It's implicit that a murder will have impact on those around the victim, and that there is an existing web of relationships that this murder exists in the context of. In contrast, a theft or similar crime often exists in a web of value and price, where the interests are usually more measurable (as in the value of the thing stolen).
The problem is, a lot of games are very cavalier about life and death. Many games encourage either an enthusiasm for slaughter by creating hordes of faceless monsters, or embrace a sort of indifference, as murder is just one more facet of a dark, dark world. If your game incorporates either of these views, it's going to change what sort of crimes you can use. If your group understands that someone might commit a crime for gain, but balks at one for less concrete reasons, you may want to stick to theft, extortion, bribery and other crimes of that stripe. In that case, some of this may not be as useful to you, but the bulk of the principals still apply.
Which brings us back to murder, the most basic, primal of crimes. What do you need to make a murder mystery, for purposes of the game?
Laying the Foundation
Obviously, you need a killer, a victim and a crime. Any of these is a solid place to start, and each allows you to develop a certain sort of mystery. Once you've worked out your starting point, you're still going to want to work out the details of the other two, but you'll be able to use your core idea as a foundation.
Building From the Victim
The victim is the usual place to start. If the victim is your core, then the mystery should be about the web of people who surrounded him. One of the people connected to the victim is the criminal, and the crime was committed because of their relationship. Other relationships may muddy things, and there will still be clues and red herrings to deal with, but when the victim is the foundation, the suspect list is usually very easily derived. Most mysteries, especially books and television, follow this model because it is so strongly character driven. You have a stable of potential characters, each with a strong incentive for interaction (proving their innocence, or someone else's guilt). Most of the truisms of the genre, like seeking motive, means and opportunity, are most strongly present here, and it's usually the most controlled of mysteries. It is also the kind that can potentially call for the least conflict, since most of what can be determined is done by investigation and conversation. How good or bad that is can depend a lot upon the group.
One of the characters connected to the victim is usually the client, who we'll get to shortly.
Victim driven mysteries include the mystery of the boardgame Clue and most episodes of Murder, She Wrote.
Building from the Crime
When you start from the crime, there are usually two reasons to do so - the crime is either a loose thread or a novelty.
Loose threads are crimes that, upon investigation, lead to other crimes. The first thread gets pulled on, and something much larger unravels. This model does a very interesting job of turning the traditional mystery on its ear because the mystery is usually not very hard to solve. There's a strong sense of progress (rather than frustration) as the layers are peeled back. Characters are rolled in to deliver a piece of evidence, and rolled back out just as quickly, and the process culminates in the final, true crime being revealed. Because it moves so quickly, there's not a lot of room for interaction between the characters, which is why this is the ideal model for the police procedural, where the bulk of the story focus is on the team of investigators more than the crime. At first glance, this may seem the most game-able of models. The focus on the investigators and the steady progress are both very compelling. The problem is that the lack of real interaction outside of the investigators makes for a very isolated sort of game. Most notably, this is a tricky model to GM, since it requires setting up tiers of mysteries, which is a lot of work, but is also (because it is so highly structured) the closest thing to a dungeon-style mystery available. This one really merits a full blown sidebar, bit for the moment, let me simply say that this can easily be the best and worst of models at the same time.
The vast pantheon of modern crime drama, tends to fall into this model, with Law & Order: SVU as perhaps the best example.
The novelty crime, on the other hand, is often less about who committed the crime as much as how they did it. Once the how is determined, it usually points quite directly at the who. What's more, the identity of the victim and killer tend to be secondary to the novelty of the crime. The classic "locked room" mystery, where a victim is found dead in a locked room, which has no apparent entrance or exit, is probably the most famous example of this.
The problem with this approach is that it's very difficult to make a novelty crime something that you can really keep focus on. They tend to be Gordian knot's - once a solution is found, everything else falls into place. Because of this, they tend to be ideal mysteries for broader, more adventurous stories, where the mystery is only a component of the whole.
In Dresden, adding magic to the mix can complicate novelty crimes even further. Magic can easily be the explanation for all manner of "impossible" crimes, but it becomes easy to get lazy and just feel that magic can do everything. The rules and limits of magic need to be kept firmly in mind for any magical crime. Without those limits, then player will look at a situation and just assume someone was powerful enough to do it. That kind of assumption can be great for the occasional hook, but if there are to many mysterious figures flaunting the rules of magic, then detective work becomes almost moot.
