CMC Training and Possible Problems

[John_d]

I was having a poo the other day when I had a stroke of genius. It occurred to me that the tactics we will have to use in CMBB sections of CMC (henceforth known as CMCTACT, as opposed to the operational level past of CMC. Which I will call CMCOP. Just for simplicity really) will be very different from those tactics that we currently use in standard CMBB. Of course, this has already been covered several times in this forum, but I came up with an idea- why not create scenarios in CMBB that will reflect the kind of things that we will come across in CMCTACT? For example, recon work where the object is not to kill the enemy or take any territory, but simply to see what is out there and then get out quickly. Or infiltrating a company of infantry through enemy lines at night (or even in a city) without making contact with the enemy. Or concealing a large force from a recon platoon as effectively as possible. Or defending an artillery battery against a far superior force (which I think would just involve running away). Or ambushing a battalion in convoy with only company (or even a platoon).

So I got to work creating scenarios that would reflect the weird and wonderful new experiences that CMCTACT would offer us. I designed the maps based on my understanding of how CMCTACT is supposed to look- 2km x 2km maps with large victory flags in each 1km x 1km quadrant. The one thing I was unable to work out from the information that we have been given is who controls which flags at the start of the battle, although I suppose that this will depend on the situation in CMCOP at the time. I didn't really bother playtesting them for playablilty to reflect that you won't always get a pretty looking, balanced, fun to play battle (in a CMBB sense- I imagine that CMC will completely change the definition of what is and isn't fun to play, given that unbalanced forces will be the norm and there will be much more opportunity and call for retreating) in CMCTACT.

However, I noticed one major problem. The AI (understandably) always behaved as if it were fighting a single, isolated CMBB battle. Recon is surprisingly easy when the enemy makes a banzai charge towards one corner of the map to capture a flag. Basically, all battles got reduced to meeting engagements. Not only is this historically inaccurate, but meeting engagements are the type of battle handled most poorly by the AI and IMHO the least fun to play generally. Also, the AI made no attempt to reduce casualties. A further unforseen problem is that the victory flags no longer really mean anything. Because of the way that they are distributed (i.e. one per 1km x 1km square) they often end up in totally inappropriate places. Like in the middle of a field or a river or a lake. Then, predictably, the AI runs to whichever flags its does not control, which doesn't even give it a slight strategic advantage like it does sometimes in CMBB, by holding a hill or a group of buildings or so forth.

Anyway, the upshot of this is that for in order for CMCTACT to be playable against the AI, some serious changes are going to have to be made. Either the flags will have to be positioned differently (i.e. by the scenario designer) so that they only mark out points of strategic importance, or the AI will have to be recode to prevent the banzai charges and conserve troops for later.

If anybody wants to try out making mock CMCTACT battles- just use the random map generator and follow the instructions above. I have a few, but it seems kinda pointless for me to send them, when they take so little time to make yourselves

[76mm]

As far as I understand, the victory flags can't be moved by the campaign designer, so I don't think that will be possible. But you did leave out another possible solution, which is for campaign designers to create their CMBB maps so that the flags aren't in the middle of wheat fields. While this might not be possible with 100% of the quadrants, I think it should usually be possible to place a small rise, a building, a patch of woods, etc. under the flag.

[Hensworth]

We're looking for some players for Onion Wars, specifically for the Soviet side. Don't be fooled by the silly name, this is a serious campaign which has covered 16 months of all out war.

If you want to get into the groove of operational thinking, it's the place to be.

[Christian Knudsen]

It surprises me that this post hasn't generated more commentary, as it points to a major potential problem with single-player CMC - how to make the CMBB AI deal with an extra level of operational requirement.

Nonchalantly, I, with no concept of how to program anything more complicated than a VCR, wonder if a system similar to that in ROQC might work, in which every CMBB battle takes place in a simple operational framework that weighs victory points. Another possibility is to utilize the handicap system - does anyone know how the AI acts when confronted with an insurmountable points handicap?

