In the roleplaying game the Burning Wheel, there is a rule called “Let it Ride” that states that when you have a lengthy, multipart task which requires a certain skill, you should only have to roll the skill check once, and that skill roll is applied to every part of the extended task. Today I’m analyzing this rule and the context behind it.
On the positive side, I can totally see where this rule is coming from. The author mentions that it is directed against the concept of “roll until you fail.” This is a good way of putting it, I have experienced and hated that very same problem myself.
The problem here involves lengthy skill uses which take up considerable "screen time" and involve many sub-elements. For instance, trying to sneak into an enemy base involves sneaking past many areas with sentries or other dangers. Each time, there is logically a chance that the character would be spotted. Indeed, in the movie, there would be a bit of suspense as the character tries to sneak past each individual area where he might be spotted. So it is probably natural for the GM to want to call for a stealth skill check each time the character does something that risks getting spotted.
Unfortunately, when combined with the ordinary skill check rules used in a normal role-playing game, this can easily go disastrously wrong.
The main problem stems from the fact that the probabilities assigned to a skill check usually assume that the check will only be performed once. The GM is expected to determine a difficulty rating according to how hard the check seems to be. So a highly skilled character might have a 70% chance of succeeding in a "hard" skill check, for instance. This is okay when the hard skill check is to quickly repair a badly damaged sports car; since the individual steps aren't very interesting and quite possibly would occur "off-screen", the GM would probably call for one check and see if it succeeds or fails. But in the example of sneaking into the enemy base, the GM might decide it seems like a hard thing to do, and call for 6 hard skill rolls. But this disregards the fact that since failure in any of those roles of essentially means total failure, you have to succeed all 6 times, and that is much much harder than succeeding once. Now your chance of success is only 12% instead of 70%.
Not only is this excessively difficult, but the whole process is very arbitrary. The difficulty of the task ends up depending on how many sub-tasks the GM decides to split it into, and there is no telling how many of these there will be. This can even mean that if the player is good at role-playing and describes his actions in a more thorough and interesting fashion, the GM will think of more opportunities to call for skill checks, and the chance of succeeding will drop. Not much fun at all. I can recall playing games where you just knew there was no way the GM would allow you to succeed in a multi-part self check, even though the GM probably wasn't aware that this was what he was doing.
Even if the GM is fully aware of this problem, however, and wants to make the individual rolls easier, there are further issues. Unless the GM has considerable mathematical sophistication, determining how much easier to make the individual rolls is difficult. After all, there are no guidelines in the game for how to adjust difficulties based on number of rolls required for the check. It would be all too easy to over compensate or under compensate. Especially when you may not be sure how many rolls you are going to end up requiring beforehand. Also, the kind of adjustments required would depend upon the type of dice rolling used in the game.
Trying to convert to making numerous skill checks with a tiny chance of failure is just awkward. There isn't a lot of granularity. If you are making skill checks on d20, for instance, the chance of rolling 3+ 6 times in a row is 53%. Improving the skill check to 2+ increases the odds to 74%. And the next step up is 100%. This means that there is no way to map the many different success probabilities the characters might have on their skills; whether your normal chance of success is 8+ or 5+, the closest you can get to the same probability when rolling 6 times is 2+ either way. You certainly can't easily use a formula like add +6 to your skill bonus, as small variations in skill bonus or difficulty number would mean major changes to the overall chance of success.
But getting the odds correct still leaves a further problem. Rolling many checks, each of which has only a small chance of failure, is not necessarily as exciting as one might like. Rolling a d20, and trying to get 3+, means that a roll of 3 and a roll of 19 are pretty much the same thing. It feels odd, like you aren't really demonstrating your skill, but rather rolling dice to see if your character screws up. When you're rolling for tiny sub tasks, it is hollow to succeed extremely well in one sub task when failure in the remaining sub tasks are just as likely to make it a totally moot point. While if you “Let it Ride” and let one skill roll cover the entire task, it is much easier for the GM to interpret the overall quality of the roll in determining the amount of success you achieve.
For all of these reasons, I think the “Let it Ride” rule is a considerable improvement over the alternative of “roll until you fail”. My only complaint is that I still feel it is pretty far from the optimum rule for complex skill uses. It goes too far the other way. When you have a multi-part skill check with many possible points of failure, you either fail right up front or not at all. As I mentioned earlier, in an actual movie or book every point of failure would be a point of suspense, and you can never be sure whether the entire operation will succeed, or if not, at what point it will fail. But when one skill roll covers everything, once you succeeded in one task, you automatically succeed in every subsequent task – not very dramatic. Of course the difficulty could increase, but if you fail the check because of this, you may feel less like you failed and more like the GM suddenly pulled a dirty trick on you. Of course, the GM could have every task become more difficult than the last, but that means the GM is really “cooking” the adventure in a very specific way. And that it runs head-on into another problem, which is that if you know your roll is low or high to begin with, you can take advantage of this in illogical ways - retreating because you know you are bound to fail the next check, or taking all sorts of extra risks because you know your roll is so high you can't fail.
I think the best solution to this kind of thing is a much more elaborate system for resolving complex “skill challenges”, something like the superb Dramatic Skill Resolution system in Torg. Indeed, the Burning Wheel itself has a very simple version of such a system. But this is a difficult area in which the state-of-the-art in game design is still under development, I don't think any existing skill system I know of would really hit the “multi-part infiltration” scenario totally on the head.