Fixing the enemy units behind-the-lines problem

[Mark_Everson]

There is one big issue regarding demo eight that I think we need to discuss now. Numerous people have complained, and still do, about the "whack-a-mole" nature of having to chase down by hand small enemy units. These enemy units can be generated either by units of another civ sneaking by your main forces, or appearing due to rioting. I find this one of the more frustrating elements of the game as it exists now myself. If we can come up with a quick and dirty fix for this problem before the release of demo eight, I think it would be well worth it.

There are two somewhat separate issues that I want to elaborate upon here. As mentioned above there is the issue of enemy units in "home" territory that can take extraordinary efforts to track down because of the simultaneous movement system. I'll refer to that as the "unit" problem. In addition there is a " land" problem that can remain even after the enemy unit is dead. Where you have to move through every single previously-controlled square to re-establish control. Generally the unit problem is worse than the square problem, since you can always just arrange a movement path to go over previously-conquered squares to reclaim them.

One solution that occurs to me is a "patrol" order. A task force with this order would have a movement path entered, and be responsible for controlling an area within a certain number of squares or movement points from that path. For a linear movement point on a flat plan the TF would then attempt to control a rectangular piece of the map. If all the ground were under friendly control and there were no enemy units with in the area that the unit is responsible for, the unit would simply patrol back and forth along its movement path. If an enemy TF , either of another civ, or a rebel, were identified within the area, the patrolling TF would attempt to attack it. The patrolling TF would only consider the enemy TF a target if the patrolling TF were significantly large to take the enemy out. If the patrolling TF does not have an enemy TF that it considers a valid target but there are enemy-controlled squares within its area, then it would move to those squares to reestablish control.

If this sort of thing is practical, then I think it would solve most of the problems of player frustration from whack a mole. It would still be possible for the enemy TF to lead the patrolling TF on a bit of a chase. But if the chase didn't usually last too long it wouldn't be a huge deal. One other modification that might help would be having the unit movement rates reduced as supply reduces combat power. That way a rebel unit might expect to be moving slower once it is expelled from its home territory, and be easier to chase down. Certainly having more than one TF on patrol for overlapping areas would also make it much easier to chase down any interloper.

I would like to see some brainstorming of ideas for how to solve this issue at least crudely in a way that doesn't take too much effort. I think solving the unit problem is more important than solving the land problem, although potential fix is that would address both are of course best. Also, I'd like to hear if the idea on patrolling units above is something that can reasonably be implemented given current capabilities.

[alms66]

A few more options to consider:

1. Give units in home territory a +1 to movement and reduce units in enemy territory to a move of 1. This simulates the fact that units defending home territory know and can use the best paths to reach the frontline quickly and the fact that any invading army will slow to cover its flank and rear as it advances. This also prevents the enemy using your road network against you, until he's captured it.

2. Instead of just 'patrol' why not go one step farther and also do a 'patrol border' order, which automatically sets the movement path to patrol the border of the civ (or the province where the order was issued). In fact, since having the unit moving in circles on the map could be distracting, I'd like the unit to disappear entirely and go into a 'province defense' force, but that can wait until later, since it would also require building some sort of GUI to deal with this 'province defense' force.

3. If playtesting shows that the above two aren't enough, and only then would I consider this option, we could implement some sort of ZOC that only occurs for the defender in home territory.

As a long term solution, here's what I'd like to see:

No units on the map at all (most of the time) and thus no tedious moving of units every turn...
There is a pool of units (which the player builds normally) for each civ, all new units enter this pool. This pool automatically defends the civ's borders (as in #2 above). If the player decides to go to war, he brings up his military advisor and says, "I need 30% of my forces to attack China (or whoever)." He can fine-tune these orders by saying "45% of artillery and only 30% of the rest" or adding an additional 50 tanks and 100 fighters, for example. Once the order has been given, a single task force appears on the map. The player then sets about giving it orders such as attack the southern province, capture the city of Hong Kong, don't loose more than 50% casualties, etc. Then the TF automatically sets out to execute those orders. When the TF reaches China, it automatically breaks into any number of necessary divisions and executes the plan issued by the player.

