Saturday 1 November 2014

Reality and Multiple Models

I may well be repeating myself here, but I would like to ponder something that has come up again in recent discussions here. The thing is, in a wargame, we are taking multiple complex reactions to a human (and very stressful) situation, breaking them into models, and then reassembling them into some sort of intelligible, usable whole to play a game.

I think I would like to leave aside, at least here, the ethics of doing this. I have considered them elsewhere, but I do confess that the above description of what we do in a wargame makes me feel slightly uncomfortable. However, I shall simply ignore that for the moment and consider the actual problems of the reductionism that we employ and the possibilities of reconstruction that we hope for.

So, firstly, reductionism. We have a set of accounts of battles for a particular period, say, plus our imaginations and a framework for what we can produce as a workable model and what we cannot. So, for example, I know that my model, in order to be intelligible and tractable, is going to have to result in some sort of arithmetical operation, a dice roll for a bit of chance and some sort of outcome based on the numerical result. This may not be the only way of resolving wargame events, but it is the most usual one. We could call it a paradigm model of event resolution.

Now, in our accounts, in our imagination, various factors come into play. The men, for example, might be confident or treacherous (think about Bosworth, for example). The leaders might be, or might be considered to be, competent or not, careful with the lives of the men or not, and so on. The weaponry the armies are issued with might be effective or not; the tactics employed could be such as to amplify or supress the usefulness of the weapons or not, and so on.

I think the issue here is that, in real life, these things are all presented as a mass. Generals, I guess, do not go around thinking that if only they deployed their archers in hollow triangles they might be more effective. These things are largely dependent on tradition, training and other things that happen off the battlefield. Generals, deploying their armies, have to use what they are given. The men are armed in a certain way, trained in a certain way, have a set of conceptions and expectations about themselves, their comrades and leaders, and, probably, the enemy as well. But these are all of a mass; the men, their officers and generals do not necessarily split them into different categories.

Furthermore, of course, all these ideas and conceptions interact. I might be confident in the performance of my long bow, but nervous about the fact that my sergeant is incapable of hitting a barn with his bow, or that there seem to be an awful lot of Frenchmen about wearing heavy duty armour. Thus, my confidence, which might be high in one sense, could be low in another, and dependent on my own mental outlook and physical wellbeing (the fact that I have dysentery might add or subtract from my confidence). In short, there are all sorts of factors making up the outlook of one individual and, of course, many individuals whose outlook is a factor both in my individual outlook and in some emergent outlook of the unit, or the army, as a whole.

There is no way in which this can be handled as a whole, I think. In order to make a set of wargame rules work, we have to reduce this mix of personal outlooks, physical capabilities and so on to a set of models which reproduce certain aspects of the whole. Thus, I have a given model for the effectiveness of the bow. I have another model for the impetuousness for the French knights. I have yet another model for the use of crossbows in a skirmishing role, and probably another for the effectiveness of crossbowmen ridden down by those impetuous knights, and so on.

In short, what I have done is taken all the formally distinct bits of the world of the archer before Agincourt, and turned them into separate models. The thing is, though, that those models are only formally distinct. How well the ‘real’ archer shoots might well be a function of whether he thinks his weapon is effective at a given range against a given target, as well as his state of health, how confident (or scared of) his leaders he is, and so on. The different models in fact represent different bits of reality that we can think about separately. Reality itself is unitary.

The next step is to model these things as arithmetic calculations. This introduces another level of abstraction. The simulation is now grainy. Reality tends to be more gradual, more graded than simply reaching, or not, a numerical threshold. As wargamers we have to accept this as a consequence of attempting to wargame at all. Although there are sudden collapses of units, usually this is the cumulation of a set of circumstances and events in the recent past. While a sudden bad dice roll might reflect this, it also washes out that slow disintegration of the individual and communal morale.

Now, we attempt to put these models together in some sort of order, to reflect the reality which they are trying to simulate. But note that we have, in fact, changed things considerably. The continuous and fairly smooth behaviour we witness in real life has been replaced by a step-wise and clunky function depending on dice rolls, among other things. The integration of a person, their environment and world view has been replaced but a set of models for each of them, hopefully statistically derived for a mass of people. And so on. The models we use only reflect some formally distinct aspects of the reality we try to model.

We do, of course, gain some advantages. The models we use are tractable and intelligible. If a unit runs away we can give reasons, which the real life equivalent might struggle to do. But we do need to be aware of the things we have to sacrifice to obtain that intelligibility and tractability. 


