Icewind Dale: Rime of the Frostmaiden is designed as a sandbox. And not just in name only. It’s the real deal: A campaign where the players are empowered to either choose or define what their next scenario is going to be.

In the latest Advanced Gamemastery video, I simulate the opening sessions of a campaign in Icewind Dale, walking you through how a master GM brings the Forgotten Realms to life.

ICEWIND DALE ARTICLES
Review: Icewind Dale – Rime of the Frostmaiden
Icewind Dale: Running the Sandbox
Icewind Dale: Travel Times
Icewind Dale: Goat-Ball
Sandbox & Simulation

Subscribe Now!

The concept of an RPG sandbox campaign often gets mixed up with a lot of other things. Some of these are common structures used for sandboxes (like hexcrawls). Others are just misnomers (like sandboxes being the opposite of a railroad).

(Quick definition: A sandbox campaign is one in which the players are empowered to either choose or define what their next scenario is going to be. Hexcrawls are a common sandbox structure because geographical navigation becomes a default method for choosing scenarios, which are keyed to the hexes you’re navigating between.)

Other conflations are subtler. A particularly common one is to conflate simulationism with the sandbox structure. One major appeal of the sandbox can be that it allows players to feel as if they’re “living in the world” because they’re free to do “anything,” which has a fairly large overlap with what people enjoy about simulationism.

But simulationism is not required for sandbox play.

A good example of this is the chardalyn dragon from Icewind Dale: Rime of the Frostmaiden.

SPOILER WARNING!

When the PCs approach a particular location on the map (Xardorok’s fortress), this triggers an event in which the dragon flies away to cause some havoc. In discussing this as part of a sandbox scenario, I was challenged: How could it be a sandbox if it was dramatically triggered by the PCs’ approach?

(Note: I’m just talking about triggering the dragon flight here. Shortly thereafter Rime of the Frostmaiden ALSO has an NPC show up to trigger a linear plot that ends the sandbox. I’m not talking about her. Just the dragon.)

The confusion here is due to the conflation of the sandbox structure and simulationism. A simulationist wouldn’t trigger the dragon based on the PCs’ approach. They’d probably do something like have Xardorok’s construction of the dragon be on a schedule with the dragon being released when Xardorok completes it, regardless of whether or not the PCs have found Xardorok’s fortress yet. (There are also other simulationist techniques that could be used here.)

But a sandbox isn’t dependent on simulationism. There’s nothing about dramatically triggering an event which is incompatible with the players remaining empowered to choose and define their scenario.

Go to Icewind Dale Index

Go to Part 1

So what does this all look like in actual play? Well, let’s simulate a campaign (with some actual dice rolls for stuff like the campaign’s starting town) and take a look.

For the purposes of this simulated campaign, let’s largely ignore the players being able to define their own scenarios. In actual play, this will almost certainly happen: Ten-Towns is enough of a living environment that the players can, for example, decide to become caravan guards from Kelvin’s Cairn; or re-open an abandoned inn; or buy mead from Good Mead to sell at a high price in a town where the taverns are running dry because trade has been disrupted. But what we’re going to focus on here is primarily just the baseline play that arises directly out of the sandbox structures in the campaign.

INITIAL STARTING QUEST: Cold-Hearted Killer

A dwarf named Hlin Trollbane believes she’s identified the serial killer who’s been plaguing Ten-Towns — it’s a man named Sephek who’s travelling with Torg’s merchant caravan. She approaches the PCs in a tavern and asks them to track the killer down and kill him.

STARTING TOWN: Caer-Konig

The quest in Caer-Konig sees the PCs stumble across Speaker Torvus, the dragonborn leader of the town who is drunkenly patrolling the streets in a vain attempt to capture dwarven thieves who have been sneaking into town. This eventually leads them to a nearby Duergar Outpost (p. 47).

