Thursday, 6 December 2012

On the Philosophy of Randomizing Tables

Earlier today I was working on a new set of random tables (a random mercenary eunuch war-band generator, if you must know) and it occurred to me that there are two basic approaches to the creation of random tables, and they are as follows:

  • The realist approach
  • The instrumentalist approach
Under the realist approach, the creator sets out to create a set of tables which will generate broadly realistic results (obviously). Under the instrumentalist approach, the creator sets out to create a set of tables which will generate immediately and certainly gameable ones. 

Another way of putting it: When the random table was created, was the first concern:

a) Generating random results which, though clearly random, are nonetheless broadly reflective of "what might happen" in reality?


b) Generating random results which are "fun" and interesting for the purposes of getting the PCs involved in hijinks?

One easy test for determining whether the approach is realist or instrumentalist is whether any attempt has been made to assign some sort of distributional characteristic to the results. A classic example of a realist approach is the method used for creating a random encounter table in AD&D 2nd edition: you have a set of results from 2-20 as follows:

2 Very rare
3 Very rare
4 Very rare or rare (DM's choice)
5 Rare
6 Rare
7 Uncommon
8 Uncommon
9 Common
10 Common
11 Common
12 Common
13 Common
14 Uncommon
15 Uncommon
16 Rare
17 Rare
18 Very rare or rare (DM's choice)
19 Very rare
20 Very rare

And you roll d8+d12 to get your results. This creates a situation in which very rare monsters are encountered very rarely, while common monsters are encountered commonly. The results are realistic in some sense: this is important, because it makes sense that dragons are only encountered on special occasions whereas goblins are ten-a-penny - it would be unusual to have an equal chance of encountering both.

Classic examples of the instrumentalist approach are what are generally found at The Dungeon Dozen. For example:

In the Saloon  
1. Depraved cretins w/strong sense of entitlement (2d4)
2. Tavern sage holds down corner of bar: answers simple questions for a drink, buy a round for the house for more complex inquiries
3. Surly drunks embittered by years of being surly (2d4)
4. Some guy who's really loud and thinks he's hilarious
5. Raucous gaggle of pickpockets emboldened by drink
6. Black lotus addicts waiting around for their connection to show, rather edgy
7. The guy who has strident opinions on anything he happens to overhear, not a particularly deep thinker
8. Pack of ruthless, armed-to-the-teeth dwarfs celebrating successful delve
9. Inebriated laborers fomenting uprising, much speechifying and little regard for alternate opinions
10. Tattoo artist plying trade in well-lit corner: save vs. infectious diseases, heavily inked sycophants openly question the machismo of the un-inked
12. Off-duty assassins amusing themselves by subtly pitting various patrons against one another then sitting back to enjoy the ensuing mayhem
I know from personal experience, as I am sure you do too, that in a given pub you very often encounter 1, 3, 4, and 7, and very rarely 2, 8, 10 or 12; but here there is an equal chance of encountering any of them. Why? Because it is fun for the players to encounter 2, 8, 10 or 12, and they provide hooks for potential adventure - why make such potential gold a rare event?

There are horses for courses, so both approaches are more or less appropriate depending on the situation; but I've discovered that it is worth thinking about the implications before sitting down to write. Do I want my random eunuch mercenary war-band generator to be realist or instrumentalist? And what are the implications of the two approaches? These are actually non-trivial questions. 


  1. I would prefer it if you created an instrumentalist eunuch mercenary warband table, but that is to be expected since I am a Marxist which means I am inherently instrumentalist. But seriously, when I create tables as examples (in the same sense that books of saints are sets of examples) I work from an instrumentalist set of assumptions: let's show people what's interesting about X. When I am creating tables for others to use - let's say for character psychoses - I am still going to make every example interesting, but the ones at the extreme of the distribution will be more over the top.

    1. Yes: the unwritten rule is that, whatever approach you use, you have to make the results interesting or else you are doing it wrong.

  2. Reminds me of something Gary Gygax said about wargames before writing Dungeons and Dragons. Some people want to recreate the actual scenario of the battle, with as much detail and "realistic" results as possible. And some people are more interested in the odd "possiblities" to see what sort of game will pan out. It is a spectrum, of course.

    1. Sure, but while I think the personality of the designer is important, I think the purpose of the table is what should be considered, first and foremost. Which approach you use ought to depend on what your aim is.

  3. I think that to some extent it depends on how much use you expect the table to get. If you're going to be rolling on a table frequently (say, a random encounter table for a hexcrawl campaign), the relative frequencies of different entries will wind up giving players a sense of what the campaign world is like, and they will learn what they can generally expect (and the rare entries will be all the more memorable when they come up). If you won't be rolling on a table that often, then there won't be much opportunity for the players to get a sense of the probabilities involved, so there's not much point in going for a realistic distribution.

    I also think the 2e random encounter method is useful not only for modelling "realism" but also for varying difficulty. Having a range of probabilities makes it easier to include encounters that are easier or more difficult without having constant wild swings in difficulty.

  4. Realist ones are better for tables you use a lot, instrumentalist are better for things you only use once in a while.

    1. Exactly; if you have a load of these instrumentalist tables that cover extreme events, but you only pull them out occasionally, then you are functionally creating the same kind of probabilities.

      It's just that instead of "when the dice rarely roll that thing" it's "when you rarely roll on that table at all".

      It strikes me then that there is a hidden instrumentalist purpose in realist tables; you want a good go-to table for normal life, but you don't just want it to be the "things go as expected" table. So bringing in some weird stuff as edge cases means that when you choose to roll on that table, you're not choosing to exclude the others.

      My lazy version of this is add a few "roll on another table" entries, but I think that's probably actually a bad solution for handling time reasons.

  5. Would it be really stupid to leave a bare column next to the one with the dice numbers so people can write on thier own distribution? Or to provide both alternatives? They could cross out the other one with pencil.

  6. @patrick
    The "cross out and write your own" thing is actually a compromise solution.

    It allows for the instrumentalist approach without the monotony of "oh, a chameleon dog on fire with extra knees _again_ "

  7. Realist is dangerous, it leads into the trap of thinking you can actually use the tables to simulate the environment. That way lies madness, IMO, if used on any but the smallest scales. OTOH purely instrumentalist tables can be unsatisfying if they don't really give sense of the locale, and often each of the entries can only be used once.

    Probably a mixed approach is best - I think of it as "simulating the daily life of adventurers, who naturally attract more than their share of trouble".

  8. You could probably extend this spectrum to other kinds of rolls. People should prefer bell curves for things that they feel should more or less go as expected. And straight flat distributions for thing that they want to provide more randomness. I've never seen a realist spell fumble table, for example, but this can extend to things like realist skill checks and instrumentalist combat (e.g. SWN).

    One idea: you can use both 1d20 and 2d10 to make a table with 20 entries, and just put the normal entries in the center with the weirdest entries on the periphery. 3d6 works, too.

    This also works for 1d12/2d6(/3d4?). The table can be realist or instrumentalist depending on the situation. 1d12 for Weirdbeard's Opium Barge and 2d6 for the Queen's Petticoat Tavern.