Slitherine

quote:

ORIGINAL: haromar

Since this is the "general" AI section, here some general stuff that should apply to all nations.

Is this the correct thread?

Some of this stuff is obvious (not all though ), but it also has to be documented, correct?

- In case the hex can be reasonably attacked, stack a third inexpensive unit to take as loss. E.g. If out in the open with an arm or mech, put a cav div on top to take as first loss. Not true if the hex is critical, in that case put another AA or AT on top. Also not true if no 3rd unit available.

- When attacking in blitz 2D10, take a Mot Div or Mech Div as potential loss.

- Place units so as to Zoc out all potential invasion hexes (applies to Egypt, Palestina, Italy etc.).

- Single Strategic Bomb hexes w/o enemy FTR cover. First in normal range, then double range. Then hexes with enemy FTR cover in single range with own FTR escort, then double range w/o own FTR but no worse than -1/+1. Follow this tactic until you can mass enough bombers (applies to allies only) that you are actually trying to destroy Factories.

- In Air to Air, place best AtA rated FTR in front. In case LBA and CVP are same value, CVP in front. The exceptions:
a. You must fight or want to fight more than one round of combat, i.e. do not want to abort, and you expect your opponent to do the same. You might place the 2nd best rated FTR in front so as to gain depth after the first AA or AX is rolled. This is especially true if it allows you to then place a CVP as best front FTR.
b. Your best rated FTR is orange circled. e.g. CW 1940 with 5-8 3 turn FTR. In that case, you might opt to put the 3 CVP in front giving 4 straight instead of 5.3 weakened.
c. You're only LBA or your opponent cannot do much damage with his naval bombers. You are fighting at negative AtA due to surpirse points or not having all sections included. You do not have naval bombers to absorb losses. You plan on aborting the AtA combat and search again or abort the sea zone. You have no good FTR arriving in this area for the next turn but must fight in the area. You expect your opponent to pull out FTR out of this area. Hide the best FTR.

- In large Air to Air, always take the AA or AX on the FTR. Exception, the combat is almost over, then abort the last or almost last enemy naval bomber

Is this the correct thread for stuff like this?
To be continued ...

quote:

ORIGINAL: DarthCycle

An interesting topic, I loved thread on AI.

For me, a good AI must have at least 3 components: a plan, a strategy that defines the steps that are required in order to realize the plan and tactics that execute each step of a strategy.

A human player thinks of all these elements at the same time (sometimes, without realizing it) when playing a game.

Within the context of a game like WIF, this could translate into (let's imagine we're emulating an Axis player playing Germany):
1-Overall plan: winning the war (ok simple enough statement)
2-Plan: Invade Poland, Invade Low-Countries, Invade France, Defend from England, Invade Soviet Union.
3-Strategy: (one example only) Invade Poland: declare war, invade through western border, capture city A-B-C ...
4-Tactics for capture of city A: identify weakest defending stack within 2 hex of objective (city A), concentrate land force within range in order to get 2:1 ratio, ...

An AI that only deals with tactics but no plan will not achieve anything. An AI with a plan but poor tactics will also not achieve anything.

This is especially important in a game like WIF. Without a plan (a vision), the AI will only focus on short term gains: conquering territories, winning battles, limit its losses. Individually, these actions are all important but they only become meaningfull when they are part of a plan, implying that each action are complementary.

IMO, this is where most AI fail. There is no cohesion to AI action from turn to turn. AI also become extremely predictable.

From a game design perspective, if you can modelize an AI that has these 3 levels of decision (plan/strategy/tactic), you can modelize separate behavior in unique objects and these objects are reusable from one plan/strategy to another.

Another advantage of this design model: you can mimick behavior from top players. Find a good USA player. Get him to explain his plan, each strategy step and each tactic (relevant to USA action). You now have a good USA AI. Another twist is to have 3 (or more) unique USA plan, and make the AI choose one randomly at the beginning of each game. You now have a USA AI that will play very differently from game to game (you don't know which plan is in effect). Combine this for each major power and the various AI combinations makes for very different gameplay behavior for each game. When a new strategy or tactic is available (based on players feedback or modders), you can incorporate it into the relevant AI plan.

Food for thoughts.

quote:

ORIGINAL: composer99

I would not make disorganizing enemy units one of the goals of a land combat - after all, ideally you are attacking hexes where the units have already mostly been disorganized by your ground strikes.

