrected.

 

[Vohen]

Even just the ability to concentrate armor into an army and using the spearhead command (though that's only available with TfV) for a straight push would provide some of these benefits.
Outside of production and garrisons, I'm not so sure the AI can currently even tell difference between an infantry division and a tank division

Indeed, the main point would be for tank armies to behave independently
But I also think armies should have different compositions to fill different roles (influenced by factors such as doctrines and tech), which would dramatically increase production and war efficiency as a whole imo.
And when you consider that they already introduced template roles (can't remember if it was on 1.4 or 1.5), I figure it would be the next logical step, even more so when you consider that those template roles are pretty meaningless for anything besides garrison and suppression.

As for my phases and front argument, I feel like it would give a lot of fluidity to the AI, being able to act and react dynamically to the war, and place an far bigger emphasis on intel on both sides of the front.
I think it's unlikely for them to actually change the game to that degree (at least on a mid term), but in the future that wight be even be a good way to reintroduce corps to the game, and deepen even further those roles (I can dream).