I suspect that another answer will be to abstract recce for the computer player - It will be enough for it to move recce assets near to human positions in CMCOP for the computer to aquire intel. I would be happy with this, for several reasons. First, a good recce isn't seen by the enemy anyways, and second, it will avoid "recce rushes" like John_d mentions.

As regards the force incompatibility and retreat battles, I trust that Hunter and the rest of the gang will be able to implement some system whereby depending on the commander, intelligence, et al, the enemy will conduct an exit battle to its own friendly map edge(s).

With regards to the other situations that John_d mentions (infiltration and ambush), I think that CMC will ned a powerful internal setup zones editor - like with the capability to make an ambush setup zone complete with flag(s) to anchor the ambush force...

However, despite the difficulties raised here, I remain supremely confident that CMC will arrive and be a great game for single player play (I KNOW it will be great for MP).

Incidentally, I wonder how many of mankind's greatest thoughts and inspirations have come while on the potty? There must be a self help movement to inspire somewhere in here. I can just see the infomercial...

P.S. - Sgt Kelly, I wish I had time to play Onion Wars. Please keep it going, and in six years when both my kids are old enough to go to school, I'll be there with bells on.

[Peterk]

Based on my attempts to sim. Red Barricades in CMBB, the AI does not react at all to the points handicap. It just wants flags. All of them.

Easy enough to test. Set up a small map with the computer AI on defense. Give it a huge bonus so the attacker (you) can't possibly win the game. Now stick two small victory locations in the attacker's setup area.

I'm pretty sure the AI will go for the flags.

[Christian Knudsen]

Ok Peter, I decided to give it a shot. I built a standard (so we think) CMC Map, 2000m x 2000m, 4 quadrant victory flags, farmland, small (normal) hills, light cover.

Gave the defending Germans the west half of the map, a fortified batallion with armour and arty support. As well a 5000 pt bonus. The whole kit and kaboodle, in other words (but set up quick and stupid - no Observation Posts or forward defenses prior to Main Line of Resistance).

Russians got a recce platoon and some scout cars. Gave them a 250m strip on the east side to set up in, as well as an exit zone running the whole east side of the map. Idea is, can't win, don't try, run away.

What ended up happening (Played without FOW) is that the Russian milled around trying to take both flags, and largely got shot to smithereens, about 60% casualties. Considering that most of the Germans never actually fired due to the 1000m plus engagement range, these are pretty ridiculous losses.

One might argue that the AI would handle things differently if fighting at night, or with extreme FOW, but I would tend to think not. If I were to try it again, I might shift the handicap so that the Russians can't lose. I suspect that this might not make a difference in how the AI reacts, however, as the 1200 points of flags too seriously outweigh the point benefits of exiting. What is needed is the ability to vary the exit points award. Say that each unit is given 100 times its point value for withdrawing, thus outpointing the flags.

Big problem with this though.

Say the AI is awarded 2000 points for withdrawing its troops in the above situation, assuming that no point handicap is awarded for either side. What will happen is that the game will hand the Russians a minor victory, (2000pts vs 1200pts) with whatever that implies. I hardly think it rational to be able to dislodge a dug in batallion by successfully recceing it with a single platoon. Even then, the AI will likely try for an even bigger victory by trying to grab at least some of the flags.

At any rate, I think that as far as recce goes, CMC would be better served by abstracting it. Make small recce units hard to detect in CMCOP. Allow them to stealthily move stationary to enemy units and gradually gain CMCOP info on them, thus abstracting the whole patrolling process. In terms of armoured recce, allow it to gain broad info quickly from adjacent squares, but make it much less stealthy - both sides get info, while the armoured recce can fulfill its role of finding and screening large enemy concentrations.

Similarly, I think that large force disparities should be handled abstractly. Let's say that our German Batallion is being observed in CMCOP by our Russian recce platoon. Suddenly, the Batallion lurches forward into the very 2x2 km square solely occupied by our recce gods. Yikes! In CMBB (barring changes), this would mean the certain destruction of the Russians. In this case I would understand and appreciate if the CMC engine withdrew the recce platoon to an adjacent friendly square, or even left it in place and have it pop up in the german rear after the Batallion passes. Much more fun and quite possibly more realistic too.

