I should be working on something else, but a question came up in conversation that my brain has been worrying like a bone: When should the GM roll dice?
For a little bit of context, it's important to be aware that there are several game designs in which the GM never rolls. The most famous example is probably the SAGA system, though Apocalypse World has also brought the practice to mind for many. For players unfamiliar with these games, this may seem counterintuitive at first, but it's fairly easy to imagine. Most systems are fairly symmetrical, so the average outcome of die roll + modifier is roughly in line with difficulty, which means that changing which side is active (rolling) and which is reactive (using a passive value) doesn't actually change the outcome.
To illustrate more concretely, let's suppose your opponents has an attack modifier of +5 and is rolling d20 to try to hit your armor class of 16. He needs to roll a 11 or better to succeed, straight up 50/50 change. However, suppose instead of an attack modifier of +5, he had an attack _value_ of 16? Rather than roll any dice, he would just act as if his attack roll had totaled 16. Similarly, rather than an armor class of 16, you had a defense value of +6. When he attacks, you have to roll d20 and add that 6, trying to beat his 16.
It's a weird shift, but it doesn't actually change much of anything about the system. Modifiers still get applied without changing the math at all. This also works for skills too: Trying to sneak? Stealth vs. a fixed difficulty. Trying to spot someone sneaking? Perceptions versus a fixed difficulty. Easy peasy. In short: EVERYTHING is a target number.
Of course, this is not the only model. Apocalypse world's system isn't symmetrical (it's not attack vs defense), it's based on rolling for outcomes, and all the "tables" you'd roll on (the moves) are purely for players. There's no question of the GM rolling because there's no place where she _would_ roll.
Bottom line, this is a design option, but it's a curious one, and also a bit contentious. Some GM's love it, viewing that it gives them less to do during a game so they can focus on other priorities (those priorities varying from GM to GM). Other GMs find it off-putting, usually because the techniques of die rolling are part of their kit, important to their sense of drama or communication.
Personally, I have a great intellectual appreciation for the GM not rolling. It keeps the players proactive, even in otherwise reactive situations, and I'm a lazy enough guy to appreciate the release. However, in practice. it falls short for me when I go cold turkey. It's something that's very hard for me to put my finger on, but I feel that I am not suitably engaging the game, and that my decisions and rulings have too few limitations. I enjoy the bounds that dice put on things, and I use them for oracular purposes (LINK: http://rdonoghue.blogspot.com/2010/03/arbiters-and-oracles.html) as often as I do as arbiters, perhaps even more often. In retrospect, one of my stumbling points with Apocalypse World was that I was not comfortable raining more-or-less unchecked fire down on the players when the dice turned against them. It felt like cheating.
Anyway, all that was a big prelude to come around to the question: When should the GM roll dice?
Obviously, different systems have specific answers to this, but that's not quite what I'm getting at. It's not something I've thought about generally speaking when kicking around design.
The big answer seems to be "When there's an outcome the GM doesn't know" but that's deceptively simple. It begs the question of why the GM doesn't know. Most often, is it because uncertainty is desirable for the same reasons uncertainty is useful for players, and perhaps unsurprisingly, those are the kinds of rolls that can most easily be passed from the GM to the player.
Other reasons for uncertainty really revolve around what the GM _should_ know, and that's kind of interesting, because that's a pretty profound design decision - it has a lot of implications for the GM's role and authority. Consider something very old school - reaction tables. Games like AD&D used tables to determine the reactions of NPCs. Now, whether this is an assertion that the GM shouldn't make these decisions or a tool to save him the decision-making (much the same way a random table for encounters save the headaches of filling in blank areas on the map) is not as important as the fact that it could be either.
Still, while that speaks to certain issues, I'm not sure it does so informatively. Turning it on its ear, maybe the question is what the GM rolling the dice signals.
Curiously, what it doesn't signal is that something interesting is about to happen. That's a key one, because one might expect that to be the case, but the reality is that when something interesting is about to happen, the GM usually calls for a *player* to roll dice. There are a few tension-building techniques that look like the the GM is rolling to make things happen, like rolling for no apparent reason to make players nervous, but they're not actually mechanically relevant.
What it does signal, at least to my mind, is that the GM is now subjecting himself to the rules. By rolling the dice, he is making an announcement that the decision which follows is going to be subject to the outcome of the dice. This is, I think, why dice fudging can be such a visceral issue for people, far beyond it's theoretical impact. It takes something the GM is communicating very explicitly and makes a lie of it.
Does it signal anything else? That something is happening, sure, but that's usually already quite clear. Sometimes it communicates details about what's happening, but that only applies in systems where the number or type of dice use signifies something. That is, a double handful of d10s is a meaningful way to communicate to Exalted players that their enemy is a scary badass, but in D&D, all d20s are pretty much the same (though you can do some stunts with pulling out the REALLY big d20).
At the end here, I don't feel like I'm any closer to an actual insight. Lots of things that work part of the time, or offer limited insight. I feel like there's something here, but I can't quite see to the heart of it. Still, I feel like I'm closer to sense of how to use the point at which the GM rolls as something to get purchase on in a design.
1 - Here's an important note - I don't have an answer to this, and I don't swear I will have one by the time this is over. If I had an answer, the question would not be nearly as compelling, so this is mostly me airing out my thought process.
2 - Ok, that's a small lie. You need to decide on defaults, so in the case of d20, you create attack value by assuming the NPCs always roll 10s or 11s. Either is about right (stupid averages) but it's an important decision. I tend to vote 11, and players win ties, which is mathematically closer to 10.5.
3 - There's also a concrete mechanical impact on systems that involve competitive die rolls (attack vs. defense) that is best illustrated by the Fudge dice used by Fate. The usual 4df roll generates a curved result from -4 to +4, but because each die is neutral, your odds are never going to get any better by changing the number of dice. This is important because when the GM rolls 4df as well, it's the same as if the GM had rolled nothing and the player had rolled 8df, for a result from -8 to +8, but with no better odds of success. In situations like this (or in any game where there is rolling on both sides of a contest), there may not be a change in overall odds, but the range of possibilities broadens. That may or may not have any mechanical impact, depending on the system, but it's somthing to keep in mind.
4 - While I don't think this reflects on the broader issue of whether or not GMs should fudge dice rolls, it definitely suggests a reason it's such a hot button topic.