Librarians & Leviathans: On variable skill resolution

So, I was doing some very tenuous rambling about skills.

For reference, "skill" here means anything that behaves roughly like a "skill" as in various major RPGs. "Attempt" means either an attempt in the traditional sense, the consumption of one arbitrary amount of resources, or an arbitrary amount of time and effort. "Roll" means whatever mechanic is used for determining the effectiveness of a skill. "Resources" include time, and this is probably the most common resource.

So last time I rambled a lot about the different kinds of situations skill use represents, and some awkwardness that results if you treat these situations as mechanically equivalent.

So, what kind of distinctions might it be useful to draw, if we're looking for skill use to roughly reflect likely outcomes in real life?

Is this an activity that attempts to change the status quo, or one that determines which of two (or more) possible situations arises? If you're trying to unlock a door, failure means the door remains locked and you can potentially try again. If you're trying to convince a police officer that you're from Scotland Yard, either they believe you're from Scotland Yard or they believe you're a liar - there's no neutral ground. In many cases, when you try to do something, a failure means the status quo continues. Sure, you might pick up a bruise or waste some fuel, but the range of possibilities is essentially unaffected.
Does the attempt itself pose a risk, other than (as in point 1) leading to a situation you don't want? Trying to win a game of Patience is basically safe. Trying to retrieve a key from a tank of piranhas is not, and each delve (or each arbitrary unit of time spend there) poses an additional risk.
Does the task you're attempting include the test of your success? To put it another way, can you reliably gauge your success at the task when you've finished it? If you're trying to swim across a lake, resolving that attempt also resolves the outcome (whether you reached the shore, how fast your swam, and so on). The same applies to climbing a cliff or opening a door. On the other hand, if you're trying to fix a hyperdrive engine, you'll only ultimately find out how well you did when someone tries to activate it. There is a distinction between skill applications where the test is intrinsic, and those where a test is separate and may never materialise.
Will you (provided you're willing to expend a realistic amount of resources) eventually succeed at the task, or is there no guarantee that you can succeed? Most people are capable of walking to another city 100 miles away if they've got the whole summer to make the journey, but the time taken will vary greatly by fitness. However, not everyone can build a functioning hot air balloon in that time.
Are the resources consumed by an attempt negligible? If it takes 1 second to try and pick up a £100 note, you're not counting the cost of any plausible number of attempts. If it costs £1000 per go on the coconut shy, you will pay careful attention, no matter how much you want that toy albatross.

Skill use taxonomy

I made several attempts at creating a decision tree to usefully subdivide skill applications. This is what I ended up with.

The first question is point 1: does this attempt alter the possibility space, or just attempt to change the status quo?

Fork in the Road

Regardless of the outcome, the attempt establishes a new set of possibilities. It's not possible to try again.

Fork-in-the-road applications are the other kind. The result of the skill use determines the new possibility space, but on a failure you are not left in the position you started in. Perhaps you're trying to hotwire the jump pod before toxic fumes overcome you. Or approaching a fork in the road, and trying to wrench your runaway carriage to the left (into the spooky forest) rather than the right (over the rickety bridge).

In a fork application, you only have one attempt to achieve your aim. Whether you succeed or fail, trying again is not possible. The result of the roll determines the consequences of your attempt, and also the new possibility space.

Here, it makes sense to me to use a one-off roll. It's the character's skill in this specific instance that is relevant.

Now, let's go back to those status quo-friendly skill applications. All the remaining applications are of that type.

Let's begin by considering non-risky applications.

Next question, how well can we gauge the success of our skill application at the time? Let's look at those where we can't judge it.

Magic Circle

A failed attempt makes no essential difference to your options, so you can try as many times as you want.
You can't judge how successful you were until something happens to test your work. Therefore, there is no guarantee of eventual success.
There is no risk.

In some cases, it's not possible to immediately judge how successful your attempt has been. It may be that the results take some time to manifest, or will only show up when your work is tested. In other cases, you don't know whether you're getting accurate feedback or not - this is particularly the case with social interaction, when you don't necessarily know what someone else's real attitudes or beliefs are.

You could maybe sum this up with the idea that the task you're trying to accomplish is separate from the test of whether you were successful. This allows you to retry the task (given adequate resources) without the consequences of success or failure manifesting; however, it also means you have only a limited idea of how successful your attempt was.

In these cases, it's only the last attempt that matters for resolving the success of the attempt. Imagine you're trying to make a magic circle to bind a summoned spirit - try as many times as you like, the only time you'll really know whether it works is when the spirit arrives.

It doesn't make a lot of sense to make multiple attempts in a case like this, but if that does happen (and in real life, it does) we should track resources used separately if we care about that. We might feel that some failed attempts should be noticeable to the character; this is going to vary by situation and genre, as in other cases we might feel that a failure represents not only wasted resources, but also a failure to notice the mistake.

