Fix Retreats

[pgr]

Does any one think that this thread should also discuss how to avoid/alleviate bad retreats under the current rules?

Im all ears for tips !

[tripax]

Im all ears for tips !

Late in the game when I have the extra resources, I'll sometimes build a depot or a fort in the direction I want to retreat (and don't fight with water in the direction I want to retreat). In general, retreat problems are most common in regions surrounded by water (Island 10) and in regions with multiple cities - especially cities larger than size 1 (central Virginia, and the Mid-Atlantic). In these areas avoid retreats by fighting with commanders with higher strategic (I think that helps avoid retreating when unnecessary and lowers cohesion drop by making travel faster), use B/R when possible, and pay extra attention to your cohesion.

[Jagger2013]

No matter what you do to tweak the Algorithm, you are going to have the oddball solution which is sometimes disasterous. Mathematical solutions lack human logic.

I still believe the solution is a human designated line of communications/retreat created by the selection of a specific town, depot, etc as a retreat designation selected during the plot portion of the turn. If a unit is defeated anytime during a turn, then the AI plots the best path back to the designated town, depot, etc. I would think that would solve the oddball retreats as the human as imput into the ultimate retreat destination if a retreat occurs.

[Ace]

The problem with that is it can create weird situation as well.
For example, you are surrounded, out of supply and wanting to go to depot 2 regions fromyou. But in one region between the depot and you is large enemy force. If current rules are tweaked that the greatest retreat pull is the depot and not the starting region, one could simply attack blocking force, and retreat through it to the desired depot. See what I am getting at? So, there is no simple thing to change here.
If I were to tweak settings, I would:
1) Make retreat impossible to 0%MC regions for all except cavalry and irregulars
2) Make MTSG Corps retreat to original region, not to the region attacked Corps is retreating to (this has been implemented in TEAW and works quite nicely)

Ideally I would love to see the game where retreat options to favorable locations increase the more cavalry you have in your stacks, but that is not easy thing to do and would require some new coding, not tweaking some setting in the existing code.

[pgr]

1) Make retreat impossible to 0%MC regions for all except cavalry and irregulars
2) Make MTSG Corps retreat to original region, not to the region attacked Corps is retreating to (this has been implemented in TEAW and works quite nicely)

Ya, I think it keeps coming back to the minimum MC %, because that is how the system identifies friendly or unfriendly regions. How hard would it be to have the minimum mc while creating an exception for Cav and irregulars? I'm assuming TEAW simply has the 5% minimum rule?

[Pocus]

(2) will be in the next CW2 1.05 public beta then... I have got sidetracked by the EAW official patch to send for validating but I'll check remaining issues on CW2 very soon.

[Captain_Orso]

We used to have that. It was the default. What happened is that although the region in front of a retreating stack was empty, it would not enter that region--because it was afraid . . . . . . . . .

Please, think about what it would mean to make such changes. How would the game play? Don't just think about the best case scenario; think about the worst case scenario. If you could put things together in the worst possible situation, what would that look like and what would happen with such changes in place? Now think about how likely or unlikely it would be for such a situation to come about? Now think again about whether that suggested change is a good idea.


I will reiterate what I believe must and should be done to fix the situation. Please comment on these; good, bad; too little, too much?:

- Parameters causing a pull toward cities, depots and forts should only work on the retreating stack if the structure is friendly owned.
. Importance: Would help fix poor logic and many bad situations.
. Why? Because even if a structure is completely devoid of a defender, the retreating stack--in Passive Posture--can have no benefit from retreating into that region. Thus there can be no logical "pull" in that direction.

- Friendly MC in a region should pull a retreating stack toward that region.
. Importance: Would help fix poor logic and many bad situations.
. Why? A demoralized forces will logically want to get to a place of relative protection, where no enemy is to be found: IE little or no enemy MC.