Also, there is what, exactly one result where the enemy can be disorganized after an attack without also being retreated or shattered, right?

quote:

ORIGINAL: Klingon

The goal of military operations is to bring a rapid and successful conclusion to the war, not to destroy the enemy's forces; Sun Tzu understood this very well.

quote:

ORIGINAL: Shannon V. OKeets

Can we get back to the very narrow question I posed?

How do you choose between Assault or Blitz? As the attacker? As the defender?

quote:

ORIGINAL: Shannon V. OKeets

Can we get back to the very narrow question I posed?

How do you choose between Assault or Blitz? As the attacker? As the defender?

quote:

ORIGINAL: Shannon V. OKeets

Can we get back to the very narrow question I posed?

How do you choose between Assault or Blitz? As the attacker? As the defender?

quote:

ORIGINAL: HanBarca

quote:

ORIGINAL: Shannon V. OKeets

Can we get back to the very narrow question I posed?

How do you choose between Assault or Blitz? As the attacker? As the defender?

actually, the thread title is: "AI (general): location v. destruction of enemy", so I think you are the one off topic ;)

quote:

ORIGINAL: HanBarca

quote:

ORIGINAL: Shannon V. OKeets

Can we get back to the very narrow question I posed?

How do you choose between Assault or Blitz? As the attacker? As the defender?

actually, the thread title is: "AI (general): location v. destruction of enemy", so I think you are the one off topic ;)

quote:

ORIGINAL: Mziln

quote:

ORIGINAL: HanBarca

quote:

ORIGINAL: Shannon V. OKeets

Can we get back to the very narrow question I posed?

How do you choose between Assault or Blitz? As the attacker? As the defender?

actually, the thread title is: "AI (general): location v. destruction of enemy", so I think you are the one off topic ;)

Nope, on topic. You have to figure how to make the AI choose a combat table

But I changed the title for you just in case

quote:

ORIGINAL: Mziln
But I changed the title for you just in case

quote:

ORIGINAL: Shannon V. OKeets
Sorry. I could have started a new thread for the question (probably should have) but I try not to start a new thread for every little thing, but instead put them with similar (usually broader) topics.

quote:

ORIGINAL: Froonp

As the attacker, I would choose Blitz everytime I can, because :
- Blitz are more easy at taking the hex than Assaults (result needed to take the hex is 16 in Blitz, and 20 in Assault).
- Blitz inflict less casualties to me than Assault.
- Blitz disrupt me less than Assault.

The only drawback of Blitz is that it directly kills the enemy less often than Assault. BUT, when the Assault kills the enemy, the Blitz retreat or Shatter them. An enemy retreated is an enemy disrupted, so it can be killed latter. Also, If a shattered (and retreated) enemy can't be killed later, it can be shattered instead. Also, when the assault table kills, at the same numbers the Blitz table Shatters (B) the enemy, which is as good as a kill sometimes.


As the defender, I'd choose Assault everytime I can, simply because it denies the Blitz benefits to the attacker. There are exceptions. I would choose the Blitz table sometimes time when I don't care of loosing ground and when I wish to save units, and the enemy only have low odds against me, so that the survivablity of my units is greater. China and Russia are two places where you will do that sometimes.

quote:

ORIGINAL: Shannon V. OKeets

Here is a problem I am working on fleshing out the answer to. What I want to achieve is a written description of the process for determining which CRT to use. I have made a small start on this and I am interested in your opinions as to how you make this decision. Most of this writeup is for when attacking. What needs to be different if you are defending?

The 3 Re is the level of difficulty I assigned to writing this rule (3: where 1 - easy and 5 - very hard). Re denotes that the the AIO has to Respond when asked this question.

===========
8.33 Resolve land combat - choose table (see RAC 11.16) (3 Re)
Calculate the expected results for both tables (Assault and Blitz). Results are pessimistic, expected, and optimistic results. The extremes occur 10% or less of the time. These 10% and 90% thresholds can be modified whenever appropriate (e.g., to 20% and 60%).

When attacking, possible goals are:
(1) to empty a hex so we can occupy it,
(2) to achieve a breakthrough,
(3) to kill enemy units,
(4) to preserve our own units,
(5) to disorganize enemy units, or
(6) to not disorganize our own units.

If the mode of attack (see below) is to push the enemy back, then the choice is between 1 and 2. If the mode of attack is to kill enemy units, then the choice is between 3 and 4. If the mode of attack is to disorganize the enemy, then the choice is between 5 and 6.

When defending, ...


