Retreat

[runyan99]

The retreat logic the game uses can be a real killer. If for example, the Confederates try to defend Memphis and lose, the computer retreats west across the Mississippi, where there is a level 1 structure (port), rather than south along the rail lines to Jackson, protecting the Confederate rear areas.

It is not very realistic to retreat in a direction away from an army's communications. Manstein in our discussion suggested the solution can be that a defeated force must be retreated one zone to the closest controlled depot independently who has the military control of that zone.

At first that seems like it would be better logic to use. Using the Memphis example though, a force at Memphis would still retreat west towards the depot at Madison, AR or even Little Rock, AR both of which appear closer to Memphis than Jackson, MS.

Is there any way to improve the logic so that armies retreat in the direction from which they draw supply?

Or, getting into new feature territory that can perhaps only be applied to future games, could certain regions be 'flagged' as beacons, and then units forced to retreat would always attempt to move towards the closest one? Players could 'flag' important regional cities and depots as locations to retreat towards.

There is room for improvement in the current Military Control and towards the nearest city checks I believe the game currently does. These currently result in non-sensical movements oftentimes.

[runyan99]

One possible solution I have always advocated is to force fewer retreats altogether. If the battle is close, let the bias be for the armies to SHARE the region until the end of the turn without further combat. Put both armies in defensive stance until the end of the turn. This allows the players to choose their movements and stances again at the start of the next turn.

Zone of Contol rules, already in place, will prevent one army from 'moving through' another, even if they share a region.

The regions are big, and there is room to accomodate two armies. The game(s) should be more willing for enemy forces to share space.

Civil War armies, once in contact, were generally slow to move very far away after a battle. Lee's army, for example, remained in the vicinity of Gettysburg for some days after the battle, before beginning the move southwest to Virginia.

Except for the rare rout, let the players move the units back and forth. Player control is good, and forced moves are bad.

[Gray_Lensman]

deleted

[Pocus]

I'm surprised that the variables controlling the destination region are not enough to have the work done properly... I can always add another, or you can tweak the existing one. Check Control&Retreat.opt

ctlContested = 5 // Minimum control gained upon entering a region (if not passive)
ctlAllowRetreat = 0 // Minimum control to have in a region to allow a retreat into it
ctlRetreatAdjCity = 5 // Interest in retreating toward a region with a city (per level)
ctlRetreatAdjFort = 30 // Interest in retreating toward a region with a fort (per level)
ctlRetreatAdjDepot = 25 // Interest in retreating toward a region with a depot (per level)
ctlRetreatLandLink = 10 // Interest in retreating toward a region, value per land link
ctlRetreatPrevSubSpaceCoeffH = 250 // Coefficient applied to the interest if the region is the one where we are coming from
ctlNoBeachHead = 10 // Minimum control to have so that a region is not a beach/riverhead if you have to cross a body of water before attacking

[Pocus]

(I moved the thread)

[runyan99]

Okay. If you want to list the coefficients, we need to know the formula the game uses to choose the region. Otherwise we do not know how the coefficients are applied.

Also, is it possible to include a directional bias? For example, can you give a bias for the CSA to retreat south and the USA north?

[soloswolf]

Would it be possible to pre-plan your retreat in the same way you set up your movement path?

So you'd set up your move for the turn then click on the 'retreat move' button on the special orders tab and plot out your steps 'home'.

I know this would involve a large interface change, as well as possibly being impossible... Just a thought.

[MrT]

s possibly being impossible...

That sounds like one of those great noncommital terms like ''Definately maybe!''

[Gray_Lensman]

deleted

[Gray_Lensman]

deleted

[Generalisimo]

Now what would make a heck of a lot more sense is a predisposition to retreat back towards the retreating units' supply source, whether north or south in direction.

Maybe Pocus can add another coefficient for that?

[lodilefty]

Maybe if we boost the ctlRetreatPrevSubSpaceCoeffH we'll get closer to what we want....

Gives player some control [by manually tracing a path to battle, keeping lines intact, etc.] without a buch of complexity [eg: bugs to chase ] in the code.

[Gray_Lensman]

deleted

[runyan99]

Yes I suspect ctlRetreatPrevSubSpaceCoeffH is already so large it is making the other considerations irrelevant.