This is another case where it makes sense to use a single roll to determine the success of the venture. It's the character's skill in this specific instance that is relevant. For game purposes, I'm inclined to say that it's best to handwave hypothetical failed attempts that were noticed by the character as just part of the one roll.

Okay, and now those situations where we can judge our success.

And finally, is eventual success inevitable in this situation, taking a "plausible and realistic" view of inevitability, rather than an irritatingly pedantic definition?

Patience

A failed attempt makes no essential difference to your options, so you can try as many times as you want.
You can tell how successful you were after each attempt (or during it).
There is no risk.
Given enough attempts, you should succeed at this task.

In this case, the task is well within the character's skill. The limiting factor is the resources available.

Some tasks really just call for effort. Perhaps a hundred thousand books need to be moved from one room to another. An average person is perfectly capable of doing this, it'll just take a long time and be very boring. There is no point rolling to see if they succeed provided the resources are available. They just do.

This situation becomes interesting when resources are limited. Can you do X with the time/resources available (typically, before Y happens)? Or, how much of X can you achieve in Y time?

In this case, I think it is interesting to test your skill to establish what you manage to achieve. However, you also probably want the achievement to be logically appropriate to the resources expended. If you have an hour to move paperbacks, it's nonsensical for a normal human to roll 1d100x1000 to see what they achieve in that time, but it might be reasonable to roll 1d10x200. However, in circumstances like this where degree of accomplishment is important, you probably will tend towards your average performance on a task, which makes me feel that a dicepool might be a better model. You're not really likely to move only 100 books in an hour, or to work at absolute maximum efficiency throughout. Something like, say, 10d20 is probably a decent bet.

This application comes off the character's average ability.

Four-Minute Mile

In these cases:

A failed attempt makes no essential difference to your options, so you can try as many times as you want.
You can gauge your success to a substantial extent; most importantly, you can tell whether you failed. This allows you to refine and improve with multiple attempts. The relevant point here is not improvement over time, though, it's that you aren't going to abandon a successful attempt and replace it with a failed one.
There is no risk in the attempt, other than wasted resources.
The resources used are negligible to you in your current situation. You have plenty of time (and are willing to spend it), costs are too small to be worth tallying, your batteries will outlast the sun, and so on.
There is no guarantee that you are capable of succeeding.

Essentially, the limiting factor in your achievement here is your ability. You will eventually do as much as you are capable of, perhaps with a slight wobble caused by sheer luck.

An example might be learning a card trick or piece of music, picking a lock, or trying to run a four-minute mile.

Imagine you're trying to run a four-minute mile. You have no particular deadline to meet this goal, and it doesn't cost anything you care about. Failing doesn't prevent you from trying again. You can judge your performance on any given run. As running a four-minute mile is difficult, you don't know whether you're personally capable of achieving it.

It makes sense to me here that a randomiser isn't really necessary. If the character has the ability, they succeed. If not, they don't. If it's non-binary, and a matter of degree of achievement, then the character's ability determines their degree of achievement. There can be some breathing space around this idea, to account for additional assistance they might obtain from other characters or resources (better running gear, steroid injections, coaching) that effectively increases their ability.

If, however, the resources are non-negligible, then the skill use models the resources used to reach that final level of success.

This application comes off the character's maximum ability.

Okay, now let's go back a couple of branches and look at risky applications.

How well can we gauge the success of our skill application at the time? What if we can't?

Piranha Magic Circle

A failed attempt makes no essential difference to your options, so you can try as many times as you want.
You can't tell how successful you were until the the outcome is tested.
Each attempt poses a risk (but doesn't alter your basic options).
There is no guarantee that you personally are capable of achieving this task.

This is essentially the same as the Magic Circle, except that each attempt poses a risk. You are, for example, making a magic circle at the bottom of a tank of piranhas. Each time you decide to redo it, it's another chance for the piranhas to bite you.

In these cases, the success of the attempt depends only on the last attempt you made, as with the Magic Circle. However, the risk you're exposed to depends on how many attempts you make, as do the resources consumed. Since it's risky and you have no way of knowing whether one attempt is better than another, it's not very logical to try more than once. That being said, there's a reasonable argument that if it's really bad you will probably notice.

This is another case where it makes sense to use a single roll to determine the success of the venture. It's the character's skill in this specific instance that is relevant.

And back to applications where we can judge our success...

Is success inevitable?

Piranha Patience

This label applies to cases where:

A failed attempt makes no essential difference to your options, so you can try as many times as you want.
You can gauge your success to a substantial extent; most importantly, you can tell whether you failed. This allows you to refine and improve with multiple attempts. The relevant point here is not improvement over time, though, it's that you aren't going to abandon a successful attempt and replace it with a failed one.
Each attempt poses a risk (but doesn't alter your basic options).
If you are willing to spend the resources, you will eventually succeed.

