Saturday 14 April 2012

Writing Rules

Writing wargame rules is, when you come to think of it, a very odd pastime indeed. On the one hand you steep yourself in the historical evidence, whatever that may be, or at least read a popular book or two on the subject, and on the other you try to reduce what you have read to a single set of cogent guidelines for reproducing it on a table top with toy soldiers.

It is little wonder that it is a more difficult task than may immediately strike the consciousness. On the other hand, the fact that it looks easier than it is is probably a good thing, or no-one would embark on such an undertaking, and we would not have the variety of rule sets that now exist.

Rule writing does, however, possess a deep problem which is, I think, insurmountable. As a human race, we like to detect patterns. If something happened this way in one instance, and then another, we predict, often accurately, that it will happen thus in a third.

This, for those of a philosophical turn of mind, is nothing but the problem of induction, as outlined by the Scottish Enlightenment philosopher David Hume.

The problem, Hume argues, is this: The sun has come up on every day of my life so far, and, therefore, I expect the sun to come up tomorrow.

How can I justify this belief? Well, Hume says, I cannot. Just because the event has happened every day of my experience so far, it does not mean it will happen again. There is no proof that it will happen the same tomorrow, absolutely none.

Now Hume was no fool, and he recognised that induction is something which we use every day, many times a day. Can you imagine the chaos that would ensue if induction did not work? If, every time you turned the wheel of your car left you were uncertain whether the car would turn left or right?

Whatever its faults, induction does work, which is probably just as well. However, Hume’s point remains, that we cannot justify the use of it. This has been, and still is, a problem for the philosophy of science, because science relies on induction to work.

In science, we make an observation of A, and we observe that associated with A is always a B. We make a sizeable set of observations, and every time we see an A, we see a B too. And so we declare a law of nature – if you see an A, a B is there as well. And so science progresses.

However, if you look closely at the structure of that argument, you should see Hume’s problem. Actually, we have proved nothing, except that there is a certain, probably (or hopefully) rather high, probability of seeing a B if you see an A. Without getting into complexities, we are relying on something called Bayesian probabilities, which allows us to use induction to get the best explanation, and accounts for increasing amounts of evidence improving the odds that our explanation is right.

So, what has this got to do with writing wargame rules?

Well, we are in a far worse situation with wargame rules and historical accounts than scientists are with induction. A scientist can always go and repeat the experiment, while we, as wargamers, cannot.

What we are left with is a series of contingent events and no way of working out how likely they are to be repeated in a similar situation. Given that each historical event has its own context, even if we had sufficient data to directly compare events, we would still be struggling because the contingent contexts are different.

Now, history is all right on the battlefield. What I mean is that no-one bothers to calculate the probability of this volley of musket fire turning out all right. They just shout ‘fire’ and hope it goes well. If, by some low probability event, all the musket balls strike home and disable an opponent, the commanders are not going to stop and explain how remarkable it was. They are simply going to shout ‘charge’.

The problem for us as wargame rule writers is that these probabilities are important to us. A set of wargame rules is exactly a means of writing some laws to govern these contingent outcomes. We are relying on induction to say something along the lines of ‘a unit of fresh musketeers shooting at a unit of militia will, on average, have X result’, and we then adjust the average by a random factor to make life a little uncertain.

The problem is, of course, Hume’s. We have no justification for doing this and, indeed, cannot. The empirical evidence is simply not there. On how many occasions did a unit of musketeers fire on one of militia under equal, controlled circumstances and recorded the outcome? I’ve not done an archive search, but at a rough guess I’d say, with, I think a reasonably high chance of being right, precisely none.

Thus, we have a major problem in rule writing. We cannot know the probabilities, and so we cannot base rules on empirical evidence. We can only work with what we have and hope that our mechanics give some sort of acceptable result.

In short, in writing wargame rules, we have to have some a priori concept of what we are about and what the probabilities of the outcomes should be, even if these are not explicit in our thinking. We expect, rationally, even logically, that a trained unit of professionals should, all other things being equal (and there is the catch), defeat a unit of hastily raised militia.

But the point is that this is something which we apply to the world, an expectation drawn from how the world works, in general. Providing specific evidence for it is hard, even impossible, so ultimately we cannot base our rules on empirical evidence.

And even if we do manage some evidence, Hume’s argument about induction still bites us.


  1. Empirical rule
    Empirical rule, empirical

  2. "On how many occasions did a unit of musketeers fire on one of militia under equal, controlled circumstances and recorded the outcome? I’ve not done an archive search, but at a rough guess I’d say, with, I think a reasonably high chance of being right, precisely none." Probably not too far off the mark (unlike our musketeer!)

    The only full-blown modern attempt to work-out the effectiveness of black powder arms (as far as I'm aware*) is Firepower by BP Hughes. And he relies on a relatively small number of incidents, and some of those he relies on "factoring out" the effect of say artillery in order to calculate the effectiveness of musketry. The clearest cut piece of evidence is when two Irish regiments conincidentally came face to face at the Battle of Malplaquet (one in the French service, one in the English) in an isolated area of a forest.

    * of course I could be suffering from an Anglo-centric view of the world not being aware of anything in other languages (but surely Duffy would have referred to them if there were any?)

    So (for this specific period) we're left with going with what you feel is right based on "inherent military probability", to quote a phrase quoted by Brig Peter Young in the Wargame. I suppose IMP is the nearest thing to induction in this context!

    Players can always tweak the rules if they disagree and who can say what is right or wrong.

  3. I think you are right, in that all we are left with is 'it feels right' and some sort of 'inherent military probability' type thing, which Burne and Young were fond of (and which, incidentally, appears in Featherstone as well).

    The problem is that we are relying on a very low number of incidents, a few experiments of dubious quality and the intuition of some retired military men from a different war.

    On the other hand, it is better than pure guesswork...