Can we have an AAR from CMC ?

[vulture]

juan, I think what Jason is getting at is that humans are quite capable of producing results from individual CMBB battles that defy expectations to a huge degree. And in a program of any complexity (and this is pretty complex), it is pretty much inevitable that this can be used to break the system in some way - in a way that simply wouldn't be apparent in playtesting using auto-resolution of most or all battles. You can work on interface and data issues that way, but not game balance / mechanics issues.

To take one possible example that has been discussed previously, consider a battalion attacking a platoon. The platoon hides very effectively, and the battalion is unable to find it before the end of the battle. The result at the operational level is that the defending platoon continues to hold the tile, and the battalion is unable to move past it. Rinse and repeat. (Auto-resolving the battle would lead to the death of the platoon for little or not loss to the battalion in all probability).

It is of course relatively simple to work around on specific issue such as this. But as I said before, this is a complex game. It is virtually inevitable that there will be a number of unforseen game breaking issues like this that crop up. Some will crop up and be solved when using the auto-resolution. But some will only crop up when highly competitive nerds^HGrogs try every extreme idea in the book to gain an advantage - there will be ways of manipulating the CMBB battle result to exploit loopholes in the rules at the operational level. The auto-resolver will never do this - it will always give you resasonable results (and if it didn't, it would be cursed to high heaven and back again).

So to beta-test reliably, you need to have the testers trying their utmost to break the system, which means playing every battle (arguably, it means playing every battle in a campaign several times - keep the first one as the 'result', and then try every wierd and wonderful approach that occurs to you in the replays).

All of which boils down to the necessity of seeing the first release as a beta product. Without having the entire CMBB community beta-testing the game for months on end, things will be missed. Game breaking exploits will be found. Patches will be released to stop some. Others will be found to be too deeply rooted in the program design, or would break too many other things in the process of fixing this one. This is the way things are with all software of any complexity.

[Sergei]

Originally posted by vulture:

To take one possible example that has been discussed previously, consider a battalion attacking a platoon. The platoon hides very effectively, and the battalion is unable to find it before the end of the battle. The result at the operational level is that the defending platoon continues to hold the tile, and the battalion is unable to move past it. Rinse and repeat. (Auto-resolving the battle would lead to the death of the platoon for little or not loss to the battalion in all probability).

No, it doesn't. In each 1x1 square there is a flag. Whoever holds that flag is the owner of the square.

If a handful of troops hides itself out of the way, the enemy will just march to the flag and claim it.

Also, I remember Hunter saying that games of a certain degree of disparity (was it 1:10 or something) are always auto-resolved.

Besides, that's a matter of testing out the GAME. Testing individual campaigns is a different matter. You can never, ever try out every possibility that can take place in a campaign, because there is a near infinite number of possible end results, depending on how clever and lucky you and your opponents are, just like within every CMBB scenario there is a near infinite amount of possibilities for who moves where and fires at whom with what results.

[JasonC]

You don't get it, the players are not trying to do the things you imagine they are trying to do, and wrote the auto resolver around.

A company is probing another company, but in the same fight a dozen tanks enter on the right flank. The terrain is open and the defenders have no AT weapon bigger than an ATR. You'd expect an easy win for the armor heavy force, with modest losses, and the defenders lucky to get anyone out alive.

But actually, the tanks just drive from the right edge to the left edge, transitioning a couple op squares along the front line, and take no further part in the battle. Defenders hold losing a squads worth of personnel, attackers lose about a platoon, tanks move from op square to the right to op square to the left. ("Attackers" were not trying to take the op square. They were conducting a pinning attack to further an operational movement through a defending location).

A company of armor with infantry and heavy weapons attached, catches a column of guns moving by truck across open steppe terrain, only a few patches of scattered trees and rough, gentle slope hills. The defending column is wiped out completely, but a whopping 10 tanks are KOed or immobilized in the process. How? Defenders grabbed a valley and put guns at the bottom, and hit the tanks from reverse slope as they crested piecemeal. Attackers made poor use of their mortars etc.

