Saturday 11 May 2013

Multiple Models and Wargame Emergence

It should come as no surprise to regular readers of this blog that I understand wargames rules as models of reality. They cannot, of course, be reality itself, or it would not be a wargame, and we probably could not get away with, for example, firing 25 pounder artillery in our quiet neighbourhoods just to find out what happens.

The situation is, in fact, slightly worse than we might think. A set of wargame rules is not a single model . We have models built in for all sorts of things, such as movement, shooting, close combat, morale and whatever else the designer thinks might be important. The way the model set is presented might, and usually is, different, but, nevertheless there is this set of models.

In fact, I think it might be argued that there are nested sets of models. For combat we have, perhaps implied, models for each different troop type against each other type. These may, indeed, sit within an overall combat model, but we could regard each combat type, by which I mean say, cavalry against cavalry, or infantry against artillery, as being a model in itself.

This might sound slightly alarming, but I do not think there is a particular cause for panic. In many areas of life assorted models are used quite happily up against each other. In physics, for example, we have wave-particle duality of quantum particles. This is, in fact, two models, that of an electron being a wave and that of an electron being a particle, which coexist quite happily alongside each other, even though the idea is counter-intuitive.

Even in less esoteric worlds of human endeavour we tend to have multiple models. I have two alarm clocks in the house, and they do the same job, but the user interface they present are different. The designers, presumably, had different ideas of how a person would come and set up an alarm clock. These interfaces are, of course, models, in that they attempt to form in my mind a mental map of how to set them.

The fact that one is more successful for me than another is not really here or there. The fact is that they are two models of a task that I would like to do.

So, two sets of wargame rules are, in essence, two different sets of models of some sort of battle reality. I think this has some implications for how much we like given sets of rules.

Firstly, there is the question of whether the individual models in a rule set fit together. If, say, the morale and command rules bear no relationship to each other, or contradict each other, then we might not like the rules. If my commander is within command radius and yet cannot encourage his men, we might look a little askance at the rules and declare the inconsistent. There has to be an internal coherence to the model set to make the rules work at all.

Secondly, the model set has to, in some way, correspond to our expectations of the real world the models are abstracted from. If, for example, our early WW2 Matildas are regularly taking out Tigers, while 88 mm shells bounce of the armour, we might suspect that the rule set was not a particularly accurate one. Similarly, if our Assyrians routed when a single Egyptian chariot appeared on the table, we might think that the rule model set needed some adjustment.

Thirdly, the way rule models sets are constructed is via the modern philosophical standby, reductionism. In my description above is reduced the battle to a set of component parts: cavalry against cavalry, tanks against tanks, or whatever. We then can model the individual bits, and build up our set of models to acquire an overall process of the battle. The battle, we imagine, is simply to be reduced to a set of interactions between these constituent parts, in more or less complex ways.

The question that arises is, of course, the one which arises for reductionism anyway. Does the sum of the parts actually give us the whole?

I have mentioned before the interesting fact the individuals behave differently in a crowd to when they are on their own. Of course, in military units that goes further. Individuals are trained to act in a different way within the unit. That is, after all, part of the point of having a unit at all. So we cannot, in principle, model a unit in the same way that we model a whole bunch of individuals. There is, in a unit, some emergent behaviour.

This is not, of course, fatal to our ability to wargame. In fact, given that we do not need to model the individual because of the emergent behaviour of the unit, it simplifies matters considerably. All we need to do is to successfully model that emergent behaviour.

The question is, then, do we need to go further. A set of units is, of course, a higher level unit, a division, or a corps, or even, depending on your scale of thinking (or the size of the historical prototype) an entire army. Do these objects exhibit emergent behaviour too, and can we, should we model it?

Now, many rule sets to at least pay lip service to army level emergent behaviour. In PM:SPQR (and, I think, PM: ECW, I must be more persuasive than I think) army morale is modelled. As a rule set they are not alone in having a model for emergent behaviour at the level of the whole force, but, given that there have been a fairly limited number of armies (as opposed to units) modelling the top level behaviour can be a little tricky.

Nevertheless, I believe that, historically, army commanders can affect the emergent behaviour of armies. Montgomery seems to have been able to do it. Wellington certainly seems to have thought he could do it. But it is not something that, it seems to me fits easily with our highly reductionist standard model sets for wargame rules.

Or, maybe, I am simply overemphasising something quite obvious again.


  1. A recent ( Ken & Robin podcast mentions something similar in the way board games are made. You can either start with a clever mechanic and wrap a setting around it, or start with a model of the world and abstract it into a game. Both can work well, but each has its pitfalls. I think that war games have similar properties of design, but with the added issue that, particularly for 'real' or 'historical' games, the abstractions can lead to some absurd strategies being optimal, that can damage the game's believability.

  2. Yes, I agree.

    The problem is that we can invent some neat mechanic and then look for a problem to solve with it. All the problems we come across, we find, can be solved by it, so long as we lop bits off sufficiently to make the problem the right shape.

    I've mentioned before the problems with DBR, trying to make the pike and sot interaction work, while the actual rules encourage the wargamer to deploy in a massively ahistorical way. It is the problem with models and mechanics: they only work so far, and then fail.

  3. To quote the (very recently) late George E. P. Box: "all models are wrong, but some are useful." Sorry, I just had to throw that in there.

