Aggregate Mod

[Schweetness101]

The only thing I am still not sure about is the automatic cohesion drop for flank attacks from cavalry units. I think it might be better to have +100POA instead (which might, or might not, lead to a cohesion drop).

against other cav?

also, what about swapping it so cav no longer can flank one another with full effect like you are suggesting, but they can flank infantry with full effect? Like the psychological effect of being flanked by cavalry as infantry is greater than if you are yourself cavalry? Because of the relative fear delta?

[stockwellpete]

The only thing I am still not sure about is the automatic cohesion drop for flank attacks from cavalry units. I think it might be better to have +100POA instead (which might, or might not, lead to a cohesion drop).

against other cav?

also, what about swapping it so cav no longer can flank one another with full effect like you are suggesting, but they can flank infantry with full effect? Like the psychological effect of being flanked by cavalry as infantry is greater than if you are yourself cavalry? Because of the relative fear delta?

That will create another anomaly for me. I think the +100POA against both infantry and cavalry is consistent and I think we need to test it over a longer period. The cavalry +100POA against infantry for flank will also cause cohesion drops on occasions. At the moment we do not know the % chance of it happening. Again we need some stats. I will set up some of these situations in my Training Ground over the next few days and do 100 of each to give us some idea. We are a bit in the dark at the moment.

[desicat]

8 ) reduced all general command radii to 4
9) removed positive effects of general unless it is the CinC or an SG and you are on the SG's team, ie SG are now like allied generals for the units under their command

-----------------------------------------
Those two changes will make it hard for the AI in SP Scenarios and Campaigns.

I confess I do not play those very much. I have not changed AI behavior in this mod, and the AI is of course written with vanilla rules in mind. For example, the AI is still going to try to line up flanks like they will cause an auto drop, when they no longer will in this mod. It may be worth making AI edits once this mod is close to complete, but for right now that is too ambitious to do alongside the rule changes, and would be inefficient if I was updating the AI with every rule change. Better to do that all at once at the end.

Specifically, why will the struggle with changes to command radii and rallying inside of the general's radius only? are there scenarios setup with units inside of an 8 tile radius that will now be outside of a 4 tile radius for example?

One more try to answer, I have tried twice with no success.

The player has the advantage of assigning units to their Generals in the setup phase prior to battle while the AI is limited to the "Team" concept. Orders are issued to the Teams via scripts, and sometimes their are more Teams (8 possible) executing certain maneuvers or restricted by defensive orders than Generals available.

The Rally rule would be even more restrictive as the player will hopefully be way more nimble in positioning and possibly reassigning their Generals to other units. Technically the AI can move its Generals too, but I have only seen that happen on a single occasion.

[Schweetness101]

8 ) reduced all general command radii to 4
9) removed positive effects of general unless it is the CinC or an SG and you are on the SG's team, ie SG are now like allied generals for the units under their command

-----------------------------------------
Those two changes will make it hard for the AI in SP Scenarios and Campaigns.

I confess I do not play those very much. I have not changed AI behavior in this mod, and the AI is of course written with vanilla rules in mind. For example, the AI is still going to try to line up flanks like they will cause an auto drop, when they no longer will in this mod. It may be worth making AI edits once this mod is close to complete, but for right now that is too ambitious to do alongside the rule changes, and would be inefficient if I was updating the AI with every rule change. Better to do that all at once at the end.

Specifically, why will the struggle with changes to command radii and rallying inside of the general's radius only? are there scenarios setup with units inside of an 8 tile radius that will now be outside of a 4 tile radius for example?

One more try to answer, I have tried twice with no success.

The player has the advantage of assigning units to their Generals in the setup phase prior to battle while the AI is limited to the "Team" concept. Orders are issued to the Teams via scripts, and sometimes their are more Teams (8 possible) executing certain maneuvers or restricted by defensive orders than Generals available.