A battalion attacks a two company position, and takes it with modest losses, wiping out a defending heavy weapons platoon (which showed themselves, defending "up" and firing). But the bulk of the "defenders" take the occasion to "infiltrate", passing through the attacker's lines and into their operational rear. Where they set up in the op-square the attackers came from, and hold.

A single infantry company, slightly weakened by losses, with a modest heavy weapons platoon of 4 HMG and 2 81mm mortars and one 105mm FO (60 rounds) in support, attacks an artillery battery position with 4 on map 105mm guns, a couple MG halftracks, a platoon of infantry and another of heavy weapons, plus HQs and trucks. Defenders are dug in to a village position in moderate woods. The attackers wipe out the defenders to the last truck driver, losing only one squad of men all told. How? Defender deployments weren't great, attackers made excellent use of arty and mortars to break sections attacked just before infantry hit them.

An infantry company supported by 3 Hetzers and a 150mm FO probes a village defended by more than a company, plus heavy weapons, 105mm support with abundant ammo, and 7 Sherman tanks, including 2 76mm models and all with W armor. The defenders also have combat engineers and mines. Result is the defenders hold the location, attackers lose 1 Hetzer and less than a platoon of infantry. Defending infantry losses are similar. But the defenders also lose 5 Shermans and a 6th bogs.

A full battalion of infantry (with HMGs, mortars, schrecks, and artillery FO support) attacking up a road in column of march, runs into a single infantry platoon of defenders, plus a little cavalry with a couple M-8 scouts cars, 2 60mm mortars, etc. Terrain is hill farmland. Result, the attackers take the position but lose 2 1/2 platoons of men, while the defenders lose a couple of jeeps.

14 tanks and a company and a half of infantry, including a platoon of pioneers, with some towed guns and FO support, attack a single green infantry company with heavy weapons platoon in open steppe. The defenders are dug in and have mines and wire, but their only AT weapons are a pair of ATRs. Result, defenders hold, each side loses a platoon of infantry, 1 tank KOed and 2 immobilized. How? Attackers got completely fouled up by a modest mine barrier.

Now, if you write an autoresolver such that events like those above happen frequently, players will correctly complain that they are unrealistic. But if you have players actually resolve the events, they will happen. In 3 campaign-days I've run, totalling about 50 human resolved battles, every one of them happened.

[Sergei]

Originally posted by JasonC:

Now, if you write an autoresolver such that events like those above happen frequently

And there's the magical word - frequently.

Of course, anything can happen. Some frequently, some less frequently. Do you have an actual point?

[Peterk]

True that auto-resolve will not know much about the actual terrain or about the players intentions (maybe other than basic stance).

The question is...is something unusual happening in some CMBB battles enough to undo a campaign design and make a campaign unbalanced?

I don't know if you play ASL but there's a nice analogy in the Red Barricades campaign. Tons of weird extreme strategies have been devised by players over the 20 or so years that this game has been around and they even have names like famous chess openings (Halftrack Rush, the ELR game, etc., etc.) .

The designer probably (almost definitely) did not think of all or any of them. And they cetrtainly weren't tested, since they came to light after the game was released.

..and Red Barricades is still considerred to be extremely challenging and a well balanced contest...because equally inventive defenders have come up with effective counter strategies which have also become well known.

Just because players come up with something unexpected that the designer did not test explicitly does not necessarily mean there's going to be a problem, if there's enough flexibility built into the campaign from the start and if the campaign is long enough.

By the way, I would hope that the auto-resolve is "chaotic" enough to show most of the results you listed above - infiltration, more movement than fighting, fortifications/bad weather stopping an assault cold are fairly "normal" occurences in my eyes. CMC can maybe use unit stances to determine probability of most of those so it's not totally random.

[Michael Dorosh]

Originally posted by Peterk:

I don't know if you play ASL but there's a nice analogy in the Red Barricades campaign. Tons of weird extreme strategies have been devised by players over the 20 or so years that this game has been around and they even have names like famous chess openings (Halftrack Rush, the ELR game, etc., etc.) .

Where can I read up on these?

[Sergei]

Originally posted by Peterk:

The question is...is something unusual happening in some CMBB battles enough to undo a campaign design and make a campaign unbalanced?