  1. A friend and I had a joke format for a super-grand-tactical level wargame. You spend about 2 hours setting up the armies, in considerable detail, then you say something along the lines of "red army has better chance of success, but not by much" (even better, you get an umpire to say this), then roll a die - 3 or better and red wins. Then spend 2 hours discussing in considerable detail why the great battle went the way it did. The heart of our joke was that we reckoned this concentrated on the enjoyable bits of the game, and usefully "abstracted" all the clunky bits of the rules which didn't actually work all that well, and which gave rise to all the argument and bad feeling.

    Board games do this all the time - one die roll and a big army counter is removed - battle over - campaign moves on. On numerous occasions in board games I have regretted the lack of visibility of these actions - often strategically crucial actions - but the need to keep the game moving means we have to sacrifice a lot of detail. At the other end of the spectrum, I have also found, over the years, that the more detailed a set of rules is, the more "realistic" the grip on the blow-by-blow narrative, the sillier and more obviously artificial the game becomes; the closer we study the action under a microscope (however worthy or scientific our intention), the more apparent is the clumsiness of the imitation.

    I have little experience of skirmishes - which represent a different activity altogether - but my preference for big, somewhat abstract games in which you can see the movement and the development (rather than argue about morale bonuses) has a lot to do with this. As you correctly state here, the more we try to subdivide the models (and submodels?) the more we offend against the fact that they are not really independent at all, and the more distortion we introduce. The 3+ die roll giving victory to red is not a lot of fun, but it is possibly less daft than some of the exceptionally clever things we have tried to do instead!

    Excellent post, as ever - thanks.

    1. Thank you.

      Yes, of course, we could just throw a dice and move on, but then what do we find that is engaging about a wargame. Some people seem to find the arguing part of the reason for wargaming at all (I've never understood that one), but there must be something at a lower level than the abstract that draws us in to spending hours painting, modelling, setting up and playing. I guess the narrative is the thing.

      Within the narrative most things can be explained, or explained away, either at a game level ('bad dice rolls') or at the action level ('good shooting'). But that is then part of us telling a story about the game, or the game world, or both. And this can get confusing when thy are muddled.

    2. I realise that my focus on the "level" (of abstraction; of detail) in the game is getting rather away from the point of your post, but in moments of quiet pondering I sometimes (quietly) ponder whether there is a correlation between the scale of the figures and the rules approach - even subconsciously. I've been re-reading the Horse & Musket period book for "Black Powder", which is great fun (haven't played the game, but the book is fun), and I wonder if there is a correlation between the style of figures and the comfortable level of abstraction. Black Powder is stuffed with fine photographs of detailed (caricature) 28mm figures, each of which obviously has a personality and (if you wish) a personal life story, and it would seem inconsistent to run such a game at too high a level of abstraction. 6mm is different, and card counters are completely different.

      Just a thought. I found it difficult to imagine someone using 6mm for Black Powder, though I guess it is quite possible. One digression on the subject of Black Powder - painted casualty figures - am I the only person who thinks it is funny to see units marching about trailing strings of dead bodies behind them? Like the tin cans behind the wedding car.

    3. I think there is such a correlation, yes. The bigger the figure, the more individual the rules, it seems to me. of course, I've not done a scientific investigation (most rules are too expensive for that), so it is just an impression.

      mind you, most, for example, Samurai games go for individual heroic action. Most reports of Samurai battles indicate the officers were the samurai and the others did most of the fighting (and dying, unless your head was cut off as a trophy).

      So I guess it comes down to what sort of reality do we wish to portray?

  2. All very true and it shows some of the difficulties and pitfalls of the bottom of approach. A bit like trying to understand human behaviour by starting at tge cellular level.

    A few years ago I started trying to approach wargames from a top down approach, starting by trying to recreate typical group behaviour as reported rather than starting by trying to understand the individuals. It has its own pitfalls and does not come naturally, neither does it explain why history unfolded the way it did. However, at least one can compare games to the range of recorded sample events that we chose and we are able to say that game events do or don't but fall within the range of possible recorded events that the game was based on.

    Obviously a sample of 1 event doesn't allow any variance so makes a poor game so we end up grouping events but the closer the events are, the better. So basing a game just on Zama allows no variance, basing the rules on all of the recorded gives us a range of possible outcomes and some patterns, bading them on 1,000 years starts to look at very generic patterns, blurring the differences between various wars.

    Different approaches delivering different things that look similar.

    1. I think there are two approaches here, a top down and a bottom up one. But I think there are other issues as well, such as what level do you want the rules to be about?

      this seems to me to be an issue, in that most rules make the wargamer successively army, division and unit commander. With PM:SPQR I tried to make sure the wargamer stuck to being, as much as possible, the army commander.

      But, battles are one off events, and i agree that we can't base a rule set on one incident. But how many incidents makes a good statistical ensemble?