The Rally rule would be even more restrictive as the player will hopefully be way more nimble in positioning and possibly reassigning their Generals to other units. Technically the AI can move its Generals too, but I have only seen that happen on a single occasion.

ok, definitely something to keep in mind. Maybe you could help out and do a few tests on how this mod is harming the ai specifically and once we've got a list of ai changes to make to rectify what the mod breaks we can look into fixing it

[Schweetness101]

there is a function called:

FUNCTION CheckPursuersSwitchTarget(me, type)

with a comment like:

// Test for potential charge target within 45 degrees of streight ahead. Returns -1 if none that they want to charge, unit id if they do.
// type == 1: Pursuing routers, type == 2: Pursuing evaders.

within which there is a section like:

Code: Select all

// Ignore flank threat if pursuing routers, but not if pursuing evaders. Note that we might want to change this, so that pursuers don't ignore flank threat when pursuing routers.
		ignoreFlankThreat = 0;
		if (type == 1)
			{
				ignoreFlankThreat = 1;
			}

and I think perhaps we can go some way toward's fixing the issue with pursuers of routers being too slippery and too unlikely to charge a new unit and too likely to storm through lots of ZoCs by simply turning this one exception off. IE, by making pursuers of routers consider flank threats in the way pursuers of evaders do. The comment that comes with the code even suggests as much.

edit: HOWEVER, this might be doing opposite of what I hope though and by taking into account flank threat the unit is less likely to charge

One way to pose this question to pete is: do you notice that pursuers of evaders have the same issues you have brought up as pursuers of routers

[stockwellpete]

One way to pose this question to pete is: do you notice that pursuers of evaders have the same issues you have brought up as pursuers of routers

Don't know. I will have to watch what happens more carefully, but it looks like there is a possible fix there.

[stockwellpete]

there is a function called:

FUNCTION CheckPursuersSwitchTarget(me, type)

with a comment like:

// Test for potential charge target within 45 degrees of streight ahead. Returns -1 if none that they want to charge, unit id if they do.
// type == 1: Pursuing routers, type == 2: Pursuing evaders.

within which there is a section like:

Code: Select all

// Ignore flank threat if pursuing routers, but not if pursuing evaders. Note that we might want to change this, so that pursuers don't ignore flank threat when pursuing routers.
		ignoreFlankThreat = 0;
		if (type == 1)
			{
				ignoreFlankThreat = 1;
			}

and I think perhaps we can go some way toward's fixing the issue with pursuers of routers being too slippery and too unlikely to charge a new unit and too likely to storm through lots of ZoCs by simply turning this one exception off. IE, by making pursuers of routers consider flank threats in the way pursuers of evaders do. The comment that comes with the code even suggests as much.

One way to pose this question to pete is: do you notice that pursuers of evaders have the same issues you have brought up as pursuers of routers

Yes, that looks like a possible fix, doesn't it? Resume tomorrow. Knackered now.

[Schweetness101]

One way to pose this question to pete is: do you notice that pursuers of evaders have the same issues you have brought up as pursuers of routers

Don't know. I will have to watch what happens more carefully, but it looks like there is a possible fix there.

as per edit after HOWEVER in above comment I may have got this reversed, but either way this is the spot to change. I am confusing myself with the double negatives ie if it is not the case that it does not take into account flank threats...anyway i think i can figure it out. I can change that and then also simply increase the odds to pick a new unit to charge if pursuing, routers or evaders, by altering the number 50 in this line:

if (AI_ChanceOfCharging(me, id, ignoreFlankThreat) >= 50)

ie if that 50 is lower then there is a higher chance to charge a new non routed unit while pursuing

[rbodleyscott]

The Rally rule would be even more restrictive as the player will hopefully be way more nimble in positioning and possibly reassigning their Generals to other units. Technically the AI can move its Generals too, but I have only seen that happen on a single occasion.

It only does it if it thinks the unit it is with will soon break because it is Fragmented or close to Autobreak. Because the general can't leave when in close combat, this is mainly when player has concentrated shooting on the general's unit.

It is nothing cleverer than that I am afraid.

[rbodleyscott]

One way to pose this question to pete is: do you notice that pursuers of evaders have the same issues you have brought up as pursuers of routers

Don't know. I will have to watch what happens more carefully, but it looks like there is a possible fix there.

as per edit after HOWEVER in above comment I may have got this reversed, but either way this is the spot to change. I am confusing myself with the double negatives ie if it is not the case that it does not take into account flank threats...anyway i think i can figure it out. I can change that and then also simply increase the odds to pick a new unit to charge if pursuing, routers or evaders, by altering the number 50 in this line:

if (AI_ChanceOfCharging(me, id, ignoreFlankThreat) >= 50)

ie if that 50 is lower then there is a higher chance to charge a new non routed unit while pursuing

The results returned by AI_ChanceOfCharging() are not a continuum though. I doubt, for example, if reducing the test value to 30 would make very much difference. Generally speaking the returned values are either very high (75% upwards) or very low (e.g. 18% downwards).