    I think you're spot on with your comment on fitting the problem to the solution. Rulesets/models are conceived to function for a specific situation/data set. When the player steps out of that range the models validity fails which the designer may have missed. He may not even have bothered about it as he has probably been focused on staying within the range. Fitting the problem. Obviously I'm not saying that DBR models reality correctly when the player deploys ahistorically. However, given that we do not have a data set (battle) where real life commanders deployed and fought "gamey" I don't see how we can ever know. No matter how unlikely it is. It's hard to make a model cope with gamers' creativity. We all have to play the part if we want historical accuracy I suppose.

    1. Oh yes. My favorite example are Newton's laws; we know they are wrong (quantum mechanics and relativity) they are just too useful to give up.

      I suspect that lurking behind the discussion is the question of what we want from rules. If they tie us to using only historical formation, we complain about be restricted and the rules not covering some unusual situation we have found. On the other hand, rules that are too loose yield unhistorical formations and results.

      Probably it simply means that there is no such thing as a perfect rule set. But having multiple rule sets is healthy, as we can see that what this one does, that one does not.

      The next question is why that should be the case, which would need analysis of the actual models used. But that is probably getting too complex for me, at least.

  4. I don't think you are overemphasising the obvious at all, it's a great topic.

    I think there is a continual tension between playability and simulation, and where a designer falls between these axes will determine the kinds of models used. For example, a designer (Richard Clarke is an example) who wants to model "friction" will introduce mechanics that limit command and control, emphasize leader quality, morale, rallying, an unstructured turn sequence, etc. These rules might achieve an atmosphere of chaos while frustrating those who want a conventionally playable game. A designer who is interested in modelling the behaviour of military hardware on the battlefield may have multiple submodels for range, penetration, ammo type, and create a technically sound but unplayable game which forces a myriad of charts on the players.

    I'm currently interested in Napoleonic rules which model the battlefield at the division/corps level. The sub models that would make a game playable for me would be:
    1)limited command and control, so that not all leaders move every turn
    2)some system that makes some commanders more talented than others
    3)some behaviour that models national characteristics or doctrines of the various armies
    4)tactical rules that recognized the shock and morale value of the cavalry charge and its limits
    5)a game that's fun to play

    So that's my shopping list for a) a desirable model of the Napoleonic world as it is conventionally understood and b) some workable nested submodels. I probably won't get everything that I want, but I'm hoping to get most of it.

    1. I suppose that it comes down to a set of models that give the aspects of combat that the designer thinks are important.

      I guess it is then up to the wargamer to decide if the designer is correct...

      The list of constraints you outline is reasonable, but probably restricts the model set quite radically, and also might, in itself, be a little controversial.

    2. Just so.
      Any set of rules will focus on the aspects of combat in the period that the designer considers important. The aspects Michael says are vital to his enjoyment of the game may not be the same as the next person's. And it may all be down to the individual reading of history.
      For one thing, I shudder a bit at the mention of national characteristics (as opposed to doctrine) as these were sometimes dreamed up later to explain what had happened.

      Bruce Quarrie's Napoleonic rules doubtless ticked all the designer's boxes, but rules that split a move into eighths and record casualties in 33rds of a figure might not be to everyone's taste.
      You pays your money......

    3. Yes; and there is also quite a lot of fashion in rules design, from 'detailed' to 'fast play' and back again.

      Part of the irony of this, of course, is that fast play rules are often as accurate as detailed ones...

  5. Just a thought Michael. Why set your aim at the division/corps level. Is this because you want to be Drouet d'Erlon rather than Boney? Or because you're fitting the problem around your solution?

  6. As intuitive as it is to start at the bottom and try to model a battle by studying soldiers and weapons etc piling each on the other. It never seems to add up quite right. A bit like trying to make a model of glass by beginning with trying to understand and model sand. I think it is probably a mix of missing knowledge, lack of personal context/experience and complexity beyond our ability, (like trying to understand the meaning of 42). One alternative is to model top down by looking at over all events and trying to make rules which will lead to a wargame leading to the same result but for unseen reasons. Not very intuitive but it can be learned. A 3rd path is to try and balance over all effect with intuitive "colour" but this becomes more art than science

    It is an added complication that most of us have a better understanding of game mechanisms than of the reality and that this affects our reaction to a new idea or mechanism.

    Take for instance the current fad for "command control". We know that things didn't always happen how and when the person ordering them wanted and so it is common to make this a random event when most were the result of things that can be traced and explained. This tends to warp player perceptions of how generalship and armies thought and worked. It can still make a good game and can sometime even provide a reasonable result if well designed. But the players will never know what really happened, they will only see/learn what the game designer thinks happened.

    probably too long a comment already so I'll stop.

    1. I suppose that it depends on what we are trying to (excuse the scare term) 'model'.

      Are we trying to get an outcome which is realistic, or are we trying to create a command system which is historical?

      It is possible, but I think, unlikely, that we can do both at the same time. In which case we have to accept the abstraction for what it is: a game mechanic to obtain the right outcome.

      Given that we cannot really know what a historical battle was like, I think we have to stick to obtaining historically reasonable outcomes, and not worry too much about the details of the models used to obtain them.

  7. Ross, I think your comment is spot on. Especially the last paragraph. I'm very keen on command and control (or lack of C&C) and the fog of war. Rules which abstract these can produce games with "realistic" feeling outcomes, but as you seem to be saying, the abstractions leave you with no feeling for what might have happened. There's a lack of narrative - unless you invent an explanation afterwards

    I'm not sure there is an answer which doesn't involve a lot of admin.

    1. Even with a lot of admin, I'm still not sure a decent command system could be created that will accurately model a historical army command system.

      There are all sorts of issues about training, expectations and world view which cannot access, even if we write orders and take reactions from individual commanders. So it might be that a historical system is simply un-model-able, at least with any degree of playability, and quite likely not even then...