You'll notice I said 'plan.' That's intentional, as the player would basically have an interface to the AI planning module, and would build a plan for it. The AI could then use this plan to execute the orders given, and learn from it as well.

-edit-
Of course, those masochists among us who enjoy moving 100 units per turn could opt to take control of the invasion rather than designing a plan.

[LDiCesare]

Let me suggest entirely different solution to the unit problem:
-Riots
-Fleeing from fights
-Militia

Riots generate many units. Probably too many. I can reduce the number of units generated. It won't really suppress the problem, but if there are less units, they are likely to die faster. Not solving the problem, but tangential.

Fleeing from fights:
Most of the time, small remaining units come from a routed army which fled away. They sometimes manage to flee back to a square which you used to control. This makes for chases. If units can flee when routed only if the target square (which they come from) is controlled by them (or noone or an ally), they would stay in place more and get reduced to pieces without requiring as much chasing.

Militia:
I don't like the Patrol proposal because I think it would waste lots of resources and not be easy to code. The border patrol with abstraction looks better to me, but I think it shouldn't even require the player creating units and dedicating them. In fights, a militia can be generated in order to help unis destroy enemy units. This could be expanded so that militia would be able to fight even if there are no friendly units. Militias wouldn't hamper big armies, but they could get rid of smaller ones. That would mean that your well populated squares would defend themselves alone from small armies, while squares with small populations might still be conquered by small forces.

[Mark_Everson]

I'm willing to go with your solutions Laurent, and see if they solve the problem enough. Changing the movement rates a bit depending on control of the territory moved into, as alms suggests, is also a possibility. However, that could cause problems with path finding etc., so it's probably best left as a backup possibility.

In another thread I suggested that movement rate be changed due to supply level. Is that another "quick fix" that might be applicable here? I have to admit I don't even know where rebels get their supplies from, or if they even need them in the current code.

Alms, the abstract stuff you suggested the end of your post, in terms of just being able to give the AI an abstract plan for it to implement , is still in the long-term spec as far as I know. But that kind of thing is a long way off, so I don't think we should divert from current activities to discuss it at this point. I don't think there are many people that want to get rid of units and task forces on the map completely.

[LDiCesare]

In another thread I suggested that movement rate be changed due to supply level. Is that another "quick fix" that might be applicable here? I have to admit I don't even know where rebels get their supplies from, or if they even need them in the current code.

I must confess I don't know either... I suspect they are not checked at all.
Answering a related point raised by alms, the riots model doesn't say that rebel units would stop being if the cause of the rebellion stops. They are supposed to try to seize power instead. Check the social model and we can discuss it in the Social Model thread.

[LDiCesare]

I coded a few things and aml not satisfied. Having militia intervene even when there are no enemy units doesn't work very well because it can't work in territory where the locals prefer your enemy to you. e.g. Persians in the social scenario won't help to fight against Persians. I'm not committing my changes because they don't seem to change much. I'll do so next week-end if a week of testing doesn't reveal bugs.
I still would like to have something abstract to help get rid of weak stray units. I'll try to see if I can tweak the existing code for militia creation to get it more effective or if it needs something else. A 'patrol' spending entry in the economy could also be used to get rid of the problem and would be easier to track as patrollers and militias are two different things.

[alms66]

Originally posted by LDiCesare
Militia:
I don't like the Patrol proposal because I think it would waste lots of resources and not be easy to code. The border patrol with abstraction looks better to me, but I think it shouldn't even require the player creating units and dedicating them.

I don't see why you think the Patrol order would waste resources, wouldn't you simply reissue the same movement queue that existed before? This seems to be about the same intensity as the current 'move' order, or a slight bit more.