Imagine a very, very long holiday where you're playing Patience to while away the time. Your objective is to win one game. Winning or losing doesn't make any difference to whether you can play again. You can tell when you've messed up and can't win any more. Given long enough, you will succeed (and, if you care about degrees of success, assume that we're tracking the moves used or the time taken).

Except that lacking any cards, you're inexplicably using live piranhas as cards, and every time you reshuffle them to start a new game, there's a chance you'll get bitten.

In this case, we know we can succeed, but we care about the number of attempts that would be necessary to reach a particular threshold of success. This will establish the risk undergone, and we probably have some risk/injury threshold where we would cut our losses and give up. It's also possible that if we can't succeed after N attempts, we would lower our target threshold of success because of the constant risk - maybe winning a game in 200 moves is fine after all, we don't need 100.

This is a bit fiddly. We might want to do something like:

Establish how many attempts we would take before succeeding.
Establish how much risk would materialise (e.g. how many hit points would actually be lost) before that stage.
If this cost is too high, establish the point where we'd cut our losses and accept a lower threshold of success.
Depending on the nature of the challenge (does each attempt replace the previous one, like hairstyles, or are they parallel, like making cakes or lap times?) we may need to start again and aim for the new threshold, or simply accept a previous attainment.

And very finally, applications where we might fail.

Piranha Four-Minute Mile

This label applies to cases where:

A failed attempt makes no essential difference to your options, so you can try as many times as you want.
You can gauge your success to a substantial extent; most importantly, you can tell whether you failed. This allows you to refine and improve with multiple attempts. The relevant point here is not improvement over time, though, it's that you aren't going to abandon a successful attempt and replace it with a failed one.
Each attempt poses a risk (but doesn't alter your basic options)
There is no guarantee that you personally are capable of achieving this task.

In this case, though, you're running in a swamp that contains piranhas. Yes, ignore all the other questions that raises. The point is, each time you try to run, you're risking getting bitten.

Your ultimate long-term achievement is limited by your skill, just as in the Four-Minute Mile above. However, it's also limited by how much risk, and how much actual pain, you're willing to endure. As attempts pile up, you may be willing to lower your target for success (say, a five-minute mile) because the threat of piranhas is weighing on your mind and the original goal begins to feel unrealistic. This will partly depend on the severity and probability of the risk - a small risk of getting a tiny nip is much less discouraging than a large risk of losing a limb.

Semi-blind applications

In some cases, it's impossible to distinguish a lack of skill from an impossible task. The best example is attempts to find something that may or may not be there. Did you miss the secret door, or isn't there one? Is this an unbreakable code or a random string of letters? Did you fail to contact the aliens, or are there no aliens?

You can think of this as a paired set of choices. Is X a meaningful achievement, and can I achieve X? Is there a secret door, and can I find it? You will only receive a definite answer if both cases are true. Or, of course, if you find another way to establish the truth of the first statement (if you have reliable evidence of a secret door, for example).

What we have is limited knowledge of failure.

In the case of knowledge tasks, you'll be aware of ignorance, but not of misremembering.
In the case of deduction or interpretation, you'll be aware of bafflement, but not of misinterpretation.
In the case of social interaction, the feedback you get depends on the person and the type of task. You'll know if they refuse a favour, but not necessarily whether they swallowed a lie.
In the case of physical tasks, you'll know whether you manage to break something, climb something or resist the cold, but not necessarily whether you picked up a disease.
In the case of creative tasks, you'll know whether a piece of art looks right to you, or if you have managed to build a car, but you don't necessarily know what other people will think of the art, or whether the machine will work.

I don't think this is especially going anywhere right now, I just thought it was a bit interesting.

Suggestions

So I'm not going to suggest all games everywhere should do this, but it seems like it might be interesting to model skills in different ways for different applications.

Let's assume a system with skills ranked from 0-5.

Any fork-in-the-road, magic circle, application would use a single d6 roll-equal-or-under compared to the character's skill.

Any patience application would instead use a dicepool. Roll one die for each point in the skill. 4+ is a success.

I imagine there is some crippling mathematical flaw or other in this, but I don't have the energy to check. Either way, I find the idea of distinguishing one-roll and dicepool skill uses strangely appealing.

Any four-minute-mile application wouldn't roll at all. Just compare the character's skill to a threshold of difficulty for the task. If they'd pass, use a dicepool as above to determine how many attemps are needed.

And as yet I have no idea how to handle the risky applications. But I already spent ages on this post, so hard cheddar.

Wednesday 12 March 2014

On variable skill resolution