The exception would be when a ridiculously smaller force is out of movement, or is being pursued by a quite faster force, in which case the annihilation of the smaller unit would be justified, and we can live with quirky AI tactics because they are doomed anyways.

Anyways, while this test has confirmed limitations in the current CMBB engine as regards to CMC situations, I hope that it will not prove difficult to build workarounds into the CMC engine.

[Peterk]

The solution in my eyes for most of the AI limitations is not to have any unposessed flags on the map for the AI to attempt to capture on defensive missions. That and/or padlock most of them in place so they don't abandon defenses.

Trying to prompt the AI to run off map is another can of worms altogether. Have no idea if it's doable.

[Christian Knudsen]

Oh crap - I never even thought about the split map defender scenario. I think that the only way around this one will be to eliminate the attacker's flags - ie have a battle in which the only flags on the map are on the defender's territory.

The downside to this is that it will in essence be impossible to lose ground so long as you are attacking - If the defender is reinforced in mid battle and takes a quadrant back, this will be very hard to portray in CMC, short of cheating by not allowing the reinforcements on until the following 60 minute turn. Another workaround would be a variant on the old 'bounceback' style rules, whereby if the attacker captures none of the defender's terrritory and losses are above a certain threshold, the attacker is forced to withdraw to an adjacent square.

Basically, the only way in which all four flags will be on a map is if the defender holds the entire square, or if both sides in or moving into a square attack; ie. a meeting engagement.

[Peterk]

Padlocking would work even with 4 flags onboard. On a normal defense mission padlock 75% of the AI forces, on an assault defense mission padlock 90%. If reinforcements come in on a flank, fine they will attack - and it's not too far from actual German doctrine of immediate counterattacks. So that way you still get good defense around the flags with just a touch of offensive activity (patrols?).

Forget about CMC; this is a feature which would be great in normal CMBB as well against the AI IMHO.

[Hensworth]

Making sure the AI starts off with flags in possession and padlocking its units doesn't do much to change the behavior.

As soon as the AI loses one flag it will send all its units out of their foxholes and streaming towards the lost flag in a doomed attempt to reclaim it.

Padlocking does nothing at all, because the AI will set up on and around flags if it can alright, it just won't stay there if any flag is lost.

[Christian Knudsen]

Well, I think the bounceback idea would work the best - not the perfect compromise, by any means, but one that would not hurt the game too bad. Make the casualty level needed to trigger a withdrawal dependant on an attack setting in CMC - probe, attack, assault, as well as ratings for the formation commander.

Besides, this would lead to some fairly exciting and fluid attack/counterattack scenarios, in which an attacking force tries to take part of a square, is ejected by a sudden reinforcement, and jumps back right in the very next turn. Look at it as a quick reorganisation, if you will.

Another possibility is to create pointless flags for the attacker. Leave flags that have no point value so that if the attacker loses them, they lose the quadrant, but the AI isn't trying to rush after them. Of course, the AI isn't going to be inclined to go anywhere near them, either...

Oh well. I have a dream (a pipe-dream, admittedly), that BF will annoumce that they are releasing CMC with a reworked, patched up to CMAK CMBB. But if wishes were fishes, I guess we'd all be swimming.

[Peterk]

Doesn't padlocking keep the unit in position until they are fired on? Thought that it did.

CMBB needs something similar to the ASL No Move counter to keep defenders defending.

[Pak40]

It seems to me that the obvious solution would be to incorporate an "acceptable loss" factor to both sides. For example, the AI defender would be set to one of the following options:

</font>

• Defend at all costs</font>
• 80% casualties</font>
• 50% casualties</font>
• 30% casualties</font>
• 10% casualties</font>
• Runaway!</font>

Similarly, the attacker AI would have options such as:

</font>

• Take Ojectives at all costs</font>
• 80% casualties</font>
• 50% casualties</font>
• 30% casualties</font>
• 10% (probe with minimal losses)</font>

For instance, if the AI chooses 30% casualties for a particular battle, then once it's casualty rate reaches 30 percent of it's entire forces then it will call off the attack and either retreat or stop pressing the attack.