As for the border patrol (abstracted form), I envisioned that all units the player creates would be used for border patrol. Thus the player does not need to intervene. He builds units for the sake of having the strongest force, not specifically for border patrol. The player would have to pull out the troops he needed when (and only when) he decided to go outside his own borders, though.

[LDiCesare]

How do I know where to patrol? Do you mean you issue a path and let it go around that path indefinitely? If so, yes it's cheap on resources. But then you have to fix it everytime the border changes, and it doesn't prevent an enemy from entering the zone you control when the patrol is at the other end of its path.
If all your units inside your territory participate to border patrol, then it's going to be very hard to invade anything.
Already, the code I put to provide automatic upheaval of militia to help defenders (when they are sufficiently numerous) gives a BIG boost to defense. Anything more would be unbalanced. I'm not even sure I'm going to deliver the code I did.
More or less: When an army enters a square, I systematically check ethnic groups to see if they should raise a militia. They usually don't, unless you have an army that needs help, or the incoming army is really small (the problem of small is very dependant on the size of units: In Delenda, an army of 5000 men is quite small, but is still big compared to the population of most squares). The result is it's far easier to defend a square when the opponent is more or less of the same strength as you.
The difference with last week-end is that militia will try to help in that order, those who oppress them the least, then the civ whose prefered nationality they are, then the defender civ. Thus Persians won't help macedonian defenders against persians, but Gauls will help either Carthaginians or Romans, depending on ethnic discrimination and who owns the square. It's nice, but I don't think it solves the problem.
There's still one other thing I think could e done: Kill units that are in really bad shape. Currently, a 5000-man unit reduced to 200 will keep fighting other battles, which is probably not a good idea. They should disband or surrender after reaching a minimum number, which would get rid of the smallest units which take forever to kill, and thus remove one source of the problem instead of trying to find solutions to fix it afterwards.

[alms66]

Originally posted by LDiCesare
How do I know where to patrol? Do you mean you issue a path and let it go around that path indefinitely? If so, yes it's cheap on resources. But then you have to fix it everytime the border changes, and it doesn't prevent an enemy from entering the zone you control when the patrol is at the other end of its path.

As an early implementation, I’d put it as a player-only feature, so yes, it follows the path the player sets, and only that. I imagine we’d refine the process later on.

Originally posted by LDiCesare
If all your units inside your territory participate to border patrol, then it's going to be very hard to invade anything.

Not at all. You simply withdraw the troops you want, at any time you want them, as though the “border patrol” were a bank account and troops were currency.

Originally posted by LDiCesare
There's still one other thing I think could e done: Kill units that are in really bad shape. Currently, a 5000-man unit reduced to 200 will keep fighting other battles, which is probably not a good idea. They should disband or surrender after reaching a minimum number, which would get rid of the smallest units which take forever to kill, and thus remove one source of the problem instead of trying to find solutions to fix it afterwards.

Rather than ‘kill’ the unit, and thus lose the population, I’d prefer to have them settle (depending on the circumstances) in the square where combat occurred. This serves two purposes:
1. It’s fairly historically correct to assume that battles far from home (we’ll need a definition for that) would settle in the local area after major loses (again needs a concrete definition).
2. It presents an interesting possibility for the player. Suppose that 200 people settled, grew and eventually became a major presence in the empire. Once the religious assimilation code is in, it could be more interesting because those 200 could spread religion as well.

[Mark_Everson]

Originally posted by LDiCesare
Already, the code I put to provide automatic upheaval of militia to help defenders (when they are sufficiently numerous) gives a BIG boost to defense. Anything more would be unbalanced. I'm not even sure I'm going to deliver the code I did.

What you did sounds interesting to me. Would it be possible to make it controlled by a command switch? If it's easy enough it would be good to have a command-line switch that also allowed the level of support to be varied. A level of zero could mean no militia support, in the level of one hundred could be what you have now. Then we could play around with the parameters and see what levels seemed appropriate.

click cancel's still one other thing I think could e done: Kill units that are in really bad shape. Currently, a 5000-man unit reduced to 200 will keep fighting other battles, which is probably not a good idea. They should disband or surrender after reaching a minimum number, which would get rid of the smallest units which take forever to kill, and thus remove one source of the problem instead of trying to find solutions to fix it afterwards.