In the Charlottesville battle from my AAR with Manstein, Lee is in Albemarle, Marches to the guns for a battle in Charlottesville, returns to Albemarle after the battle, but then is forced to retreat from Albemarle. Where does he go? Back to Charlottesville, which is 'where he came from' despite the fact that this region by now was in enemy hands, rather than back toward his supply at Fredericksburg. This is an example of a non-sensical result that a real general would never make.

ctlRetreatPrevSubSpaceCoeffH should be decreased, while ctlRetreatAdjDepot ctlRetreatAdjCity should be increased.

Also, a new coefficient for rail would be helpful to add to the city and depot factors. All things considered, I almost always would prefer for my units to retreat along their rail lines.

Taken together, the city and depot and rail considerations should probably overwhelm the previous space coefficient. This helps to stress supply lines.

This still will not lead to best results. Using the Memphis example again, Confederate units will still want to retreat west across the Mississippi, where there is a rail and a port, rather than south towards their base of supply at Jackson.

Or take the important example of Nashville, TN. Confederate units forced to retreat will chose to go east, where there is a level 1 town, uncovering the rail lines to the south at Murfreesboro. (This is fixable if the rail coefficent is larger than the city coefficient).

Take the example of Chattanooga. A confederate retreat will probably go NORTHEAST to Decatur, where there is a town, rather than south along the rails towards Atlanta. I don't know how to fix this one because there is a rail in Decatur too, so the retreats will always go northeast here, unless there is some other factor introduced.

[tagwyn]

Rebel forces retreating from Memphis will always go to Corinth if Held by Confederate forces. t

[runyan99]

Thanks tag but by the time Memphis is besieged, there will be Union forces at Corinth or at least between Memphis and Corinth.

And no, they will not go there, because Corinth is two regions away and the retreat calculations only look at adjacent regions.

[mikee64]

Here is an odd one that just happened in my pbem. The black line is the move, the red is the retreat. Even looking at the retreat parameters I have no idea what caused this:



Jackson attacked from Winchester into Loudon. He then retreated across the river into Frederick MD which was 100% MC for the USA. This despite the fact that all the other options to retreat to in VA were 100% MC for the CSA, including the regions of Manassas and Hapers Ferry which included intact depots. All the adjacent regions in VA also included friendly but no enemy troops.

Any ideas on this one?

[Pocus]

Can you post your retreat parameters (opt file), perhaps there are some typos and they are not read properly. I remember vaguely there was a problem like that on retreat some months ago.

[mikee64]

This is what I have, I've attached the actual file also. Looks like what you posted above.

ctlContested = 5 // Minimum control gained upon entering a region (if not passive)
ctlAllowRetreat = 0 // Minimum control to have in a region to allow a retreat into it
ctlRetreatAdjCity = 5 // Interest in retreating toward a region with a city (per level)
ctlRetreatAdjFort = 30 // Interest in retreating toward a region with a fort (per level)
ctlRetreatAdjDepot = 25 // Interest in retreating toward a region with a depot (per level)
ctlRetreatLandLink = 10 // Interest in retreating toward a region, value per land link
ctlRetreatPrevSubSpaceCoeffH = 250 // Coefficient applied to the interest if the region is the one where we are coming from
ctlNoBeachHead = 10 // Minimum control to have so that a region is not a beach/riverhead if you have to cross a body of water before attacking

Let me know if you want me to check anything else or if you would like the saves.

[runyan99]

I can already see your AllowRetreat is different from mine. This is what I have for 1.13b

ctlContested = 5 // Minimum control gained upon entering a region (if not passive)
ctlAllowRetreat = 5 // Minimum control to have in a region to allow a retreat into it
ctlRetreatAdjCity = 5 // Interest in retreating toward a region with a city (per level)
ctlRetreatAdjFort = 30 // Interest in retreating toward a region with a fort (per level)
ctlRetreatAdjDepot = 25 // Interest in retreating toward a region with a depot (per level)
ctlRetreatLandLink = 10 // Interest in retreating toward a region, value per land link
ctlRetreatPrevSubSpaceCoeffH = 250 // Coefficient applied to the interest if the region is the one where we are coming from
ctlNoBeachHead = 10 // Minimum control to have so that a region is not a beach/riverhead if you have to cross a body of water before attacking

Also, I don't understand what the following means, is this a base value?:

ctlRetreatLandLink = 10 // Interest in retreating toward a region, value per land link

[mikee64]

You're right, I checked both my installs and sometime between 1.13b and betaRC17 this parameter

"ctlAllowRetreat = 5 // Minimum control to have in a region to allow a retreat into it"

changed from 5 to 0.