- No region should be forbidden to enter strictly on the basis of the past: EG based on enemy MC alone. Actual presence should still reduce pull.
. Importance: Current situation. Experience shows that it works.
. Why? Has already been proven to be better the way it currently is. It makes no sense to prevent a retreating stack from entering a region with little to no enemy strength, just because the stack is retreating.

- Friendly SU's should increase the pull toward a region.
. Importance: Would help fix poor logic and many bad situations.
. Why? Better than getting to a region without enemy forces can only be getting to a region with friendly forces.

Situation: The "ctlRetreatPrevSubSpaceCoeffH = 250 // Coefficient applied to the interest if the region is the one where we are coming from". The reason for this parameter is to force stacks moving into a region, if they retreat, to return the where they came before going into battle and retreating.

Fix 1: If there are stacks which MTSG'ed to take part in a battle and a stack must retreat from that battle, all MTSG'ed stacks must return to their original locations, before the retreat-target location(s) is/are determined.
. Importance: High. Situation cause the illogical destruction of retreating stacks.
Why? Otherwise the retreating stack(s) cannot take into account that MTSG'ed stacks will return to their original locations.

Fix 2: The "ctlRetreatPrevSubSpaceCoeffH" parameter should only be in affect the first time a stack retreats in that turn.
. Importance: Very High. Situation cause the illogical destruction of retreating stacks, plus it is very illogical.
. Why? After the first time a stack retreats in a turn, all other retreats will cause that stack to return to a region, where the retreating stack "knows" that a strong enemy force is already waiting on it.

Please comment.

[Ace]

Captain Orso, you made some good remarks here. But let's focus on the most annoying part. It happens often that I push the enemy, while I have flanking force behind him. The enemy is pushed towards my flanking force, than back to my original force and back and forth till the turn's end. I think we should FORBID retreat into regions where enemy has more strength (PWr) than us. It would make flanking possible without the clumsy back and forth marching. The units could still retreat to empty 0 % MC regions, just not to occupied regions with more powerfull military force.

[Pocus]

1 - Retreating toward friendly only structures.
This is supposed to work like this. If it's not, then it's a bug.

2 - Friendly MC pulls retreating force
Already done since years

3 - Not sure I understand, if you allow by modding retreating even in 0% MC, then nothing else will prevent the army from entering the region in theory

4 - Friendly SU pulls retreating force
Already done since years

5 - Big bonus for retreat toward origin region, only the first time
Not done. I'll have to check what it implies...


As I said, the MTSG forces are now put back in EAW into their region and are not subject to retreat. As CW2 shares the same code, it sounds logical to me that CW2 gets the rule too, and the EAW betas are happy with it. Is it a problem.

So what's remain TBD? Check if bugs and add (5) ?

[ohms_law]

As I said, the MTSG forces are now put back in EAW into their region and are not subject to retreat. As CW2 shares the same code, it sounds logical to me that CW2 gets the rule too, and the EAW betas are happy with it. Is it a problem.

It's not a problem.

5 - Big bonus for retreat toward origin region, only the first time
Not done. I'll have to check what it implies...

Please do. It just makes sense.

[Captain_Orso]

1 - Retreating toward friendly only structures.
This is supposed to work like this. If it's not, then it's a bug.

2 - Friendly MC pulls retreating force
Already done since years

3 - Not sure I understand, if you allow by modding retreating even in 0% MC, then nothing else will prevent the army from entering the region in theory

4 - Friendly SU pulls retreating force
Already done since years

5 - Big bonus for retreat toward origin region, only the first time
Not done. I'll have to check what it implies...


As I said, the MTSG forces are now put back in EAW into their region and are not subject to retreat. As CW2 shares the same code, it sounds logical to me that CW2 gets the rule too, and the EAW betas are happy with it. Is it a problem.

So what's remain TBD? Check if bugs and add (5) ?

Many thanks for shedding some light on this issue.

If the code currently work as you've described it, then something is very bugged.