Obviously this would require some re-coding of CMBB but could be provided in a patch.

[Christian Knudsen]

It seems to me that the obvious solution would be to incorporate an "acceptable loss" factor to both sides.

And this is what I would expect when dealing with the auto-resolve feature. For it to work in CMCTAC, however, it would need an 'ignore losses' button, so that in an emergency you could take that important flag despite going over the casualty limit.

To make things more interesting, I think that the percentage levels should not be spelled out, but instead be largely based on troop quality and commander traits. "Heavy" losses for a poorly led Luftwaffe infantry company and "Heavy" losses for an aggressively led SS Panzergrenadier company would likely be different.

I still think that for this to be effective in terms of the AI counterattacking as a defender without flags, the 'ignore losses' button can only be accessible if one or both of the defender flags is either attacker owned or unowned - In other words only if the attack is making good progress.

Either that or make it totally the players choice. You know (or don't) that the enemy has recieved major reinforcements, and that your attacking forces may be in danger. At the appropriate point, the game asks you the following:

You are approaching an unacceptable level of losses. Do you wish to continue the attack?

If you say no, your forces are withdrawn to an adjacent friendly square, maybe with some random losses that take place during the withdrawal. If yes, then you are locked in for the remainder of the 60 minute battle - for better or worse.

[Pak40]

The 'acceptable loss' level should only be on for the computer controlled AI. The human player doesn't really need it since he/she can decide what is acceptable and what is not.

I think there should be two levels to the 'acceptable loss' system when attacking. For instance, let's say the acceptable loss is set for 30% for a particular battle. When 30% casualties are reached then the AI will stop the attack and order units to find best cover available but basically not concede most of the ground gained. However if the AI continues to receive casualties past a certain threshold, then the AI will pull out all together. Maybe the 'retreat' threshold should be something like 15% more than the original setting, so in the example above the attacking computer AI will retreat from the map at a 45% casualty rate.

The Campaing designer will set the 'acceptable loss' levels for each map and each battle group. Towns and cities will have higher acceptable loss levels because they are more imporatant than a flat and featureless land. Maps that have defensable positions or strategic importance such as a bridge or hill may be deemed higher acceptable loss.

Battle groups will also have acceptable loss levels set by the designer since different units have different personality commanders as well as different rolls. For example, a Battalion HQ unit will most likely cut and run before accepting a lot of casualties.

The Map 'acceptable loss' value will be averaged with the battle group 'acceptable loss' value. I.E, if the map is set to 50% and the AI battle group is set to 30% then actual value used can be 40%.

To make things more interesting, I think that the percentage levels should not be spelled out, but instead be largely based on troop quality and commander traits. "Heavy" losses for a poorly led Luftwaffe infantry company and "Heavy" losses for an aggressively led SS Panzergrenadier company would likely be different.

I agree that they would be different but it's really already inherent in the game. SS Panzergrenadiers will usually get elite or crack status while volksgrenadiers will get green or regular status. Therefore the volksgrenadiers will more than likely cut and run before the SS troops would anyway.

Anyway, it should all be determined by the Campaing Designer, and it should be totally behind the scenes & never known to the human player.

[ February 14, 2006, 08:12 AM: Message edited by: Pak40 ]

[Christian Knudsen]

True enough. I would not want the setting of a loss level to overly dictate how I play in CMCTAC. However, I would concede it as realistic if certain loss levels led to action from higher - I needed your batallion combat-effective more than I needed that junction, for example.

The problem returns, however, to how to effectively handle AI defense and counterattack without having every battle degenerate into a meeting engagement. Clearly the flags that are owned by the attacker have to be negated somehow within the context of CMBB. This is easy enough to do*, but raises the question of how to handle a battle in which one force already in the square attacks, and then an enemy force launches an attack into the square later in the turn, thus in CMCTAC arriving as reinforcements. If you have set the first force as the attacker without flags, what does the new AI force attack?

