4586

Just got back to this thread. Great discussion. My long answer is, I dont know and I think it maybe impossible. But I need a few more years to think about it.

My short answer is I have noodled on and off with this problem for a few years. The best starting point it seems to me is generating “win” conditions then working back from that. Or creating weights for the AI is perhaps a better way of putting it. Once you have goals and some variables then the usual AI game playing techniques can be reversed to create a “tunnel” into the system. That tunnel through a vast area of possibilities becomes the boundaries for future players and so a “game”.

But an artificial umpire that can create a new game maybe impossible in principle. That would be a fine result by me because it would give a special role for consciousness.

4587

That makes sense to me, it is called backwards chaining. The challenge, from my pov, is how one can find a set of rules expressive enough to make a compelling game and at the same time, regression over the winning conditions and so on, recursively, is a feasible computation.

Keep doodling @Rod_Humble those are interesting thoughts.

4588

Hey thank you! I had no idea that technique was so formalized in backwards chaining, very helpful!

4589

I see that we’re in agreement to a certain degree, and I will grant that in many cases new data requires new models. This is quite true of games based in Ancient times, and if I am referring to P. Sabin work quite a bit is not by accident. When you only have heavily corrupted accounts of the historical event, new data - as new archeological evidence, somebody raiding the Vatican Library - can turn the narrative upside down. For example, and as already pointed, the matter of scale - size of armies, frontages, etc. - is problematic. Several “classical” accounts of major battles were shown by Sabin to be wrong at a fundamental level. Wrong as in not making any sense at all, once some makes some modest assumptions about the laws of Physics, and human psychology and physiology now and then.

Another, more recent example where the same you refer to happens is early Horse & Musket, like the games covering the War of the Spanish Succession. The historical issue with gaming in that era is that we know that platoon firing was identified by a number of contemporary memoryists as a key advantage of the English army (and of any army who used the same firing tactics). Yet, several studies point out that the overall weight of fire over a period of time, regardless on how it is distributed over that period, is the defining factor to gain a tactical advantage by attrition of the enemy, morale effects being difficult to quantify. So systems where model French units as fundamentally incapable of overpowering their English counterparts in close quarters need to have massive question mark over them, and I would say that it is not crazy to think that the importance of platoon firing has been overexaggerated by chest beating British historians. This can be appreciated for instance in this (non-academic yet very well researched) article by Jeff Berry

that suggests a very balanced way to account the effects of platoon firing. It would be interesting if anyone here is aware of systems that follow Jeff’s reasoning.

When it comes to WW2… I think not, new data will hardly cause a re-evaluation of any model… unless the model itself is making undue assumptions of efficiency due to “nationality” (for instance, Trevor Dupuy’s QDJM or some other jumble of letters). If you have been desigining a model around a particular outcome - Germans in good defensive terrain are hard to dislodge - it is way too easy to make a call to introduce a magic number to make the model “work”. Dupuy’s model is probably one of the most influentials in both board and computer war games, as his algorithms probably have been re-implemented, simplified or reinvented in spreadsheets by countless war game designers over the past 40 years.

That’s an example of what I call overfitting.

Another issue which doesn’t affect computer simulations is when humans misinterpret the rules or play them as if they were computers rather than people. That means that the rules where written expecting a very specific emotional reaction. For instance, your game with @Navaronegun had a bit of that… no Spanish player would have allowed your Republicans to take Zaragoza or keep Segovia.

Pat decided to go for the KO blow - take Madrid, damn all those funnily named Spanish cities - forgetting about the SD rule, which is modelling the fact that during the very few first months of the war there weren’t two sides, there were many sides, cooperating with each other to some extent (or not at all). The fall of Zaragoza - a major religious center and home of the Spanish military academy - so early would have been a major political disaster for the Alzamiento and a major boon for the lawful government of the Republica. Does that sound harsh from a gameplay perspective? Yes. From a historical standpoint? A bit of a stretch, but within the realm of the possible, as the sudden death wouldn’t imply that the revolting army officers would decide to disband their troops and go back home or exile. But rather, it is just a shortcut to avoid 150 turns of inexorable defeat.

On the other hand, GMT SCW serves me as well as an example that illustrates how hard can be to fix problems in models & data - rules, charts and materials - as used by boardgames. In GMT SCW, rating the early Republican columns with attack factors is actually ahistorical - I don’t think it is controversial to observe that ability to initiate offensive operations requires some level of staff work, something which was simply not existing for the Republican side at that point in time. Other than the North African Army, the readiness of Spanish army units in the Peninsula was abysmally low too, yet I would concede to them some offensive power.