After my most recent set of play testing last weekend I had the following thoughts:

1. Eliminate all units that fall to below 10% of starting strength. (If a similar unit exists in that square that could take the reinforcements that it should get them.)

2. Any unit that is in a fight at worse than 10:1 odds should be annihilated. (There may be some exceptions, for example if the weaker force has higher mobility it should have some chance of getting away.)

Number 2 above points out what I think is a big flaw in the combat system. It restricts the number of units that can fight a given unit to a very small number. (I think it is either one or two.) In the real world there are many ways were the excess troops can make a huge amount of difference. I think that at a minimum the combat matching should allow greater odds. Perhaps such a change is beyond the scope for demo 8, so I will just copy this part of my post over to the military thread.

[LDiCesare]

A command switch can be done, but figures wouldn't help much, as it's already very hard and somewhat subjective to judge the effect.
Eliminating too weak units I will do. There's already a check like that, but it's based on starting health in the turn, and not unit maximum health. I'll do it. The proportion will not be a straight 10% but something along those lines based on morale because I think that's what morale stands for (if remains less than 1/morale, knowing morale is about 10 for an average troop).

[LDiCesare]

I think the problem should be fixed with the last code update I sent. If that doesn't seem to fix it, I suggest lowering the morale of elements. See the military model VI thread for details.

[Mark_Everson]

The recent changes seem to have helped significantly! I think it's good enough for the demo at this point (although I need to playtest a bit more). I will do that while I make some scenario changes to fix typos etc.

Thanks Laurent.

[demipomme]

I think that the 'land' problem is more problematic than the 'unit' problem. The only reason why I want to track down and kill the stray enemy units is because if I don't, they conquer my entire province except by well-defended city. If it were the case, and I have already suggested it, that control of a square reverted to its original owner unless held for several turns, then a single enemy unit wouldn't be able to take large swathes of my territory. The second advantage is that the enemy unit would have to stay still to gain anything and thus I can track it down while its stays still.

I think this solution is realistic as it models for the change of local government & police etc. of the square. The same can be applied for province capitals i.e. you get to keep the entire province if you hold the capital for several turns.

Perhaps the number of turns needed to 'conquer' a square could be related to its chance of rioting (as an indicator of dicontent with current regime).

I'm also down with the militia idea.

[alms66]

The idea in and of itself is sound. A unit should have to move into a tile and stay there for that turn, at least, to change ownership. Having more than 2 turns would probably not be worthwhile though. It would bog wars down as players have to leave individual units behind the lines to "occupy" the tiles in order to gain control, or worse if we require the occupying force be larger depending on the population and resistance of the tile. Then, they'd have to move them once ownership changed, it would be a hassle.

I'd rather have the automated "provincial defense" pool we discussed above and elsewhere (I think) than to be forced to leave those units behind. Like I said though, holding a tile for a 1 turn minimum would be good on the changing of ownership.

This is also still a good option to consider, IMO:

Originally posted by alms66
1. Give units in home territory a +1 to movement and reduce units in enemy territory to a move of 1. This simulates the fact that units defending home territory know and can use the best paths to reach the frontline quickly and the fact that any invading army will slow to cover its flank and rear as it advances. This also prevents the enemy using your road network against you, until he's captured it.

[Mark_Everson]

Originally posted by demipomme
I think that the 'land' problem is more problematic than the 'unit' problem. (Snip) If it were the case, and I have already suggested it, that control of a square reverted to its original owner unless held for several turns, then a single enemy unit wouldn't be able to take large swathes of my territory. The second advantage is that the enemy unit would have to stay still to gain anything and thus I can track it down while its stays still.

I think this solution is realistic as it models for the change of local government & police etc. of the square. The same can be applied for province capitals i.e. you get to keep the entire province if you hold the capital for several turns.