The short of it is that an entire corps, which was not moving, was attached and retreated into a region where enemy MC was 100%, when there were 3 other adjacent regions in which friendly MC was 100%. In one of those 3 regions was a small friendly force and in another a larger friendly force. All 3 of these regions were ignored. In the next turn the same thing happened again. The only think that I can think of, which might entice the retreating corps into retreating to where it did, is that that region has a depot, but in enemy hands.

In regard to a stack which MTSG'ed and its side lost the battle, I would question whether it were logical or realistic for that stack to automatically return to it's starting region.

From simple logic, if a number of corps come together through MTSG and fight a battle, which they loose, why should they then divide up and go their separate ways, especially if one is to consider that they are in the face of an enemy force which just drove them from the field and into retreat.

Historically it can be argued that at Gettysburg Ewell's Corps MTSG'ed from Carlyle, PA down to Gettysburg--in game from the Cumberland, PA region--. When Lee retreated from Gettysburg, Ewell's Corps most certainly did not return to Carlyle.

With regard to 3-, if enemy and friendly MC, the presence of enemy and friendly SU's and the presence of friendly held structures add to and diminish the total pull of each region as described, then there should be no need for a minimum amount of friendly MC to be able to retreat into a region. Implementing such a restriction will only lead to the situations which lead to that restriction being lifted in the first place.

[Ace]

1 - Retreating toward friendly only structures.
This is supposed to work like this. If it's not, then it's a bug.

Perhaps the code checks ownership at the beginning of the turn. Midturn the possesion of friendly structures can change hand and units still retreat towards them?

5 - Big bonus for retreat toward origin region, only the first time
Not done. I'll have to check what it implies...

I am not sure what you mean, let's give this example:
[ATTACH]31850[/ATTACH]
1) Magruder assaults FtMonroe. Under current rule, if Union cannot retreat anywhere else, Union units outside FtMonroe will retreat towards Hampton roads (the very place Magruder held at the start). This kind of retreat is illogical. I am advocating that the code be added so defeated units cannot retreat toward region containing much more units than they have nor to the regions attack originated from(even better would be if the code would check/compare Pwr values).

Example 2) Let's say Magruder has Union division behind him (Westwards of him in between Richmond and Hampton roads). Union division assaults him from the West towards FtMonroe. If there are strong Union forces in FtMonroe, he should have no place to retreat. There has to be possibility to corner units on the map. It would be absolutely wrong to allow Magruder to retreat towards Richmond through unit attacking him only because the code steers him away from FtMonroe...

Example 3) Butler attacks Magruder from Ft Monroe. Magruder retreats 1 region West were he is attacked with another unit already present there. Currently he would retreat towards Butler again which is bad. But even worse would be if he would just pass through flanking units towards Richmond. IMO, the second battle after the first retreat should be to the last men.

[ohms_law]

Example 3) Butler attacks Magruder from Ft Monroe. Magruder retreats 1 region West were he is attacked with another unit already present there. Currently he would retreat towards Butler again which is bad. But even worse would be if he would just pass through flanking units towards Richmond. [color="#FF0000"]IMO, the second battle after the first retreat should be to the last men.[/color]

This is what I really want to see.

[Captain_Orso]

Perhaps the code checks ownership at the beginning of the turn. Midturn the possesion of friendly structures can change hand and units still retreat towards them?

Obviously not! As with everything, changes are enacted on a daily basis--EG cohesion, MC, ownership--everything. Just march through a region with an enemy held town and watch it change to your ownership while your stack is passing through.

I am not sure what you mean,

Then, please read the previous posts. There are examples and descriptions of what is meant.

let's give this example:
[ATTACH]31850[/ATTACH]
1) Magruder assaults FtMonroe. Under current rule, if Union cannot retreat anywhere else, Union units outside FtMonroe will retreat towards Hampton roads (the very place Magruder held at the start). This kind of retreat is illogical. I am advocating that the code be added so defeated units cannot retreat toward region containing much more units than they have nor to the regions attack originated from(even better would be if the code would check/compare Pwr values).