With a relatively speaking, simple change, changing the counters from 1-1-3 to 0-1-3, you would get the kind of highly assymetric nature of the first period of the SCW. How hard would be to “fix” that in a digital version of SCW? Like entering a relatively complex Search & Replace query on Excel… And in the real world? You need to redo the counter art (even for VASSAL).

4590

Great analysis. I really do like your points on the difficulties inherent in modeling capabilities and outcomes.

When I lived in Virginia, I once visited Dupuy’s office (home? store? business?) and bought several of their books from them, and chatted a bit. I never bought into the modeling, really, though it’s pretty fascinating for what it is.

I sometimes think that you really have two choices in wargames. Either make something that is a tool for exploring, in effect, modeling, and seeing how close you can get to a system that replicates history, or make a competitive game based to a greater or lesser extent on historical events. In the former case, you aren’t interested in win states for the player, but in simulating the win conditions that would have been accurate for the conflict being studied. Here, the goal is to provide a sandbox where appropriate inputs produce appropriate outputs. In the latter case, the goal is a balanced, competitive game, where the primary objective is having both sides enjoy relatively equal chances of victory within a roughly historical context.

When you try to do both, I often think things get very wonky. Things you need to do to make accurate historical models of events usually do not lend themselves to balanced competitive play. And as we all know, wars and battles are very rarely evenly balanced affairs, so focusing on competitive gameplay is going to screw with historical modeling something fierce.

4591

Yes, we mostly agree.

The problem is that, barring a total and complete simulation of all combat factors (including psychological factors all along the chain of command), every model will have to make assumptions and introduce magic (soft) numbers in the equations to make each side work as it historically did. Stuff like CMANO doesn’t make too many assumptions, but it also doesn’t model most of the factors involved in a conflict. In a traditional wargames, assumptions are everywhere. And new models that looks at different data appear frequently (here I’m thinking about Biddle’s 2006 model that diminishes force ratios as a factor in modern -WWI to contemporary- combat).

At any point a simulation is going to be subject to the model it uses for the simulation, and while partial fixes are indeed easier on a computer, more complex fixes (new rulesets or partial rulesets) are easier to acomplish in cardboard (for example the 2 set of alternate Napoleonic rules that use LNBs counters, maps and setups but otherwise work very differently to simulate command more directly).

So going back to the point about fidelity: if you are going to have these kind of assumptions into a model, I prefer them to be clearly visible through abstraction rather than hidden in one of the several datapoints a particular unit has. That is, when I play a wargame I’m more interested in learning and playing with those assumptions and trying to understand what they mean or try to say. That’s why I find many computer wargames somewhat dry.

4592

Not, so. Pat actually couldn’t advance (supply situation in the South) until he had no choice but to dream an impossible dream of a Blitzkrieg.

As well, those “funnily named cities” (that he was desperate to keep) couldn’t be kept-there were no reinforcements to give them at the time. “allowed” didn’t enter into it.

I think your “emotion” argument is spurious here. Good post though.

I agree regarding combat factors for Republican Columns, but they did make attacks historically during the column phase. So history belies what a good analyst or program would “assess” about that situation hypotheticallythat situation. :)

As well a computer AI would make a hash out of either side. AIs are horrible at attacking creatively, deception, and maneuver. All important command attributes.

The counter art (in Vassal is as simple as a “find and replace” as well. The same would go for changing the counter art (in Adobe pro).

4593

Ah, you met the man… what a fascinating character, really.

Very well put. I have nothing to add to that. Observe how this observation is independent from the medium, it applies to both types of war games.

4594

One concerrn with removing the attack factor on Republican columns is that it might allow the Nationalist player to take too many risks with token garrisons. You’re on point with the problems that the columns had as military organizations, but i’m not sure removing their offensive capability won’t just cause other problems.

4595

The designer is of the opinion that the SD rules need to be tweaked (again).

4596

How is abstraction transparent? I would say that it does the contrary, especially if designer notes aren’t available.

4597

Spurious? Perhaps. From the AAR write up I didn’t get the sense you were desperate.

Indeed, columns engaged in offensive operations, but not at the scale of time and space portrayed in the game. You can attack a few villages here and there, go for a piece of key terrain… But to organise the siege of a provincial capital, keeping your forces supplied for a couple weeks it wasn’t until late 1937 that the Republican Army could do so.

Edit: there even was a fully organised Army, in the conventional sense, until 1937.

4598

Yes, probably you’d have need a more active supply system, a la OCS to get things moving.