If you want to see this in game, AI_ChanceOfCharging() does write its return value to the Debug Log, which you can see in game by pressing F6 when your /Documents/My Games/FieldOfGlory2/User.txt file contains the line

DEBUGMODE 1

(Takes effect next time you start the game)

Since it probably wouldn't be a good idea to mod the AI's chance of charging in normal circumstances, the best solution would probably be to write a variant version of the function (obviously with a different name) and use that for pursuit charges.

[stockwellpete]

I think the smallest radii that would probably work on all maps would be 5 squares, but that is a very big area indeed on small or very small maps and it would affect rallying and anarchy too. Tomorrow I will playtest a 4 square command radius on a very large map to see what happens. If there is going to be problem it will be there, on the smaller maps everything will be fine.

I did this playtest today using the 4 square command radius on the largest map size and I thought it was fine. I used a Carthaginian 340BC army against the Greeks 460BC and was able to deploy my infantry in 2 lines (African spears with one sub-general on my left of centre and Warband/Scutari with another sub-general on my right of centre. The second line of infantry was mercenary hoplites and my cavalry and chariots were deployed on my left flank. I also had a 2 unit mounted reserve with my C-in-C. The Greeks attacked in a broader line than mine as I had deployed in depth and after a fairly even skirmish battle the two infantry lines met. My African spears and Warband/Scutari gradually got on top of the enemy hoplites forcing them to commit some of their second line of raw hoplites to the fray. Out wide on the left my cavalry and chariots beat up the few Greek cavalry units there and began to sweep round behind the enemy centre. Then the enemy C-in-C was killed causing a lot of cohesion loss in the centre and this enabled me to overwhelm the infantry there as my cavalry started to smash into his rear. My C-in-C was never required to move and all 3 of my sub-generals survived the battle having provided command to their contingents throughout (apart from when they were melee-ing themselves.

I think it would obviously been a bit different if any of my sub-commanders had been killed, but I think the trick with smaller command radii is to not commit them to melee unless it is really necessary.

[Schweetness101]

Don't know. I will have to watch what happens more carefully, but it looks like there is a possible fix there.

as per edit after HOWEVER in above comment I may have got this reversed, but either way this is the spot to change. I am confusing myself with the double negatives ie if it is not the case that it does not take into account flank threats...anyway i think i can figure it out. I can change that and then also simply increase the odds to pick a new unit to charge if pursuing, routers or evaders, by altering the number 50 in this line:

if (AI_ChanceOfCharging(me, id, ignoreFlankThreat) >= 50)

ie if that 50 is lower then there is a higher chance to charge a new non routed unit while pursuing

The results returned by AI_ChanceOfCharging() are not a continuum though. I doubt, for example, if reducing the test value to 30 would make very much difference. Generally speaking the returned values are either very high (75% upwards) or very low (e.g. 18% downwards).

If you want to see this in game, AI_ChanceOfCharging() does write its return value to the Debug Log, which you can see in game by pressing F6 when your /Documents/My Games/FieldOfGlory2/User.txt file contains the line

DEBUGMODE 1

(Takes effect next time you start the game)

Since it probably wouldn't be a good idea to mod the AI's chance of charging in normal circumstances, the best solution would probably be to write a variant version of the function (obviously with a different name) and use that for pursuit charges.

thanks! that is excellent advice. I noticed it wasn't a continuum, but I did not spend the time to dive into all the values. I just saw things like chance to charge going up or down based on relative combat modifiers, shooting distances, etc...of which there were a number of different values in different conditions, so it is very helpful to have those high and low approximations. Do you think 10% would be too low to set it to start testing? And ought I to just try that and not bother with changing the ignorethreatenedflank parameter?

I am not entirely clear on the which function you are suggesting to write a variant of. Are you suggesting I make a new AI_ChanceOfCharging() just for pursuits, or a new CheckPursuersSwitchTarget() just for the human player? Something else? If I just adjust the value in CheckPursuersSwitchTarget() on the line like:

if (AI_ChanceOfCharging(me, id, ignoreFlankThreat) >= X) //where X is 50 by default

would that not edit the odds of changing pursuit target to charge target while pursuing, and only for pursuing, for both AI and human players? Or are you saying that would be insufficient to fix the issue under discussion, and I should instead look to write a somewhat different Pursuit_ChanceOfCharging() method that is only called by CheckPursuersSwitchTarget()?