Example 2) Let's say Magruder has Union division behind him (Westwards of him in between Richmond and Hampton roads). Union division assaults him from the West towards FtMonroe. If there are strong Union forces in FtMonroe, he should have no place to retreat. There has to be possibility to corner units on the map. It would be absolutely wrong to allow Magruder to retreat towards Richmond through unit attacking him only because the code steers him away from FtMonroe...

Example 3) Butler attacks Magruder from Ft Monroe. Magruder retreats 1 region West were he is attacked with another unit already present there. Currently he would retreat towards Butler again which is bad. But even worse would be if he would just pass through flanking units towards Richmond. IMO, the second battle after the first retreat should be to the last men.

This is a completely different issue from the one being discussed, and if you'd read the whole thread, you'd know that. Jumping in with -I don't know what your issue is, but this is mine and it's important to me- ....


Honestly, I'm so P.O.'ed at the moment, I don't want to say anything else. Ciao

[Pocus]

Ook, let's proceed with calm and caution, things can get out of hands with this issue... Please no quarrel between gentlemen, the forum don't need that!

I'll first check what is not working as expected, i.e ownership. I also agree that retreating into a region with more power value than you is weird, is everybody ok to forbid that? Then we can proceed with 'only retreat once toward origin region' and 'only retreat once per turn' perhaps, but that's work. The first two are either bugs, or small work.

[pgr]

1 - Retreating toward friendly only structures.
This is supposed to work like this. If it's not, then it's a bug.

2 - Friendly MC pulls retreating force
Already done since years

3 - Not sure I understand, if you allow by modding retreating even in 0% MC, then nothing else will prevent the army from entering the region in theory

4 - Friendly SU pulls retreating force
Already done since years

5 - Big bonus for retreat toward origin region, only the first time
Not done. I'll have to check what it implies...


As I said, the MTSG forces are now put back in EAW into their region and are not subject to retreat. As CW2 shares the same code, it sounds logical to me that CW2 gets the rule too, and the EAW betas are happy with it. Is it a problem.

So what's remain TBD? Check if bugs and add (5) ?

Well based on the two example files I posted, then there needs to be some checking for bugs:

Point 1: The repeted choice to retreat to the Union Depot in Culpeper by Beauregard, shows that the enemy controlling the structure was not being taken into account enough for him to choose another destination.

Point 2: As above, friendly MC didn't pull Beaureguard (or at least not enough). (As an aside, I see nothing in the retreat logic for depots, cities, and forts, that considers their ownership...just their size. The Minimum MC line seems to be the only spot in the retreat logic file that seems to take into account ownership, unless I am missing something)

Point 3: exactly...this seems to be the problem, with the line set to 0 a retreating stack can go anywhere.

Point 4: Didn't seem to happen in the example, and I don't see where it appears in the retreat logic file (although it is pretty clear that enemy units should repel....I don't know why that didn't stop Beaureguard from choosing Culpeper)

Point 5: The ping-ponging seems to imply the preference for province of origin keeps kicking in.

If it helps, I can have FelixZ post the backup files, with orders, for the examples I provided. I think though, that you find that if Grant attacks Beaureguard from Culpeper, Beaureguard consistently chooses to retreat to Culpeper.

[ohms_law]

I'll first check what is not working as expected, i.e ownership. I also agree that retreating into a region with more power value than you is weird, is everybody ok to forbid that? Then we can proceed with 'only retreat once toward origin region' and 'only retreat once per turn' perhaps, but that's work. The first two are either bugs, or small work.

This sounds like a terrific place to start (and possibly finish), to me.

[ohms_law]

Regarding:

Point 2: As above, friendly MC didn't pull Beaureguard (or at least not enough). (As an aside, I see nothing in the retreat logic for depots, cities, and forts, that considers their ownership...just their size. [color="#FFA500"]The Minimum MC line seems to be the only spot in the retreat logic file that seems to take into account ownership[/color], unless I am missing something)

We should be careful not to make assumptions with issues like this. Pocus could shed some light on the issue, but I'm betting that there are some hard coded checks (variables, most likely) along with the constants that are modifiable in the config files. Not everything is going to be in the configs, if for no other reason then you can't put a variable value into a config file.