Or - is it not something that can be attributed to randomness and leadership stats?

[JasonC]

My point is that human beings behind the wheel are an improbability drive. 20% of all the fights in my campaigns had decidedly unexpected outcomes, for one reason or another. Never the same sort of outlier, but enough of them that the flow of the campaign was seriously affected by them. And these don't just happen to the players, sometimes they plan on them, playing to the in game situation in strange ways to accomplish operational ends.

Other examples - a single platoon of veteran German infantry with fausts is on a forested road. The Americans send past their position a tiny force of a platoon in trucks led by a single Greyhound. Result? Nobody hit on either side, Americans pass straight through the position. In game, a single squad ambushed the Greyhound with a faust, but missed. The Greyhound fired back and pinned that German squad, without hitting anyone. The whole US column then blew past the ambush position on "fast" without getting shot by the other Germans.

A US force of a single engineer platoon and 2 Sherman tanks, with 105mm support, "attack" a full German company in dense woods, with fausts and shrecks. By auto-resolve numbers, you'd expect a repulse and perhaps a bloody one.

What actually happened is the US lost less than half a squad, while the Germans lost a full platoon's worth of men and kept the position. The Americans were deliberately just showing up to toss HE, using their FO observed on the middle of a mass of woods, while the Shermans plastered the treeline with direct 75mm. The Americans made no attempt to advance and used their infantry just to protect their tanks.

They attrited the Germans with complete safety. But any auto-resolve feature would throw such tactics into the same averaging mechanism as attempts to seize the whole location, and would thus make the tactic risky. If repeated, e.g., there would be a high chance of losing a tank. In fact, it was not risky - the Americans were in no danger at any time.

Another example - a US reduced company is attacked from 2 sides by Germans with more than a company of infantry, plus 2 Hummels on map and a 105mm FO. Terrain light woods with a few hills. Result, no real contact or losses to speak of on either side. The Americans ran before the Hummels arrived. They weren't trying to hold the position to start with, only to run. The Germans were trying to catch them, not take the position. But failed to slow them enough with their flanking force, by not handling it aggressively enough.

Another - a single regular quality platoon with 2 attached sharpshooters "attacks" a full company with heavy weapons, on open steppe with a few trees etc, at night in a rainstorm. Result, the regulars pass cleanly through the position with no contact at all, inflitrating into their operational rear.

The moral is to test with human players. A campaign with human player resolution of all battles will flow quite differently from one with all auto-resolved battles. All auto-resolved battles will give relatively predictable slugfest results, as though both sides were fighting with determination to possess each location, each time there is a clash. In reality, barely half of human fights have that character.

[76mm]

The whole point of CMC is to provide an interface between the CMC campaign and CMBB battles. It seems pretty pointless to conduct "play-testing" that only involves auto-resolved battles.

While you might want to play-test a specific campaign in this manner (several times) to get an overall feel for campaign balance, it doesn't achieve much for determining how well the game itself will function.

[Michael Dorosh]

Originally posted by JasonC:

But any auto-resolve feature would throw such tactics into the same averaging mechanism as attempts to seize the whole location, and would thus make the tactic risky.

And if an auto-generated result did happen to repeat the odd performances, you'd see no end to the number of "FIX OR DO SOMEFINK" posts on this very forum, as everyone and his dog would be whining to high heaven about how "unrealistic" the auto-resolves were.

Truth is stranger than fiction most times...

[Sergei]

Originally posted by JasonC:

All auto-resolved battles will give relatively predictable slugfest results, as though both sides were fighting with determination to possess each location, each time there is a clash.

That is a baseless claim.

[JasonC]

Mike is right, if an auto-resolver gave wacky results all the time, players would object and rightly so.

Sergei, I know what I am talking about. You can get the mean performance about right with a spread of possible outcomes. You can get the variance about right for the first standard deviation or so.

Beyond that, out in the tails, there is simply no way to model a giant state space with a tiny one. And you couldn't even test a programmed version of a state space of the same size as real human games, in finite time.