Hi demi pomme:

I agree that we still have problems in this area that need to be fixed. However, I don't think you're idea is very practical since it requires an awesome amount of micromanagement. Every time I move a unit I would need to decide if I want to have it sit for the requisite number of turns on each of the squares in its path. I think your idea would require significant AI tweaking also.

My best guess for what we should do now to fix the "land" part of the problem is to have squares revert to the original civ quite easily if certain criteria are fulfilled. A square would be likely to revert to the original civilization if all these conditions are fulfilled:
1. it had significant affinity for the original civilization (for example same ethnic group), and to a lesser extent had minimal affinity for the conquering civ.
2. Units of the conquering civ are not too close by (say one turn movement)
3. The military power balance nearby does not favor the conquering civilization (nearby means say within a five square radius, or three turns movement, or something like that).
4. There would be a bonus to reversion if there is at least one square of the original civ ownership adjacent to the conquered square.

What the rules above would do is that a few individual units loose inside your civilization would still be an irritation, but could be safely ignored from a "land" standpoint. Just to be clear, I don't mean for any of these criteria above to be completely rigid, for example the "not too close by" criterion in #2 above would not go from "yes" to "no" when a unit was exactly 1 turn of movement away. But rather then when the unit is one movement turn or closer this criterion would be fulfilled at a high level, where as beyond 1 turn of movement the level of satisfaction for this criterion would rapidly diminish toward zero. The reversion chance would depend on something like the psalm all the levels of satisfaction of the first three criteria.

I'd also like to add that in demo 8.1 we should have the "move to enemy TF" command active, which will make hunting down enemy units much easier. Laurent, please correct me if I am wrong. I think this may even be implemented, I have to apologize but I haven't been able to check code updates for the last few weeks.

Alms, we of course always have your idea for a generalized provincial defense force as a backup if our other approaches don't work. However, I think my proposal covers several real-world situations better. For example when Sherman marched to the sea in his southern campaign during the American Civil War his army lived off the land and controlled the local territory wherever it went. That army was too big for the defending Southern armies to directly attack, or a least defeat. However, Sherman was of course not able to retain control of all the territory that he went through because as soon as his army had moved beyond a certain range the locals immediately reverted back to their desired civilization, the CSA. I think my approach handles this correctly, where as the generalized provincial defense idea, at least as originally proposed, would leave Sherman in control of that large string of territory since the defense forces wouldn't be able to eliminate him.

What do you guys think of my idea both in terms of at least temporarily fixing problems in the game, and also in terms of difficulty to code and clock cycle usage?

[LDiCesare]

3. The military power balance nearby does not favor the conquering civilization (nearby means say within a five square radius, or three turns movement, or something like that).

I like this one. I'd like units to "project" their power around them and based on this projected power, a square would know whether or not it's safe to revert or outright rebel.

[alms66]

Originally posted by Mark_Everson
Alms, we of course always have your idea for a generalized provincial defense force as a backup if our other approaches don't work. However, I think my proposal covers several real-world situations better. For example when Sherman marched to the sea in his southern campaign during the American Civil War his army lived off the land and controlled the local territory wherever it went. That army was too big for the defending Southern armies to directly attack, or a least defeat. However, Sherman was of course not able to retain control of all the territory that he went through because as soon as his army had moved beyond a certain range the locals immediately reverted back to their desired civilization, the CSA. I think my approach handles this correctly, where as the generalized provincial defense idea, at least as originally proposed, would leave Sherman in control of that large string of territory since the defense forces wouldn't be able to eliminate him.

I don't want to get into that discussion again. I'm tired of explaining stuff. The approaches taken thus far have done minimal changes to the problem, so continue down that path and who knows, maybe sometime next year the problem will be solved.