Speaking of variables: since we're talking about retreating here, I'd like to suggest that units should take the weather into consideration. If the region where the battle is occurring is, or the surrounding regions are, muddy or snow covered, I think that should be taken into consideration (units ought to be less likely to retreat into regions where movement is difficult).
Let's keep in mind that unit movement and the map that we play on is strategic rather than tactical. We're discussing massive unit level routs here, similar to what happened to Pope's Army of Virginia following Second Bull Run where the whole army ran for days back to DC.
Generally, I don't think that units should retreat across provinces at all, as often as they do in CW2.

[Pocus]

By the way I can't check anything without the order files... I tried playing the turn but got nothing worthy, Beauregard is blocked in Spotsylvania, can't retreat and get butchered. Please provide save. Even better, a small scale engagement with stupid retreat...

[tripax]

Let's keep in mind that unit movement and the map that we play on is strategic rather than tactical. We're discussing massive unit level routs here, similar to what happened to Pope's Army of Virginia following Second Bull Run where the whole army ran for days back to DC.
Generally, I don't think that units should retreat across provinces at all, as often as they do in CW2.

I want to second this. Given what this game is, retreats are tricky and I am impressed with how well it does work. In my mind, the strategic goal of an army when contemplating retreat is that they move keep between the opposing army and the opposing army's goal and that they move in the direction of their supplies. The problem with the current rules is that there is often a depot in a neighboring region to a battle that pulls the retreating force but that depot isn't really in the direction from which an army's supplies come.

I know this is a bit off topic, but I agree with ohms_law that retreats (which are necessarily across provinces) are too common in the game. I suppose there might be some creative way to improve this situation, but in lieu of this I think it makes sense to minimize the cost of retreating somewhat, as these involve a situation in which an army is often very vulnerable over a longer amount of time and a larger geographic area than is historically realistic. Good luck and thanks for looking into it.

[pgr]

By the way I can't check anything without the order files... I tried playing the turn but got nothing worthy, Beauregard is blocked in Spotsylvania, can't retreat and get butchered. Please provide save. Even better, a small scale engagement with stupid retreat...

Ya, but I assume he at least tried to go to Culpeper, but the ZOC rules may have prevented him... I'll try to round up the order files and such.

[Ace]

Pgr, was Culpepper last region Beauregard occupied prior to Fredericksburg ?

I did a test here. I attacked CSA cavalry which was standing next to Union controlled Harper Ferry. It didn't retreat there, so enemy depot is not pulling his units towards it (at least it didn't in this test). It retreated to the last region it occupied few turns ago. Here is picture and save game:

[ATTACH]31851[/ATTACH]

[ATTACH]31852[/ATTACH]

Here the most logical retreat route was to Morgan, WW which contained friendly units, and friendly MC.

Question:

Does the code remember last region you occupied even if several turns have passed, so the Cavalry retreated towards Chambers it occupied 2 turns ago?

[Ace]

I did further test:

I built CSA depot in Morgan, WW. But reb cav still insisted to go back to Chambers it occupied few turns ago. Currently, the game is designed to prohibit moves from Chambers to Fulton, then to depot at Morgan if there is enemy force in Fulton. The problem I see with this is it seems the game remembers you arrived at Fulton from Chambers even when several turns have passed in between.

[ATTACH]31853[/ATTACH]

[ATTACH]31854[/ATTACH]

[FelixZ]

Question - Has anyone ever had a retreat to an enemy depot before this alleged one that was not caused by a retreat to the region the stack came from?

I re-ran the turn with one change - Beauregard was moved by rail to Louisa and back to Fredricksburg. Beauregards' subsequent retreat was to Louisa!

I have communicated this to PGR and Captain Orso. The captain also has the files so he can recreate my change.