[Schweetness101]

...but I think the trick with smaller command radii is to not commit them to melee unless it is really necessary.

I think this is a key and very interesting difference with the new command changes (both to radii and rallies being dependent on being within that radii and the general not being in combat).

It makes you use your general like a general and not like a hero.

[rbodleyscott]

...but I think the trick with smaller command radii is to not commit them to melee unless it is really necessary.

I think this is a key and very interesting difference with the new command changes (both to radii and rallies being dependent on being within that radii and the general not being in combat).

It makes you use your general like a general and not like a hero.

Which, of course, would be unrealistic for some generals. (cf Alexander)

Ancient generals were not like WW1 generals, several miles behind the line, drinking wine in a chateau. A lot (probably a majority, if you include sub-generals) did get stuck in.

Their troops often expected them to do so - which is why the +1 CT modifier for general nearby only applies if he is in close combat.

[stockwellpete]

Which, of course, would be more realistic for some generals than others. (cf Alexander)

Ancient generals were not like WW1 generals, several miles behind the line, drinking win in a chateau. A lot (probably a majority, if you include sub-generals) did get stuck in.

In the playtest I did today all the sub-generals did get involved in melees at some point. Only the C-in-C did not. In a more even battle I expect he would have done so too. I think the point is that sometimes you have to think a bit harder about what to do with your commanders when there is a smaller command radius in operation.

[Schweetness101]

...but I think the trick with smaller command radii is to not commit them to melee unless it is really necessary.

I think this is a key and very interesting difference with the new command changes (both to radii and rallies being dependent on being within that radii and the general not being in combat).

It makes you use your general like a general and not like a hero.

Which, of course, would be unrealistic for some generals. (cf Alexander)

Ancient generals were not like WW1 generals, several miles behind the line, drinking wine in a chateau. A lot (probably a majority, if you include sub-generals) did get stuck in.

Their troops often expected them to do so - which is why the +1 CT modifier for general nearby only applies if he is in close combat.

right, sorry I didn't mean to emphasize the difference too much in the other direction. I understand why the generals are important for both reasons. The while in combat morale bonus is very important for the simulation and gameplay purposes, especially to represent men like Alexander like you suggested, and I guess maybe more for barbarian type nations that expect to see the chief in combat.

I guess it's just for me in MP games most of the time the presence of a general is just used to make an elite unit even more elite, and a good way to win is to stack up generals on elite units in the same place to near guarantee a breakthrough. That is perhaps an artifact of playing in an unintended and gamey sort of way, but it is often a good move in multiplayer, and in multiplayer people play to win, often to a silly degree. I was just saying it's interesting to see generals implemented in a way that encourages you to hold them back a bit more, if just for a time, or even most of the battle, and to spread them out to their respective unit groups, in order to benefit from rallies at all, just to see how that changes gameplay.

[Schweetness101]

Ok I have swapped out the call to AI_ChanceOfCharging at the end of CheckPursuersSwitchTarget() for a call to a new function I've named Pursuit_ChanceOfCharging. Keeping in mind that the old AI_ChanceOfCharging() method is used to calculate lots of ai charges, not just the ones that derail a pursuit, so its general purposeness is accounting for all those other cases as well, hence why Richard's suggestion to just write a new one for pursuit charges, if that's what we want to change, is a good one.

for Pursuit_ChanceOfCharging() I just copied AI_ChanceOfCharging at first, and broadly made the following changes:

for if the combat margin of the would be charger is less than zero (less than 50% chance of winning the charge):
----it used to be that chargeChance was reduced to 50 (from base 90) against an enemy that might be able to shoot it (ie it should still be decently likely to charge such a unit), to 20 otherwise, and all the way down to 10 if the combat margin was less than -25 (very low chance of winning charge). Those considerations are more appropriate for consciously selected ai charges as it were, rather than for the more impetuous charge from hot pursuits. Instead, in this less than 50% chance to win charge section I have changed it so that the chargeChance is 50 for cav vs foot, 75 for cav vs cav, and 10 for foot vs mounted, and 20 for foot vs foot, but only 10 for foot vs foot where the combat margin is < -25. That's mostly to make cav less slippery when pursuing.

