I spend probably more time thinking about skills than is good for me, but this is something I touched on previously and would like to discuss a bit.
Essentially, I've been thinking that (for understandable reasons) games tend to take a one-size-fits-all approach to "skills", and that this can end with some unsatisfying results. Call of Cthulhu is my core example here, but it's not unique to that game. The same mechanics that adjudicate whether a single bullet fired from your handgun hits a crazed cultist also determine whether your attempt to eavesdrop on an hour-long conversation between two Russian spies results in a) absolute and flawless understanding of every nuance of the conversation, or b) a glazed expression and a slight headache.
There are some differences between these two situations that aren't reflected in the single d100 roll you'd typically make. One is that using a universal resolution system ironically results in widely different ways of resolving outcomes, because of the way the challenge is modelled.
The handgun roll is unlikely to be the one and only die roll used in the Crazed Cultist scenario, because the scenario itself is much broader than "shoot at a cultist"; more likely, we're looking at a situation with interplay between a number of rolls and non-rolled actions, forming an overall scenario along the lines of "do we overcome a crazed cultist". This adds a large degree of variability to the overall outcome. In contrast, the eavesdropping Russian Language roll is likely to be the only one resolving the question of "do we learn anything from the spies", making for a binary outcome.
This contrasts oddly with reality, where bullets either hit or miss a target, while varying degrees of language comprehension form the basis of entire grading structures and not a few plotlines.
Breaking down skill applications
Off the top of my head I think there are three distinct skill applications I want to consider, but there may be more. For clarity, I'm thinking here about real-world accomplishment rather than how they're modelled in games.
Imagine firing at a cultist. Either your bullet hits the cultist or it does not hit the cultist. Of course, some hit locations will be more advantageous than others, but simply establishing a result one way or the other is adequate as a baseline for resolving what happens. Moreover, this is an all-or-nothing outcome: one bullet, one outcome. The timescale, and the immediacy and importance of the results, mean there is no ongoing feedback and improvement (unless you're doing some kind of time-travel reality-warping game, of course).
You could, of course, keep firing at a target. In a shooting range or something, rather than tracking individual shots we might want to measure overall success. But that would turn it into a different kind of application, probably the next one. In a combat situation, the moment-to-moment success of each shot is important because it creates possibility branches in terms of what other characters might do.
Let's say you're working through a set of maths exercises; you have the answers available, but not the methods; your objective is to reach the answers by calculation in order to improve your maths. How very studious of you! And you're allowed to use reference materials if you want. Your initial success at the task can vary across a very wide range, but more importantly, you can redo your work endlessly until you get it right, limited only by your overall ability and the time available. Getting an answer wrong initially doesn't put a stop to things.
Similar ideas might apply if you're trying to draw something, practicing a piece of music, training on a climbing wall, and so on.
It seems to be unconsciously assumed that skill applications are attempts to succeed. However, in some cases they are more like attempts not to fail. This is particularly true when groups come into play. If you're trying to sneak around, one person stepping on a creaky floorboard harms the whole group. If you want to infiltrate a building, one person looking insufficiently non-suspicious can blow everyone's cover.
This contrasts with something like, say, moving rubble. Each person's contribution to rubble-moving adds on to everyone else's.
It also partially contrasts with climbing cliffs or resisting thirst, where each person's success affects them directly, with only secondary effects on any companions.
Covert skills are probably the queen of these applications; how many dungeon-crawling parties have been undermined by one PC in rusty plate armour failing a Move Silently roll? This is kind of interesting because it comes down to consequences. In a much more direct way than most other applications, the stealth skill is a Three Musketeers situation, where if anyone fails their roll, the whole party tends to take the consequences.
Some skills come in useful all the time, either in reality or in games. Good vision is much more convenient than near-blindness. A persuasive way with people is often handy. An excellent knowledge of bartitsu, however, or juggling, or astrophysics, is of limited day-to-day use unless your lifestyle regularly calls for those skills.
In games, this is very common because some skills simply crop up much more regularly than others. Perception, Persuasion and Stealth tend to see an awful lot of use in most games (depending on genre, of course); Biology, Forgery and Pilot [Balloon] are pretty rare.
Some skills are only rolled a small proportion of the time they're relevant - vision being an excellent example. They're used mechanically only when something of particular significance is happening. Call of Cthulhu, for example, suggests not rolling on Drive skill most of the time. Some skills are rolled every time they might be relevant - typically those with immediately serious consequences, such as shooting skills. A hazy middle ground features skills that are often handwaved if the PC has skill at some arbitrary threshold.
Rare skills present problems for a couple of reasons. If you have invested significantly in a skill that's used once in ten games, it can be very frustrating to then fail that roll - far more so than if you purchased a more common skill that you're likely to use more often, where overall competence will tend towards your actual skill. Because costs are typically the same for rare and common skills, this makes it mechanically advantageous to buy skills that are mechanically common. Rare skills also create a tension for players, who are forced to decide whether they want to invest in a skill that's unlikely to manifest very often. Interesting skills can lead to interesting characters, or creative approaches to problem-solving.
One aspect of the rare skills issue, and the increased significance of each roll, is how GMs handle these skills. If a PC does have a rare skill, then it becomes more important that they get to use it whenever possible. Conversely, if no PCs have the skill, then it's relatively common for the GM to work around this whenever the skill would be relevant. A strange outcome of this situation is that it can be mechanically beneficial to actively avoid taking rare skills.
Threshold use can produce some statistically odd results. Taking 10% in a skill is likely to mean you succeed at 10% of attempts. Taking 50% in a skill may well mean you succeed in 80% of attempts, because the GM rules that you don't need to roll sometimes.
Back to theory
So at the moment I'm inclining towards the idea that there are several two-way splits in skill applications. These include (but are not necessarily limited to):
- Positive vs. negative accomplishment (does success accumulate or does failure detract?)
- Action vs. process (is it a single step, or a sequence of actions with potential for ongoing review and adjustment?)
- Qualitative vs. quantitative consequences (does it change the possibility space or just affect the amount of resources used?)
- Repeatable vs. non-repeatable (can you meaningfully re-attempt the task or is it essentially a new task?)
- Occasional vs. always (how often do you roll the skill for that kind of application?)
For example, let's go back to our cultist-shooting. This is positive because each person's success adds to the success of the others in Having Shot The Cultist. It's an action with no possibility of adjustment as you go along; you fire and see what happens. The consequences are qualitative because a wounded or incapacitated cultist (and, of course, a dead one) will respond differently from an intact one. It's non-repeatable because between one shot and the next the situation is likely to change. And it's always because I've never seen a game where you didn't roll attacks except for occasional coups de grace or really good surprise attacks on minor adversaries.
If instead you were steering a guided missile towards a drifting asteroid, this would likely be a process because a missed shot can be turned around to try again, providing you don't hit anything else.
Let's say instead that you're climbing a cliff. This is probably a negative task because each misstep reduces your overall success, so in a group some of you might slip back down. I'm inclined to call this a process because you can adjust what you're doing based on how you're doing so far. The consequences depend on how you approach it; they're quantitative if all the safety measures are up and you only lose time, but they're qualitative otherwise because you might break a leg or die. It's repeatable because providing you don't really mess up, you can rest and have another bash at getting up that ledge, or try a different route. And it's always because while climb doesn't get rolled out for every poxy ladder, cliffs are one of the cases where it does.
Some fairly scattered thoughts there. I'm working towards the idea that it might make sense to actually model different kinds of task in different ways.