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

Now why did Beauregard (or the original underlying stack) move from Culpepper to Fredricksburg? At some time in the game there was a stack move plotted from Culpepper to Fredricksburg. The rail button was pushed after the move was plotted. For this move (and others) the game engine does not change a direct move from Culpepper to Fredricksburg by changing to the available rail path.

There was a Union move on the same turn which exhibited identical behavior. Kearny was plotted from Clarke to Culpepper. The rail button was then pushed. But Kearny took 6 days to make the move instead of adopting the rail path which would have only required 2 days.

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

I would not mind if Pocus can make a change to prevent a 2nd retreat to a heavily occupied region. But I expect that most experienced players have intentionally used this tactic to set a trap in AACW and CW2.

There really was no bug in the turn.

[Ace]

It's not a bug in a sense the game behaves as it is coded. But, an improvement in the code to prevent such ping pong movement would be welcome.

[Captain_Orso]

By the way I can't check anything without the order files... I tried playing the turn but got nothing worthy, Beauregard is blocked in Spotsylvania, can't retreat and get butchered. Please provide save. Even better, a small scale engagement with stupid retreat...

I discovered that Felix had sent the hist and order files per email from the turn before the first battle. I ran the turn to see what happens. We probably have different game setting, because Beauregard got mauled even worse through ping-ponging during the execution of that turn, than he did in the original game.

Everything attached.

[pgr]

Ok so I'll comment (and sorry if I quote a bit out of order!)

We used to have that. It was the default. What happened is that although the region in front of a retreating stack was empty, it would not enter that region--because it was afraid . . . . . . . . .

I think this perception is the core of the resistance to the Minimum MC rule for retreats. People keep saying it prevents retreats, with the idea being that the CnC has already decided to retreat and the empty province is preventing the movement.

In my mind, retreating is first and foremost a decision to keep fighting or break contact and withdraw passively to a safe region. Most of the calculations are based on force size and how the battle is going, but the minimum MC rule is the only moment where the question is asked "is there a safe region to retreat to?" (For the stack to go G/G, avoid the enemy, and have a chance of resupply and recovery).

The Minimum MC rule, to me, serves two useful purposes. It screens neighboring regions and excludes enemy ones form consideration as a possible retreat destination. In the rare case that there are no friendly regions, the commander decides to stand his ground and fight it out.

Logically it makes sense that if there is no safe place to flee to, one should stand and fight. In the limited testing I did with the MC rule set to 5% (Gettysburg battle, giving orders to both sides, I surrounded Lee and attempted to overrun him) the stand your ground decision actually worked in the favor of the surrounded stack. The defenders fought on until the attackers were exhausted. At the start of the next turn, the surrounded stack was in much better shape (and in a much better position to attack in a break out attempt) than if it had fought a few rounds, gone passive (received pursuit damage), and risked creating a passive ping-pong situation described above.

From what I see in the logic file (and have observed) is that with the minimum MC disabled, the retreat decision making process no longer asks if there is a "safe" region when deciding to end a battle or when choosing where to retreat. Those questions need to be re-integrated into the retreat process.

Please, think about what it would mean to make such changes. How would the game play? Don't just think about the best case scenario; think about the worst case scenario. If you could put things together in the worst possible situation, what would that look like and what would happen with such changes in place? Now think about how likely or unlikely it would be for such a situation to come about? Now think again about whether that suggested change is a good idea.

I like to think I have thought about it a bit . It seems it would play out a lot like it does in other AGE based games. (Correct me if I am wrong, but ACW is the only title where the rule has been disabled.)

On the positive side, retreats would become more logical. Retreats would only move towards provinces with a minimum amount of friendly MC. It would entirely prevent stacks from moving to entirely enemy controlled regions when friendly ones are available. This prevents certain absurdities, like stack retreating into an enemy stack, a stack retreating into the province vacated by the stack attacking it (I have seen that "swap" happen), or a stack attempting to retreat to a 100% enemy controlled province and failing to arrive because ZOC rules blocked it from arriving.