[Lord God Jinnai]

My best guess for what we should do now to fix the "land" part of the problem is to have squares revert to the original civ quite easily if certain criteria are fulfilled. A square would be likely to revert to the original civilization if all these conditions are fulfilled:
1. it had significant affinity for the original civilization (for example same ethnic group), and to a lesser extent had minimal affinity for the conquering civ.
2. Units of the conquering civ are not too close by (say one turn movement)
3. The military power balance nearby does not favor the conquering civilization (nearby means say within a five square radius, or three turns movement, or something like that).
4. There would be a bonus to reversion if there is at least one square of the original civ ownership adjacent to the conquered square.

It would be nice if say the unit disbanded and the population integrated into their from the army would affect then the affinity for the civilization if they were too far away from home.

The bonus for #4 should be cumulative though, but overall should be neglible so a province that really hated the original civ won't revet 99% of the time.

Also another thing to condier, is rather than staying under the current rule, if they decide not to revert, is to simply become an autonomous region.

[Mark_Everson]

Originally posted by Lord God Jinnai
It would be nice if say the unit disbanded and the population integrated into their from the army would affect then the affinity for the civilization if they were too far away from home.

Sorry, not sure exactly what you mean here.

The bonus for #4 should be cumulative though, but overall should be neglible so a province that really hated the original civ won't revet 99% of the time.

Agree. And yes, the bonus was more meant to boost squares that would like to revert anyway.

Also another thing to condier, is rather than staying under the current rule, if they decide not to revert, is to simply become an autonomous region.

Good point.

[demipomme]

Originally posted by Mark_Everson


However, I don't think you're idea is very practical since it requires an awesome amount of micromanagement. Every time I move a unit I would need to decide if I want to have it sit for the requisite number of turns on each of the squares in its path. I think your idea would require significant AI tweaking also.

Micromanagement is good grounds for an objection, AI tweaking is not (unless you're trying to make a rush bodge game like all the other civ clones which you obviously aren't cos you're not finished yet, not that I'm complaining, I just think its worth getting it right).

Micromanagment is only a problem if you want to take a province square by square. If you want the whole province, take the capital and hold that for x turns. Micromanagement = v. low. I agree with alms comment that holding for 1 turn is long enough.

You suggested reversion, however my suggestion is not about reversion but conquering. In the scenario Attila, where I have had most problem with this, an enemy horsemen rides up and down my road covering I think 6 squares and "conquers" all of them. Realism = v.low.

A compromise suggestion is: only keeping the square you end your turn in? This way mobile units wouldn't conquer the entire world instantaneously and infantry units would move and conquer normally.

[demipomme]

Originally posted by Lord God Jinnai
It would be nice if say the unit disbanded and the population integrated into their from the army would affect then the affinity for the civilization if they were too far away from home.

I guess this means that if an enemy unit dispand in your square, its ethnic group joins your population and increases the squares affinity to the enemy civilisation.

One point I missed from the above post is that I don't disagree with reversion. I think undefended enemy squares should revert if the conditions are favourable.

[Mark_Everson]

Originally posted by demipomme
Micromanagment is only a problem if you want to take a province square by square. If you want the whole province, take the capital and hold that for x turns. Micromanagement = v. low. I agree with alms comment that holding for 1 turn is long enough.

Sometimes you need to take ground square by square. FE If the provincial capital has walls and you don't have local resources to besiege.

You suggested reversion, however my suggestion is not about reversion but conquering. In the scenario Attila, where I have had most problem with this, an enemy horsemen rides up and down my road covering I think 6 squares and "conquers" all of them. Realism = v.low.

We were talking about solving the "land" problem. Both proposals are potential solutions to that issue. Your example though is certainly compelling in terms of illustrating the problem.

A compromise suggestion is: only keeping the square you end your turn in? This way mobile units wouldn't conquer the entire world instantaneously and infantry units would move and conquer normally.

Perhaps we could have two movement modes, pure movement and conquest movement. There would be a TBD tick penalty for the conquest part but you would gain control of the squares traversed. That would at least be less MM than what I'd been thinking of. The movement mode could just be a switch that you set in orders. Lets see what Laurent and others think, you've convinced at least me that it'd be worth trying.