======================
Tactical mode
What the Field Marshal (FM) needs to determine is which of 4 modes of attack to use:
(1) Destroy enemy units this impulse,
(2) Disorganize enemy units this impulse so they can be destroyed in a later impulse (or turn),
(3) Maneuver so better attacks can be made in the future, or
(4) Push the enemy back.

Maneuvering can change supply status for friendly and enemy units, increase the number of hexes from which to attack an enemy hex, and improve the selection of hexes from which to attack. In situations where the enemy has the ability to counterattack, maneuvering can improve the FM’s resulting defensive line at the end of his impulse.

The calculation of estimated results provides the expected changes (for both sides) to the front line. This includes kills and disorganizations. In addition to combat results, units can also become disorganized because they were committed during the impulse (e.g., air units). And disorganized units might be reorganized. When the impulse is over, the FM makes a new assessment, ‘B’, for each front line., and compares it to ‘A’.

By extrapolating the change from A to B over the remaining impulses in the turn, the FM judges whether disorganization is a viable tactic for the current turn. It works if the enemy units are mostly disorganized with no reorganization capability left while the FM still has a viable, organized, attacking force.

Direct attacks to kill enemy units is usually the best tactic but it might cause too many friendly casualties or disorganizations. Or, if the FM limits attacks to only those with excellent odds, it might be too slow. Yet again, maneuvering is rarely fast and there might not be enough impulses in the turn to use the disorganization tactic. The FM must be willing to accept that none of the modes will work as well as he would like and simply go with half measures, or no attacks at all.

quote:

ORIGINAL: ullern

quote:

ORIGINAL: Shannon V. OKeets

Here is a problem I am working on fleshing out the answer to. What I want to achieve is a written description of the process for determining which CRT to use. I have made a small start on this and I am interested in your opinions as to how you make this decision. Most of this writeup is for when attacking. What needs to be different if you are defending?

The 3 Re is the level of difficulty I assigned to writing this rule (3: where 1 - easy and 5 - very hard). Re denotes that the the AIO has to Respond when asked this question.

===========
8.33 Resolve land combat - choose table (see RAC 11.16) (3 Re)
Calculate the expected results for both tables (Assault and Blitz). Results are pessimistic, expected, and optimistic results. The extremes occur 10% or less of the time. These 10% and 90% thresholds can be modified whenever appropriate (e.g., to 20% and 60%).

When attacking, possible goals are:
(1) to empty a hex so we can occupy it,
(2) to achieve a breakthrough,
(3) to kill enemy units,
(4) to preserve our own units,
(5) to disorganize enemy units, or
(6) to not disorganize our own units.

If the mode of attack (see below) is to push the enemy back, then the choice is between 1 and 2. If the mode of attack is to kill enemy units, then the choice is between 3 and 4. If the mode of attack is to disorganize the enemy, then the choice is between 5 and 6.

When defending, ...


======================
Tactical mode
What the Field Marshal (FM) needs to determine is which of 4 modes of attack to use:
(1) Destroy enemy units this impulse,
(2) Disorganize enemy units this impulse so they can be destroyed in a later impulse (or turn),
(3) Maneuver so better attacks can be made in the future, or
(4) Push the enemy back.

Maneuvering can change supply status for friendly and enemy units, increase the number of hexes from which to attack an enemy hex, and improve the selection of hexes from which to attack. In situations where the enemy has the ability to counterattack, maneuvering can improve the FM’s resulting defensive line at the end of his impulse.

The calculation of estimated results provides the expected changes (for both sides) to the front line. This includes kills and disorganizations. In addition to combat results, units can also become disorganized because they were committed during the impulse (e.g., air units). And disorganized units might be reorganized. When the impulse is over, the FM makes a new assessment, ‘B’, for each front line., and compares it to ‘A’.

By extrapolating the change from A to B over the remaining impulses in the turn, the FM judges whether disorganization is a viable tactic for the current turn. It works if the enemy units are mostly disorganized with no reorganization capability left while the FM still has a viable, organized, attacking force.

Direct attacks to kill enemy units is usually the best tactic but it might cause too many friendly casualties or disorganizations. Or, if the FM limits attacks to only those with excellent odds, it might be too slow. Yet again, maneuvering is rarely fast and there might not be enough impulses in the turn to use the disorganization tactic. The FM must be willing to accept that none of the modes will work as well as he would like and simply go with half measures, or no attacks at all.

I think your suggestion is quite good Steve.
And I would want to keep all the possibilities you have suggested.
The other arguments presented so far I feel is very Axis oriented, and early game (pre Barbarossa) oriented. And there are better ways if you are Allied.