It is not a matter of how you program it but that you program it. Repeated independent trials with randomness give gaussian statistics whatever you do. Human manipulations of strategic variables are not randomness, and simply will not add up in the same way.

The real tails will always be fatter and populated by the bizarre (from the standpoint of any random model), bizarrely soon in the distribution. You can't approximate what human mind A cross human mind B will produce, with a die roll. Period. You can get an average right and you can get the outcome of 2-5 on that die right. The rest will diverge.

Diverge a fifth or so of the time in ways that are not independent and you will get a totally different campaign, above all a less predictable one. Because randomness becomes predictable by being repeated independently, while strategic interaction does not.

I don't know why you are resisting the proposition so much, anyway. It is not like it is some great hardship to playtest a beta with humans. The thread promised that as already ongoing. Is it or isn't it?

[Sergei]

I'm not resisting anything, I just don't find your view correct or your arguments relevant. Something that could be done in your campaign might not be doable under CMC (like breaking a maneuver element up), and whatever strange combat results there might have been, a resolver can always be set to bring up similar results at similar likelihood.

[Peterk]

I think it's largely academic anyways. What owner of CMC is going to be able to resist jumping into CMBB to play things out? I _do_ plan on using the auto-resolve to quickly get a good general feel for things I'll work on though. And I'm hoping that that alone leads to something that's playable and fun, even if it may not be perfectly balanced per-se right off the bat.

========================================

Mike, do a search on Red Barricades in the Strategy And Tactics section of Warfare HQ. Lots of good Red Barricades strategies there. Little typo up above...it's the Armored Blitz and not the halftrack rush. I don't think people try it too often any more but you still have to defend against it just in case.

Are you planning on playing RB, or just reading up? I've been playing it solo for a bit and just started a PBEM of it. Quite intense.

[Steiner14]

It's so easy: the auto-resolver just has to deliver the most probably results.

If you expect something else, just play it out.

[JasonC]

Sergei - some won't be possible, check. If you realistically play them out human to human, then they will happen, but they won't fit in the system. As for the statement that "whatever strange combat results there might have been, a resolver can always be set to bring up similar results", here is how easily I can prove it isn't so. You can't even name the variety of strange results there might be. So how are you going to program them all in? If you can, go ahead and name them, right here right now.

You can't program something in you can't even think of at the time you write the program. You not only can't put a probability on it, you can't even name the set of things you don't have any estimate of the probabilities, of.

I agree with the last fellow's comment, that the autoresolver is going to give common results, right mean and a few SDs one hopes, and thus be more predictable overall. If you want wackier possibilities you will have to human-play, and I agree that is fine. It just also means, if you want to test whether a whole campaign will work properly (losses, time used, terrain gained, etc) then you will need to test human resolution campaigns as well.

[BFCElvis]

Peter are you playing a CM conversion of RB or somehow playinmg ASL through email? Mikes "Little Stalingrad" operation makes me think of RB. I have been playing it for a long time and am currently in battle 18 of 20.

[Peterk]

I'm playing the ASL version of Red Barricades by e-mail, and also by using solitaire rules to practice...

http://thewargamespot.blogspot.com/

...but I have also played 2 Red Barricade style long campaigns using my own rules and scenario file using CMBB. They might still be around here somewhere in the Scenario section of the board. The CMBB version worked well but lacked a little bit of Stalingradesque chrome that the ASL rules have (fortified buildings, mortars on rooftops).

I was the first playtester (I think) of Little Stalingrad and yes it is a fun operation. There was/is a gamey method of pushing the front line back without fighting too much that hurt it a little though when I played it back in CMBO days.

So when is this CMC thing going public anyways (taps fingers impatiently on desk)

[juan_gigante]

Originally posted by JasonC:

You don't get it, the players are not trying to do the things you imagine they are trying to do, and wrote the auto resolver around.

*A whole bunch of examples of battles the auto-resolver probably wouldn't deliver.*

Now, if you write an autoresolver such that events like those above happen frequently, players will correctly complain that they are unrealistic. But if you have players actually resolve the events, they will happen. In 3 campaign-days I've run, totalling about 50 human resolved battles, every one of them happened.