Of course, you could stop the first battle when the reinforcements show up, padlock and freeze everything, and launch a new battle that is in effect a meeting engagement, but that would be far too easy*...

* easy for me to conceptualize and talk about, that is. Hopefully it will be easy to program, too.**

** but probably not.

[Pak40]

If you have set the first force as the attacker without flags, what does the new AI force attack?

My guess is that you just don't use flags. Program the AI to conquer the map using the best methods possible, i.e. appoach under cover, use topography to your advantage, use artillery etc. If the AI force is attacking, it doesn't need flags to tell it to attack. Consider that the objective is to eliminate the enemy from the current map or maybe to probe enemy lines. It's also possible for both sides to start a battle in a defensive position both with orders to defend.

[Christian Knudsen]

My guess is that you just don't use flags. Program the AI to conquer the map using the best methods possible, i.e. appoach under cover, use topography to your advantage, use artillery etc. If the AI force is attacking, it doesn't need flags to tell it to attack. Consider that the objective is to eliminate the enemy from the current map or maybe to probe enemy lines. It's also possible for both sides to start a battle in a defensive position both with orders to defend.

Yes, but it seems that this would require a complete reworking of how the AI works in CMBB, and I think it highly unlikely that we will see that, given BF's current Schwerpunkt in CMSF.

This is why I think we need to help come up with ideas that feature absolutely minimal changes to CMBB. Don't get me wrong - I'd LOVE to see CMBB reworked to CMAK standards and then made totally complementary to a perfectly realistic CMC. But while CMC will rock, I have a nagging feeling that we won't be seeing that major a reworking of CMBB.

(Prove me wrong, kids, prove me wrong.)

As for double-defenders, why even fight a battle? I would hope that it will be possible to have opposing troops in the same square without a battle being triggered - otherwise it will be too easy to attrit the AI - gain a quick lodgement in a square, take one flag, then sit back, play defense, and take advantage of those stupid AI tricks we all know and love...

[Pak40]

Originally posted by Christian Knudsen:

Yes, but it seems that this would require a complete reworking of how the AI works in CMBB, and I think it highly unlikely that we will see that, given BF's current Schwerpunkt in CMSF.

Not at all. Advance and Assault operations in CM don't even use flags, so I don't think there's much need to recode.

Originally posted by Christian Knudsen:

As for double-defenders, why even fight a battle? I would hope that it will be possible to have opposing troops in the same square without a battle being triggered - otherwise it will be too easy to attrit the AI - gain a quick lodgement in a square, take one flag, then sit back, play defense, and take advantage of those stupid AI tricks we all know and love...

I agree, why fight the battle? But this has to be accounted for because both sides may be in the same map yet neither side may wish to attack. Like a Sigfried Line situation. I suppose they could code CMC to ask the player which maps he wants to defend. If the AI also wants to defend the same squares then the battles can be skipped. Or have the battles take place but both sides can call a truce anytime during the battle.

[John_d]

There is always the option that the human player is 'noble' towards the AI and just doesn't exploit weaknesses (as far as possible). This could include routing but not killing forces and retreating before a banzai charge.

The only way around this would really be to reduce the scale of each battle from 2km x 2km to 1km x 1km. The original design for the CMCTACT battles was that 4 tiles could be fought over at once, which makes sense, given that you could capture one objective and then move on, but I think that a reduction of battle size would be a small price to pay for solving the AI bugs. Alternatively, the flags are placed by CMCTACT at the start of each battle, rather than by CMCOP or the editor so that the victory flag only appears in the middle of the defender's positions when the game starts up (and only have one flag for each 2km x 2km battle). If the flag is captured by the attacking force, then the defenders are moved back 1km. If the attackers fail to capture the flag, then they are moved back to where they started from (i.e. 1km back)

Maybe the upshot of CMC will be that it will just be a game that is only really worth playing against humans. But even then, its gonna take a long, long time to finish a game.