If you are genuinly out to take out as many units as possible, assult is better also as an attacker. Sure if you get Blitz bonuses and the choice is between +8 blitz or +4 assult the choice is easy. But in many cases you don't get Blitz bonuses and the choice is between a +4 Blitz and a +4 assult. It's quite a different matter.

With the Allies I am often able to get a production ratio of 2:1 on the Axis. With such production ratio I am happy to accept +6 assult in small size attacks (max two hexes attacking - if not too many units flips), which I did quite a few of in my latestet Campaign, which I won.

Example:
With attacker prefered losses beeing (in order): MOT division, MIL, INF
With defender losses beeing (in order): INF, INF.
The average +6 assult would be: attacker 2.3 BP loss, defender 3.3 BP loss
The average +6 Blitz would be: attacker 1.2 BP loss, defender 1.4 BP loss.
If I am having 2:1 production ratio, doubling the attrition in combat would increase my strategic tempo, so much that I would accept quite a lot of additionally flipped units for the increased attrition and increased loss ratio.
_ As a defender this is usually even worse than what I picture here, since there second defender loss is often a high value unit like a ARM, MECH or HQ.

Disorganizing units also becomes important when dealing with OOS units. Ground striking the units is not always the best choice. Even Germany with excelent ground strikers are sometimes so pressed for air missions that they should seek other choices.
A more common use for disorganize, may be for fast moving fronts. If the Russians are running, a disorganized Russian is a dead Russian. A simple R results can be very valuable.

quote:

ORIGINAL: HanBarca

The alghoritm should work, but be careful with computer calculation based on the paradigm EVALUATE_HEX - CALCULATE_PROBABILITIES - TAKE_ACTION.

Almost every AI out there decide by taking a single, isolated problem, making some calcs and choosing a course of action.
Usually, the result is an AI that keeps sending troops trying to conquer the enemy capital and ignore completely the fact that it's going to be encircled, or that keeps sending out piecemeal units trying to conquer a position instead of waiting a couple of turns and then make a single, decisive attack.

Back to our Blitz vs Assault, what are you going to do, as Germany in '40, if France is practically falled and you have the chance to destroy some big, juicy UK land units there? An EVALUATE_HEX procedure would almost surely use blitz, but what you need there is assault.

And if you're in front of Cairo with your (only) 2 german units in africa, are you going to use assault, maybe losing 1 ?

What I'm suggesting is to build an AI with a more strategic view; as far as i know, only "railroaded" AI are able to do that (AI that choose among predetermined behavior, like "France40=Blitz1","Russia41=Blitz2" and so on)

quote:

I think your suggestion is quite good Steve.
And I would want to keep all the possibilities you have suggested.
The other arguments presented so far I feel is very Axis oriented, and early game (pre Barbarossa) oriented. And there are better ways if you are Allied.

If you are genuinly out to take out as many units as possible, assult is better also as an attacker. Sure if you get Blitz bonuses and the choice is between +8 blitz or +4 assult the choice is easy. But in many cases you don't get Blitz bonuses and the choice is between a +4 Blitz and a +4 assult. It's quite a different matter.

With the Allies I am often able to get a production ratio of 2:1 on the Axis. With such production ratio I am happy to accept +6 assult in small size attacks (max two hexes attacking - if not too many units flips), which I did quite a few of in my latestet Campaign, which I won.

quote:

ORIGINAL: peskpesk

If we focues only on the attacker choice.

The default choose is Blitz, because:
- Blitz gives a higher chance of taking the hex than Assaults
- Blitz inflicts fewer casualties to attacker than Assault.
- Blitz disrupts the attacker less than Assault.
- Blitz gives a chance of a breakthrough.
- Defending units with no valid retreat path are killed.

Possible exceptions are:
- Not having adequate casualties on hand (usually a motorized division).
- Not having chance for mop up attacks (usually late in turn or risk of bad weather).
- Not wanting the defending units to escape (usually an expensive or key unit(s), ex Chinese HQ).
- Expendable unit attack (usually a low DRM attack with a expendable unit(s) on a expensive or key unit(s), ex as Italy in front of Cairo you made successful GS and flipped the juicy UK land units that also made them oos, risking 3-2 attack with your 3-3 against the two flipped black Mot’s).
- Attrition (a usually Japan vs China or an Allies late war strategy. Because the higher production ratio reduces importance of the BP losses ratio).