My point is two-fold. Firstly, it is true that a nearly infinite number of things can happen in a CMBB game. That's part of the fun. But even human playtesting will not catch all those. It won't even catch most of them. In my opinion, any game or campaign or whatever must either be extremely heavily scripted (which no one wants) or will have opportunities for this sort of irregular results. So I don't see why it should be a priority to check for this sort of thing.

This leads into my second point - just because the player does something extremely strange that happens to "break" the campaign, that shouldn't necessarily be a bad thing. After all, military history is full of examples of commanders doing strange, unexpected things that led to incredible success. If a player, playing one of my campaigns, happens to get lucky and discover that doing a specific thing would 100% guarantee victory, I would congradulate him. They merely had the same kind of unlikely, lucky victory that happens from time to time. Think about the paratrooper/glider engineers that took the fortress Eben Emaul (probably spelled wrong) in Belgium in 1940, easily taking out a major obstacle to the German advance. If someone had created a France 1940 campaign beforehand, they probably would not have taken into account the possibility of an airborne attack on the fortress, and the attack would be an example of "breaking" the campaign. But that does not mean that the campaign creator failed by not taking that into account. It just meant that human ingenuity triumphed again over planning. Besides, many possibilities of "breaking" the campaign only came up in one specific setup, a state that could only be replicated by the exact same events happening again, which is exceedingly unlikely.

In short, I am saying that even human-playtesting would not take into account many of the possible strange outcomes of CMBB battles, and that just because battles have strange outcomes that the campaign designer did not think of, even to the extent of breaking the campaign, does not mean that it is automatically bad. Just the opposite, in my mind.

[Ardem]

I will tell you where auto resolves will be useful.

Single Player campaign where you just want to play a single battalion in a divisional battle. So you auto resolve and you hope to get unexpected conquences.

[76mm]

even human-playtesting would not take into account many of the possible strange outcomes of CMBB battles

JG--I don't see the statement above as a convincing--or logical--argument not to conduct human-playtesting prior to an AAR. The main point of play-testing at this early stage is to determine whether the game works at all without glaring bugs, not whether every possible contingency and outcome has been covered in the auto-resolve feature.

Sure, while you will be able to play CMC on a stand-alone basis by auto-resolving all of the battles, that's not the point of the game, and its not how it is intended to be played. The whole point is to use CMC while resolving the battles in CMBB. If the interface between CMC and CMBB is not rock solid, then the game doesn't work as intended and there is no point to continue on to see how the game works if you auto-resolve all the battles. Properly testing this interface will require some amount of human play-testing, if for no other reason than you can't resolve the CMBB battles AI vs AI.

My point being: let the AAR come out when the developer thinks it ready, and let's drop the requests for some half-baked AAR which would be unsatisfying at best and misleading at worst.

[juan_gigante]

76mm - I think we might be talking about different things - I was refering to creating one's own campaigns using the completed game. Oh, gosh yes the game should be play-tested heavily by humans before release. I wouldn't want to play a game that was unfinished and released broken. I absolutely agree with you that humans should play the game. I don't think that is an arguable point.

But when using the completed game to create one's own campaign scenario, what is the proper role of the auto-resolver? That was the question I was discussing.

[76mm]

JG, hmmm, I must have misunderstood and read too much into the title of thread.

I don't have any coherent thoughts on the particular topic you've raised, so 76mm OUT.

[BFCElvis]

From what I have read about the game any playtesting that doesn't involve human played CMBB will be a waste of time. If I understand how the game will work one of the primary (if not THE primary) functions of CMC will be that it triggers a CMBB battle that is then played in CMBB and the surviving forces are brought back into CMC. Without this function this is little purpose for something like CMC. So playtesting and writing up an initial AAR without that (even if it is only human vs AI) would be a waste of a testers time.

Just one pods opinion.

[juan_gigante]

I'm not even talking about an AAR anymore. I'm talking about when I have the finished, completed game in front of my, and I decide to make a campaign of my very own, for humans to play using both CMC and CMBB, how much of the play-testing will the auto-resolver be able to do for me? I think that the answer is the majority of it.