Me, I don't really care one way or the other and will happily play either type of game without much fuss. It's rare that I dislike a game based on whether it uses classes or skills as opposed to the alternative. In playing skill-based games, I have noticed a couple of interesting things, though, both in myself and even in those who are strong advocates of the "superiority" of skill-based over class-based systems. First, I don't demand rolls for trivial uses of a skill, unless they're done under extreme or challenging circumstances. This is, of course, standard procedure and has been for a long time. For example, there's this passage from Chaosium's Basic Role-Playing:
This term [automatic actions] describes activities which are always successful under normal circumstances. There is no need to roll any dice for these. They are assumed 100% successful. These include walking, running, talking, seeing, hearing, and any other normal basic function.Of course, we all remember referees -- we may have even been them ourselves -- who didn't abide by this advice and required skill rolls for everything: "I ride to the next town." "Make a skill roll." "I fail." "Oops, you fall off your horse and take 1D6 damage." "But I'm a knight; I've been riding a horse since I was a boy." "Too bad, you failed your skill roll." That rather narrow-minded interpretation of skill use doesn't long survive contact with actual play, so the advice quoted above only makes sense. In these circumstances, there is no actual difference between a class-based or a skill-based game, except that a character sheet in the latter lists a level or percentage associated with riding horses, while in the former it's just assumed.
Attempting to do these things under extraordinary conditions, or trying to do them with close scrutiny, requires a die roll, as outlined in the next section.
The second interesting thing I've noticed about skill-based gameplay is a consequence of the first one already discussed. If we accept that skill rolls ought not to be demanded under "normal circumstances," it follows, then, that, as BRP suggests, "extraordinary circumstances" are when a skill system is actually needed. What I've found, though, is that, in such extraordinary circumstances, referees are often quite prone to fudging the results of a skill roll; I know I used to do this unashamedly and I was not the only one to do so. Part of the reason, I think, is that it seems wrong to many gamers to have the outcome of some extraordinary circumstance hang on a single dice roll. "You have to defuse the bomb quickly or everyone will die. What's your Demolitions skill?" "35%" "Make a roll." "Uh, 74." "Boom!" The situation is even more common when the skill rolls in question pertain to finding hidden or unusual things, like secret doors or clues. "I search the room." "Make an Observe check." "I fail." "You don't find anything out of the ordinary." And so we see elaborate rationalizations as to why the skill roll didn't really fail or even a further narrowing of the definition of "extraordinary circumstances," all in the name of seeing a character succeed at something the rules otherwise say he'd failed.
None of the foregoing is meant to suggest there's anything wrong, let alone bad, about skill systems in RPGs. I mention this simply because I've noticed that, in practice, skill systems are rarely used much differently than the non-quantified skills assumed to exist in class-based games. The primary difference between the two approaches is esthetic and, believe me, I'm not one to knock esthetics. When I designed Thousand Suns, for example, I didn't hesitate to use a skill system, because, as a sci-fi game, it seemed to make more sense to have characters defined by their areas of training and knowledge rather than by an archetype-based class. It's the same reason why the game uses the metric system -- it feels right in a science fiction game, even though the game would have worked just as well if I'd have adopted US standard measurements. It's the feel of a skill system that appeals to a lot of people, not its actual mechanics. It's the sense that skills are somehow more "realistic" or better reflect reality, and I can raise no objection to such subjective notions. But they are subjective and that needs to be borne in mind in discussing the merits of skill vs. class systems.
This brings me to a broader, final point. I don't think we can underestimate the impact that conceptualizing "adventures" as "stories" has had on RPGs and how they're played. If we see an adventure as having a definite, specific end before play even begins, it seems inevitable that, rather than being an aid to play, a skill system, with its quantification of many activities and areas of knowledge, is an impediment to it. On the other hand, if we see an adventure as simply being a situation into which the PCs are thrown, whose ultimate outcome depends on a combination of choice and luck, a skill system -- or, indeed, any random element -- is a contributing factor to that ultimate outcome.
In my second Thousand Suns session last weekend (which I'll post about eventually), a character failed an important skill roll and I let it stand without any fudging. As a result, the PCs could not achieve their current objective and were forced to move on. Had I conceived the "climax" of that adventure as depending on the success of that skill roll, I'd have been disappointed and might well have succumbed to the urge to fudge its result. But the adventure was just a situation in which the PCs found themselves at that moment. Failure meant only that one particular outcome to the adventure was closed off, but there were other possible outcomes, not to mention other adventures/situations. Rather than being an obstacle to our fun, the failed skill roll enabled it, pushing the characters and thus the campaign in a direction the players might not have chosen if they'd had the choice. Speaking only for myself, that's what all the best game systems do, regardless of whether they use classes or skills as their foundation.