The disadvantage, a stack completely surrounded would stand its ground and fight on. In a lot of cases, I don't see this as a problem. There is no combat disadvantage to being surrounded. On the defensive, you benefit from all defensive bonuses. Staying there is better than going passive and "attacking" into enemy MC controlled provinces. (Remember that your stack will take an attack posture on entering a completely enemy province with troops, and may have river crossing penalties etc.) In addition, the fight it out approach allows a stack to survive and the player to plan his break out attempt in the next turn. Current "retreats" into enemy territory are effectively break-out attacks that the player has no control over.

Deep raiders would be relatively more venerable to being destroyed, and the AI might have a harder time of it. With raiders though, it just means keeping them moving and spreading them out. If they are in a B/G stance and avoid combat, they automatically generate 5% MC in every province they enter and thus safe retreat areas. If raiding is really impacted, then an exclusion to the Minimum MC rule could be created for cav and irregulars could be created.

The AI problem (which I guess was the main justification for disabling the minimum MC rule) seemed to be that it was too easy to cut off and destroy enemy advances. That's a problem that still exists, primarily because the AI tends not to protect its flanks and rear, and tends to hang on to forward positions. The focus should be on tweaking AI campaign behavior rather than making retreats effectively "blind."

On the balance, I think the benefits of rational predictable retreats governed by the min MC rule out way the costs.

I will reiterate what I believe must and should be done to fix the situation. Please comment on these; good, bad; too little, too much?:

- Parameters causing a pull toward cities, depots and forts should only work on the retreating stack if the structure is [color="#FF0000"]friendly owned[/color].
. Importance: Would help fix poor logic and many bad situations.
. Why? Because even if a structure is completely devoid of a defender, the retreating stack--in Passive Posture--can have no benefit from retreating into that region. Thus there can be no logical "pull" in that direction.

I think this is absolutely correct. The critical bit is how do you determine what structures are "friendly." It seems the game is very clear MC=Ownership. Your point about the passive posture is exactly right. It is a retreat is a passive move, and it only makes sense to do it in a direction where you have ownership i.e. MC.

- Friendly MC in a region should pull a retreating stack toward that region.
. Importance: Would help fix poor logic and many bad situations.
. Why? A demoralized forces will logically want to get to a place of relative protection, where no enemy is to be found: IE little or no enemy MC.

Good idea. Concretely, this would look a lot like the ctlRetreatPrevSubSpaceCoeffH, except the coefficient would be a function of friendly MC. If done it should be BIG, so that % of MC becomes the dominant factor. Forces should consider 100% friendly MC regions as THE MOST IMPORTANT consideration of retreat location.

My only criticism would be that it seems like a duplication of what already exists. Exclude low % on friendly MC is effectively the same as saying go to high friendly % MC regions. If you really want the Min MC rule dead, then this criteria needs to be created, and made big enough to replace min MC.

- No region should be forbidden to enter strictly on the basis of the past: EG based on enemy MC alone. Actual presence should still reduce pull.
. Importance: Current situation. Experience shows that it works.
. Why? Has already been proven to be better the way it currently is. It makes no sense to prevent a retreating stack from entering a region with little to no enemy strength, just because the stack is retreating.

I don't know about this. The problem with not excluding regions as retreat targets, is that all regions could still see retreats. This is precisely the problem that current experience seems to be illustrating. By having an exclusion rule, "enemy" provinces are immediately ruled out and any province selected afterwords will have a high probability of a successful retreat movement. (Not bouncing into enemy units or having the retreat stopped due to ZOC issues)

The ultimate issue is should a stack with no neighboring region with a minimum of friendly MC choose to retreat, or to stand its ground. I vote for them fighting in place, for reasons cited above.

- Friendly SU's should increase the pull toward a region.
. Importance: Would help fix poor logic and many bad situations.
. Why? Better than getting to a region without enemy forces can only be getting to a region with friendly forces.

