Our extended social group is in the habit of running week-long holidays of roleplaying where everyone pitches in and runs a game – the ubiquitous “one-off”.

Running a one-off game is like a short story – you need to get stay focussed on the plot, get to the point, and wrap it up with a (satisfactory) conclusion.  All in 4-6 hours.

Now and then I manage to make a poor choice either with setting, or plot, or something that makes the game longer than it should be or harder to run or just not turning out the way I thought it would.  I’m writing this post so that when I come to write another game maybe I’ll avoid making the same mistake again…

Dead ends (misdirected players)
Since a RPG is driven by player action, the plot has to be signposted clearly – otherwise the players could be milling around for ages.
Learning: signpost the plot clearly, and don’t let players explore dead ends for too long – have a back up plan to steer the PCs back.

System over plot (temptation to model everything with dice)
If the players need to learn the system, that could be an hour out of your time slot right there.  Unless there’s a nifty tutorial section to help the players learn the system, it should be kept very light.
There’s even an argument that you should cut down on dice rolling.  One-off games will have a few plot points where the players have to overcome adversity – if this hangs on the outcome of a dice roll, you risk stalling the whole game.
Following this advice is a big problem for me since I like trying out new systems – but  the evidence is that every time I’ve tried a complex system that’s new to my audience, the game has been harder to run and enjoy, and definitely taken longer.
Learning: For system, summarise the character abilities with four or five bullet points.  Even better, write it on an 3×5 index card.  Anything more than that, the player may well forget that they have half the abilities on their character sheet.

Combat over plot (fights take longer than you think)
Combat will definitely dominate the plot if it’s system heavy (see above).  But it will also dominate even in simple systems.  I have planned combat heavy games and ended up running only 1 combat, narrating subsequent combats because otherwise we’d run out of time.
Learning: combat is like any other plot point – the reasons for fighting need to be signposted to the PCs.  Have just enough system to make the combat interesting.  No more than 2 (preferably just 1) scenes should be dominated by combat.

Too much information
I have swamped my players with handouts in the past (due to temptation for too much world, below).  They rarely remember all the information and sometimes won’t bother to pre-read, so you have to tell them the information in-game.  All of this takes time, upsets the flow, and god help you if you buried a crucial plot point in a player brief.
Learning from this:  Following the “sparse character” rule, character + backgrounds should be 1 sheet of A4, no more – otherwise the player will forget.  An exception to this rule will be in-game aids like maps, found notes etc.  Handouts can be pretty but must be brief.
Addendum: I tried making little 8-page A5 booklets for some games – I was terribly proud of myself but I suspect they made the information harder, not easier to read.  In future I’ll stick to A4.

Too many events
Each plot point will always take longer than you thought.  Any more than 1/2 dozen scenes risks over-running by a big margin.  Some GMs can probably plot a dozen scenes and keep to schedule, but I have trouble.
Learning: allow 1 hour per scene, roughly, making maximum of 6 scenes in the session.

Splitting the party
After combat, splitting the party has the biggest potential to suck time away.  If the party starts up split up and must find each other, then find a motivation to work together, that’s an hour or more right there.  If the party fractures during the session – usually because there’s no motivation for everyone to stick together – that’s another time-sink.
Learning: give the PCs a reason to stick together and work together.

World Too Big
Too many locations dilute the area where you want the PCs to focus, and may obscure plot signposts.  And I am a big fan of big, well developed worlds with history – which falls into the Too Much Information trap.
Learning: Make it clear which is the play area and which is outside the sandbox.  Outside should be one location to minimise information (even if it is technically more than one location – like two nations at war), and there should be no more locations than scenes.  Don’t name every city or village, you have better things to do.

Time perception
I’ve sometimes plotted a single session to represent several days.  This rarely if ever works.  Why?  Because it’s very, very hard to get a group of PCs to agree to skip time together.  There is always more plotting and more investigating to be done (either by all the party or just one person), and even when you convince the party that they’re really, really tired and the caravan isn’t going to turn up to the ambush point until tomorrow there’s always the need to roleplay setting up watches and tripwires around the camp.  It’s just not credible that the players should stop what they’re doing and go to bed right now, so don’t force it.
Learning: Let the players decide when they need to rest (and whether they can risk resting while the bad guy is getting away).  Otherwise, keep it moving.

Summary
Maybe not a definitive list of rules, and probably open to some challenge.  But I’ll keep these rules in mind for the next time I run a one-off (probably next year).
1. Signpost the plot clearly
2. Character attributes should fit on an index card
3. Budget for a maximum of 2 fights, preferably 1.
4. Individual character information should fit on a sheet of A4.
5. Stick to 1 scene per hour (on average).
6. Give the PCs a reason to work together.
7. World – keep 1 location per scene maximum, plus one location outside the sandbox.
8. Assume your PCs will never sleep until the game is finished.