[Lord God Jinnai]

Originally posted by demipomme
I guess this means that if an enemy unit dispand in your square, its ethnic group joins your population and increases the squares affinity to the enemy civilisation.

Yes, as long as their civ isn't close by, which would be relative based on technology, but with some maximum even for modern day.

And i wasn't saying 100% no matter where, but if you disband a unit, without proper technology breakthoughs, even in your own civ, anywhere far from where they originated, say more than two squares minimum, then a certain percentage would stay there. This would increase signifigantly if it was in another civ.

However, reguardless of the number that stayed or left, such movements should be represented in the population by the next turn so we don't haveto continue to keep track of such small number of people.

Thus essentially they'd settle wherever they could reasonably move to in 1 turn, primarly for their own civ (their own square if possible), secondarily, an area that liked them, next to an indifferent populated area, then to settling down where they were. Some might move to the neigbring squares FE, but the majority would stay put, or go home if possible.

[Lord God Jinnai]

Originally posted by Mark_Everson Sometimes you need to take ground square by square. FE If the provincial capital has walls and you don't have local resources to besiege.

Also if the rural population resists quite strongly

[demipomme]

Originally posted by Mark_Everson


Sometimes you need to take ground square by square. FE If the provincial capital has walls and you don't have local resources to besiege.

Hadn't thought of that. Good point.

Originally posted by Mark_Everson

Perhaps we could have two movement modes, pure movement and conquest movement. There would be a TBD tick penalty for the conquest part but you would gain control of the squares traversed. That would at least be less MM than what I'd been thinking of. The movement mode could just be a switch that you set in orders. Lets see what Laurent and others think, you've convinced at least me that it'd be worth trying.

Like the pure/conquest idea. But it's got me thinking...

-Every square should have a certain number of law enforcement personnel
-Number of law enforcement personnel related to amount invested in law enforcement by owning government
-Invading units kill or disband existing law enforcement – some loss of personnel to invading army (1st reduction of stray unit power)
-In pure movement mode, invading unit moves to next square leaving previous square in hands of original civ, but in economic turmoil as no law enforcement
-In conquest movement mode, invading unit moves to next square leaving behind some personnel in previous square (2nd reduction of stray unit power)

Also:

-for realism, pure and conquest should be the same speed
-law enforcement also reduce riots
-border police – can select to have higher law enforcement on border

Advantages:

-realism
-simple choice between pure and conquest
-stray enemy units diminished in size by moving through many squares
-stray enemy units diminished further in size if they conquer squares

[LDiCesare]

I said I liked this:

3. The military power balance nearby does not favor the conquering civilization (nearby means say within a five square radius, or three turns movement, or something like that).

Here's what I propose, and an illustration of it.
If armies of owning civ are "stronger" in a given square than the invaders, then that square xon't change hands. The "stronger' means there is a strong attack force nearby. It's a kind of power projection. The image shows in red squares where the Carthaginian military influence is stronger, in green where the Romans are stringer. Several roman squares have been conquered by a stray carthaginian archer. The proposal is to prevent the squares from changing hands if they are in the green area of influence. The power projection I used in this screenshot is (total attack of army)/(time in ticks to reach square + 10). This means an influence of 1 where you are, 0.5 where you'd be in one turn, 0.33 in 2 turns. I limited the computation to 2 turns of movement but could do more.
As for reversion, I'd like it to be handled by the social model. The nationalistic feeling is not coded, but it could probably handle that.

[LDiCesare]

I forgot to attach the image.

[alms66]

I really like the way the hills look gray in the bottom right of that pic, Laurent, that's what I wanted all tiles to look like whenever you turned on one of the filters, except those that are colored by whatever filter you are looking at.

-Edit-
Also, I think it would look better in the long run if only the terrain tile & terrain overlays were colored (not roads, cities, units, etc.)

[Mark_Everson]

I like your implementation idea Laurent, sounds like it will help a fair amount, even if it doesn't completely solve the problem.