Many pulps use the novelty crime, as does Scooby Doo. The murders in Storm Front, Summer Knight and Death Masks as well as many of the other issues Harry runs into are also novelty crimes. And lest we seem critical, many of the tales of one Sherlock Holmes are novelty crimes.
Building from the Killer
Focusing on the killer is usually only appropriate when he is a continuing threat (he may kill again or whatever). This is more of a model for thrillers and horror stories than pure mysteries. Thankfully, the Dresden files have more than their share of horror and thrills, so that's just fine. Serial Killers, Man-eating monsters and spiteful Spirits are all possibilities of foundation villains.
If the killer is your foundation, the relationship to the victim can be much more tenuous. Victims may simply have been in the wrong place at the wrong time, or they may have been chosen for non-obvious reasons. Because of this, a suspect list can be far more tenuous to create or maintain. Additionally, a common complication is that the killer is just an agent of a true villain, and that villain must be stopped.
The agenda of the killer (or the villain) is also usually quite different than in other mysteries. The crime was not his end, but rather a stepping stone leading to his end. It is usually this end which is what really needs to be stopped, and by extension, the true mystery is determine what the agenda (and potentially who the villain) is. This can have a lot in common with loose-thread mysteries, in that there's often a progression of issues to solve.
Killer-based mysteries are the easiest to make into adventures because the Killer/villain is usually a worthy antagonist who must be confronted and defeated. This allows for a whole lot of exciting conflict. The price for this is that they're often less intricate as mysteries, but the tradeoff can be more than worthwhile.
Building the Frame
Once you've picked your foundation, you need to frame it with the other two elements. Often, the details of the foundation will suggest the nature of the frame, but it's usually worth thinking of at least one or two complications to throw in here, if only to serve as a red herring.
If your foundation is the victim, a novel method of murder can be an excellent way to muddy the waters or add color to the proceedings. Often, a novelty can be a component of the larger mystery, so that when it's resolved, some component, like the falseness of an alibi, is revealed. Similarly, if the villain is still at large and willing to kill again, more bodies may show up, ratcheting up tension.
If your foundation is the crime, you're going to need to put some more structural work into the frame to help keep everything together. You want to flesh out the victim and the killer's relationship enough that it's easy to see motive. Remember that in most novelty killings, the killer may be obvious, but it's impossible to prove unless someone can figure out _how_ he did it.
If your foundation is the killer, the victim is a great place to hide a clue. The apparently random victim may later be revealed to have an unexpected connection to what's really going on. Additionally, the killer may kill in some distinctive way, so that the crime originally looks like a novelty crime and the real problem is only revealed with the second victim.
Building The Doors
After the victim and killer, the next important character is the client. This is the character who is responsible for drawing the PCs into the mystery. In a classical detective story, this is the client who hires the detective to solve the crime, but it could be almost anyone: a friend of the characters, or someone who is owed a favor, or even someone seeming wretched enough to inspire their pity and willingness to help. The client is most often tied to the victim in some way, and may themselves be suspects
There doesn't always need to be a client - coincidences can occurs, but usually those coincidences will be chance encounters with a potential client. Television programming is the best example of how this is handled - you have a character who is involved in a crime every week and you need to keep it looking rational. The easiest solution is to have characters who solve mysteries for a living, like P.I.'s or cops. In the absence of that, characters should have enough social ties that there are reasons friends or family would turn to them.
It's also possible to draw players in directly, by making them suspects or witnesses. By giving them a direct stake, they become their own client. However, drawing the characters directly into the mystery is a trick to be used sparingly, as it quickly stretches credulity.
Building the Walls
The next question is what the bounds of the mystery will be: the big three to look at are
1 - Most of the best capers exist at an intersection between these models, where there is one web of value, based on what is being stolen, and one web of personal interest, usually tied to why the thing is important. These can make great crimes to investigate, but it's actually a model that sees it's greatest reward when the characters are the criminals. Good examples include Ocean's 11 and Donald Westlake's Dortmunder novels.
2 - In a theft, while the person being stolen from maybe the victim, it's also possible that the thing stolen can be considered the victim, especially if the thing stolen is of specific value beyond pure money. If many characters have relationships to, say, the stolen Maltese falcon, it serves the same purpose the victim did. In these cases, the item stolen must have some value beyond pure money.
I'm off at Origins, so I've dusted off some old articles for this week. This is an old one, from 2005, that I had entirely forgotten about. You can find the original here.