[runyan99]

There is another aspect of the retreats I'm not super happy with. That is how all retreats automatically force the retreating force into passive stance.

Consider the following battle.



In this case the CSA cavalry only is in offensive stance (which is why the CSA as a whole shows as offensive) while an infantry force under Lee is in defensive stance against Meade.

After an indecisive battle in which each side lost a few thousand, Lee was forced to retreat from Amherst. I suppose he was forced into passive stance, making him useless for the rest of the two week turn. This was to have consequences later in the turn for me in this game.

But sometimes two weeks is too long to be useless, and sometimes real armies were able to fight multiple engagements on the move over a course of two weeks. The game just will not allow that.

I don't see why after a middling battle like the Amherst one why the retiring force cannot remain in defensive stance. I big skirmish, like the battle in question, should not force one side or the other into impotence for a week or more. It would be better if passivity was forced only in the case of routs or heavy defeats, not all battles that end in withdrawl.

[runyan99]

There is another aspect of the retreats I'm not super happy with. That is how all retreats automatically force the retreating force into passive stance.

Consider the following first two battles.



In the first battle the CSA cavalry only is in offensive stance (which is why the CSA as a whole shows as offensive) while an infantry force under Lee is in defensive stance against Meade.

After an indecisive battle in which each side lost a few thousand, Lee was forced to retreat from Amherst. I suppose he was forced into passive stance, making him useless for the rest of the two week turn. This was to have consequences later in the turn for me in this game. In the second battle shown at Richmond, Lee's forces are counted as present, but are passive and do not participate in the battle, causing another loss although my side has more forces on hand.

Sometimes two weeks is too long to be useless after a battle, and sometimes real armies were able to fight multiple engagements on the move over a course of two weeks. The game just will not allow that.

I don't see why after a middling battle like the Amherst one why the retiring force cannot remain in defensive stance. I big skirmish, like the battle in question, should not force one side or the other into impotence for a week or more. It would be better if passivity was forced only in the case of routs or heavy defeats, not all battles that end in withdrawl.

[Pocus]

Please send over the save Mike, I will see if I can reproduce the same path.

ctlRetreatLandLink = 10 // Interest in retreating toward a region, value per land link
is a bonus of +10 interest points for each link the target region has. In effect this penalize dead ends and favor junctions.

[Gray_Lensman]

deleted

[mikee64]

I emailed you the saves Pocus.

Thanks for the info on that change Gray.

[bigus]

Please send over the save Mike, I will see if I can reproduce the same path.

ctlRetreatLandLink = 10 // Interest in retreating toward a region, value per land link
is a bonus of +10 interest points for each link the target region has. In effect this penalize dead ends and favor junctions.

Then we can put in 2500 or 25000?
Would this guarentee a force will retreat into an adjacent "region" of less movement points?
Or is this command for map edge "land links"?

[Generalisimo]

Then we can put in 2500 or 25000?
Would this guarentee a force will retreat into an adjacent "region" of less movement points?
Or is this command for map edge "land links"?

If I get that right, it will "force" the unit to retreat to the region that has "more land links"... so, it penalizes "dead ends" and favour regions with "more junctions".

[mikee64]

thankyou very much sir.
I would presume the weird retreat directions is related to runyuns post about his abstract retreat paths?

MrT, I moved this quote over from the beta patch thread so as not to bump up the discussion there. I am pretty sure the fix Pocus put in w/ regards to your question just fixed a specific problem that caused Jackson to retreat in the "wrong" direction as I posted above in this thread.

To be more specific: I had ordered an attack into Loudon but inadvertently dropped the path onto an enemy unit there. A battle occurred in which Jackson retreated, but he continued to follow the enemy after the retreat. He should have instead followed retreat parameters and gone back to one of the depots in VA.

I think for this patch Pocus addressed and fixed this one specific "bug" and not really any of the more "generic" things runyan was asking about.

[bigus]

If I get that right, it will "force" the unit to retreat to the region that has "more land links"... so, it penalizes "dead ends" and favour regions with "more junctions".

Thanks Generalisimo.

MC still seems to be the driving factor for retreat paths. I can have a depot with 68% mc and a region with 90% control and I bet they will retreat to the region with 90% control regardless of where this region is. Maybe just me but this has been the pattern in my tests so far.