A good idea, retreating forces should tend to concentrate. I suppose it would look a lot like the cltRetPenaltyPerNmySU rule except friendly SU would have a pull effect. My concern is that my example seems to illustrate the the enemy SU repel effect didn't seem to stop the retreat, even though it should have. So if cltRetPenaltyPerNmySU isn't working properly, it will need to be sorted out before we can expect a "pull" equation to have some kind of effect.

Situation: The "ctlRetreatPrevSubSpaceCoeffH = 250 // Coefficient applied to the interest if the region is the one where we are coming from". The reason for this parameter is to force stacks moving into a region, if they retreat, to return the where they came before going into battle and retreating.

Seems like this should only apply if a stack was attacking into a neighboring region. (Although if retreats always happen away from enemy MC regions, then ping-pongs wouldn't happen.)

Fix 1: If there are stacks which MTSG'ed to take part in a battle and a stack must retreat from that battle, all MTSG'ed stacks must return to their original locations, before the retreat-target location(s) is/are determined.
. Importance: High. Situation cause the illogical destruction of retreating stacks.
Why? Otherwise the retreating stack(s) cannot take into account that MTSG'ed stacks will return to their original locations.

Sure, good idea.

Fix 2: The "ctlRetreatPrevSubSpaceCoeffH" parameter should only be in affect the first time a stack retreats in that turn.
. Importance: Very High. Situation cause the illogical destruction of retreating stacks, plus it is very illogical.
. Why? After the first time a stack retreats in a turn, all other retreats will cause that stack to return to a region, where the retreating stack "knows" that a strong enemy force is already waiting on it.

makes sense.

To conclude, there seems to be a problem of blind retreating, or retreating into hostile areas. The AGE engine avoids the problem by using MC to define who owns provinces, and excluding completely enemy controlled provinces from being valid retreat targets. (And then favoring forts, cities, and depots as retreat destinations). The MC exclusion step has been dis-activated, so stacks are choosing retreat locations based on cities, forts, and depots without regard to enemy MC or presence. Theoretically this shouldn't be possible, because cltRetPenaltyPerNmySU should repel retreaters. Perhaps it is failing because those provinces appear "empty" when chosen as retreat destinations.

The simplest solution would be to reactivate the min MC rule, and create exclusions for unit types that we would like to be able to see keep the ability to retreat into 100% enemy MC regions.

I am not against new coding that would favor friendly MC regions rather than excluding enemy ones, but it seems like a more complicated (and perhaps unreliable) approach.

[pgr]

Pgr, was Culpepper last region Beauregard occupied prior to Fredericksburg ?

FelixZ is fond of the theory that there is some memory of where the stack was that led the previously occupied province factor to kick in. It's possible but...

In my case, Beaureguard was never in Culpeper. The stack he commands got shuffled quite a bit, with Longstreet commanding, divisions coming and going etc. But none of the divisions in the stack, or the commanders, were ever in Culpeper. (The exception was that several turns before the F-Burg corps MTSG to support Jackson defending Culpeper, but I would hope that that wouldn't cause it).

[pgr]

Question - Has anyone ever had a retreat to an enemy depot before this alleged one?

Of the top of my head in a current tournament game I am hosting, Jim-NC and Ace me thinks, CSA troops in Winchester retreated during a battle and chose 100% Union controlled Harper's Ferry (with a garrison if memory serves) over CSA controlled Strasbourg (with its friendly depot).

I also seem to recall AI games where a Union stack next to Richmond was attacked by a corps coming from Richmond and chose to retreat to my waiting army in Richmond rather than move away in the direction of FT Monroe. Of course the memory could be a bit fuzzy

[Ace]

In my game vs Jim it was again the "ctlRetreatPrevSubSpaceCoeffH = 250" dominating retreat choice. His stack was in Harper's Ferry, pushed to Winchester and attacked again in Winchester since I was advancing to Winchester through Harper's Ferry causing the undesired ping-pong effect. It is problem as the retreat choice when the stack is defending. When the stack is attacking it is and should remain the rule with the most pull.