there is another section that multiplies charge chance by 3 if target enemy is in combat, which I kept. If enemy is not in combat then there used to be considerations like if you are a non shock ranged unit, and you aren't cav charging fragmented units or disrupted inf, then your chance to charge plummets down by a factor of 5. I got rid of that, which is a pretty big change, but it's again just for from pursuit charges. So, before, your shooty troops on a pursuit would avoid almost all but the most advantageous charge opportunities, which I think, and I think pete thinks, makes shooty mounted in particular too slippery when pursuing.

There is also a section under that same charging an enemy who is NOT occupied condition where:

If 'me' is better in close combat if the enemy I am considering charging instead charges me (like 'me' is lancers receiving foot charges type scenario I think? ie lancers would rather receive the charge than initiate against foot spears for example), 'me' is much less likely to initiate a charge, which makes sense for normal charges where the lancers want to wait for receive for example, but I don't think for pursuit charges. For from pursuit charges Lancers and others should be more impetuous I think, so i got rid of that charge reduction chance.
Also, it was the above consideration unless me's advantage if the enemy charges is sufficient that the enemy is very unlikely to charge (because then nothing will happen in normal ai charges, but in pursuit something is already happening as it were), either way got rid of the charge chance reduction.
Also there, likewise if the enemy has shooting advantage then charge chance was increased or something. Sorry this paragraph doesn't make much sense if you aren't looking at the code. The point is pursuers are more likely to charge a new enemy now, even in situations where they would not really initiate a charge from standing still otherwise if controlled by ai.

I also got rid of any consideration of flank threats

Also keep in mind that the check in CheckPursuersSwitchTarget() that calls Pursuit_ChanceOfCharging will pass, and initiate a charge, if Pursuit_ChanceOfCharging returns a value >= to 50, so like RBS said it is not a continuum, ie there could be a big difference between 50 and 60, but not between 60 and 80, or whatever, depending on how you line up all the numbers. I have not done much in the way of calculating out potential results, but instead just chose some plausible values. They will likely need a lot of adjustment.

Maybe once I upload this version we can just test and see what needs adjusting, or I can point you to where the values are changed and adjust them. And, because this is an all new method called only in this one place for charges coming off of pursuits, feel free to provide any random conditions you can think of, unit matchups or morale states or whatever, that should affect the chances of making such a charge from off of a pursuit.

[stockwellpete]

OK we can try that next. It sounds very promising.

I have just done a small stat in vanilla mode using my Training Ground where I set up 50 combats with a cavalry unit engaged to its front and then charged in its flank by an enemy cavalry unit of the same type. The results were as follows . . .

50 automatic cohesion drops (obviously)
17 out of 50 impact combats resulted in no further cohesion drop (many CT tests though)
23 out of 50 impact combats resulted in a cohesion drop to fragmented
10 out of 50 impact combats resulted in a double-drop to routed

So one-third saw no further drop in cohesion while two-thirds did (and 20% were double-drops). This is much higher than I realised. The next step is to compare this with data from a test where there is no automatic cohesion drop and only +100POA on impact instead of +200POA.

[Schweetness101]

OK we can try that next. It sounds very promising.

I have just done a small stat in vanilla mode using my Training Ground where I set up 50 combats with a cavalry unit engaged to its front and then charged in its flank by an enemy cavalry unit of the same type. The results were as follows . . .

50 automatic cohesion drops (obviously)
17 out of 50 impact combats resulted in no further cohesion drop (many CT tests though)
23 out of 50 impact combats resulted in a cohesion drop to fragmented
10 out of 50 impact combats resulted in a double-drop to routed

So one-third saw no further drop in cohesion while two-thirds did (and 20% were double-drops). This is much higher than I realised. The next step is to compare this with data from a test where there is no automatic cohesion drop and only +100POA on impact instead of +200POA.

updated op with new version of mod that changes cav flanks and introduces first changes to charge off of pursuit chances

[stockwellpete]

updated op with new version of mod that changes cav flanks and introduces first changes to charge off of pursuit chances

Is that Aggregate Mod v1.1? If so, still getting automatic cohesion drops for cavalry flank attacks.