4599

Abstraction tends to make intent more explicit. Of course it depends on how it is presented (UI, in sort) and how well focused the abstraction is, but in general a well abstracted mechanic clearly presented will show design intent clearly. Lack of abstraction makes the player responsible for discerning intent. Which, granted, sometimes might be preferable. Lack of abstraction does highlight cause better for the player who is able to discern intent, though. For example, supply in OCS.

Designer notes will help in both cases.

4600

I am very familiar with he history; the designer was attempting to simulate the Column War period and then have the units themselves, and the time scale change to reflect the transition to the grinding war of divisions.

It’s pretty well done, and the Republicans are horrible until the Brigades begin appearing later. But the columns are very wonky with ZoCs and what those ZoCs do in the early game. They don’t inhibit movement or cut enemy ZoCs but the do cut supply. It really slows down the Army of Africa. Historically, the were taking Estramadura, while I was securing a rail line. :)

I highly recommend you pop open the rules and give them a read, it’s a noble failure. Tweak the SD rules and you have a very interesting game.

Even flawed, it was good enough for @Juan_Raigada and I to play twice in a row. I think it deserves criticism, but suggest that he issues with it lie elsewhere than the horrifically bad Republican columns in the first 6 months. :)

As well the lack of staff work, etc. that you speak of is actually baked into the system. For instance all Republican attacks , if declared, have an even (or close to it) chance of not occurring. I’d really check out those rules.

4601

Well, that’s because I am an experienced cardboard “Commander”. Channeling my inner Yague y Blanco. :) There wasn’t much I could do about Aragon. You just have to ride the tiger early in that one

4602

I do agree with the ZOC rules thing being a bit arbitrary, the ability to interdict transport within a 50 kms range would require a degree of mobility and active patrolling that simply wasn’t either there.

I also think the game vastly overrepresents the combat value of Republican forces in Extremadura. Historically, the “campaign” was basically a no contest, as there were no organised militia forces in the region.

I was familiar with the 1.0 rules - some of the Spanish peeps that posted reviews, materials etc. in BGG are gaming acquantainces of mine (played together or with friends of them) and this game was widely discussed (not so sure how often was it played).

I checked out the 3.0 rules before making my post :-) I can’t help to note that the SD rules are Optional Rules, and they suggest that there’s a bid before the game starts on the number of locations necessary for the Republican SD rule to trigger.

I salute you - it is a fascinating topic.

4603

I’ll let @Juan_Raigada speak to that one. IIRC, the latest issue of the playbook mandates the SD rules. It’s the bid that is optional. Having said that, we wanted every bit of historically relevent chrome we could throw on the thing. And it was worth it. I had a blast. Sometimes you can lose, and be at the top of your game. I felt like it in that game. I’ve rarely had the highs of a El Día de Los Cuatro in my years of wargaming.

FWIW, I think they kept trying to slow down the Nacionalistas in playtesting…but used a Playtest group that was too internecine. Outliers may have looked like norms. BGG really shows that the SD rules have had issues for some time.

4604

Now I get it. What you mean is that since the rule is written English, it is true that what the factors on the counters do is readily apparent to the player. In computer war games, unless you have access to the source code (and you have the technical knowledge to make sense out of it) you are none the wiser unless there’s some good supporting documentation. The manuals in computer war games are sometimes complete enough to explain what the factors do and how they’re used in English too, they are in effect an attempt at conveying what the codes do with the data, besides also covering for proper designer notes.

But properly documenting is not consistently done - many designers don’t bother or aren’t generally good writers :-) In terms of documentation, Command Ops was very good, but eventually the documentation went out of sync with the code in some significant ways.

Also, when the editor was opened up for the public, a bit as an afterthought, instead of seeing a ream of new scenarios what we saw was quite a number of people scratching their heads in utter confusion trying to figure out what exactly “Ronsonability” was doing, or why anti-personnel fire accuracy curves had a particular shape. Not that either of those concepts were important for the player - the former didn’t have much impact as tank recovery routines were never implemented, the latter actually had a small effect in combat outcomes - but actually detracted people from engaging with the editor.

Grigsby’s WITE and WITW - to which I will come back soon to answer to both @Lohengrin and @Navaronegun - manuals are also good examples of documentation that I think serves the purpose of illustrating how the different components of complex computer war games work together, with varying degrees of success, to obtain a plausible simulation of the subject matter to support a game.

4605

I’m looking forward to this explanation: in my experience, the WitE manual was a great example of the “here are some things you’ll see, sorry we can’t explain how any of it works” theory of documentation.

cf https://www.quartertothree.com/fp/2011/03/24/war-in-the-east-l2p-nub/