Right off the bat here my instinct is to have the PCs enter Caer-Konig, encounter the drunken Speaker, and get the Duergar Outpost quest. When they get back to town (having leveled up to 2^nd level), Hlin says, “I’m impressed. I think you might be able to help me bring a little more justice to this cold-blighted Dale.”

The success of the first quest diegetically justifies Hlin approaching them for the bigger job. I now roll to see which town Sephek will be found in.

SEPHEK’S TOWN: Easthaven

Alternative: If you wanted to more strictly adhere to the published structure (starting quest first!), the campaign starts in the Hook, Line, and Sinker (p. 46), where Hlin hires them to kill the serial killer. They leave the tavern and immediately stumble over Speaker Torvus, who leads them to the Northern Light tavern on the other side of the town and starts the Duergar Outpost quest.

STRANGE ALLIANCES

Either way, they head down the road and pass through CAER-DINEVAL.

Caer-Dineval is one of the towns without a proactive quest, so the PCs could just pass right through without getting one. But they’re looking for the serial killer, right? So they’re going to head to the local tavern, the Uphill Climb (p. 38), and start asking questions.

The adventure tells us that Roark, the proprietor of the Uphill Climb, won’t explicitly tell the PCs what’s going on in town (most likely out of fear), but he will try to point them at the caer (castle) in the hope that they’ll get involved. So when they start asking questions, he’ll say something like, “If any caravan was looking for permission to set up here, they’d inquire with the Speaker up at the caer.”

So the PCs head up there and knock on the door.

The caer has been secretly invaded by a cult called the Knights of the Black Swords that, among other things, wants to stop the duergar invasion of Ten-Towns: They’ve killed the guards, taken the Speaker hostage, and are ruling the town in his name. The way this quest works is that the PCs can bust up the cult and rescue the speaker, OR they can end up allying with the cult. The cult has some divine guidance which, if the PCs have taken any anti-duergar actions, will have informed the cult that the PCs can be useful allies and that they should go out of their way to accommodate them.

So if the PCs did the Duergar Outpost quest, then the likely outcome here is that the Black Swords form an alliance with them. (“Your coming has been foretold!”) That’s a second quest complete.

Alternative: If the PCs heard about the duergar thieves and said, “Doesn’t seem like our problem,” or if they tried to follow the duergar tracks, got lost, and never found the outpost, then when they go up to the caer to ask questions about Torg’s caravan, they’re simply told, “Nope, no Sephek here,” and turned away.

As the PCs head back down the hill from the caer, they meet Dannika Graysteel, who’s heading back from another disappointing attempt to find a type of fairy called a chwinga. This kicks off the SECOND STARTING QUEST — “Nature Spirits”, p. 25 — when Dannika asks them to look for chwingas in the other towns of Ten-Towns.

Now the PCs head down the road to where it intersects with the Eastway.

The choice of which way to go is now basically random. So, for the sake of argument, I rolled a die and determined that this hypothetical group is heading to GOOD MEAD.

As they approach Good Mead, they encounter a trapper. They ask him about chwingas, but he shakes his head; he hasn’t seen any chwingas around here. (There’s a 25% chance each town has chwingas. I rolled a 47 for Good Mead, so no chwingas here.) But he does tell them that he just discovered five dead bodies out on the tundra. This is the quest hook for the Verbeeg Lair (p. 71).

These players, however, decide NOT to follow the trail from the dead bodies back to the verbeeg’s lair. Tackling a giant all by themselves just sounds too tough. But they want to do the right thing, so they gather up the bodies and take them into Good Mead for a proper burial.

In Good Mead they hear that the verbeeg has stolen the town’s mead supply and killed the Speaker, threatening to ruin the town’s economy. But this mostly just confirms that the giant is going to be too tough for them to handle.

SAVIORS OF GOOD MEAD

One of the PCs, however, decides to rally the townspeople: Alone they can’t stop the verbeeg menace. But together they can triumph!

The stat block for Good Mead conveniently lists what the local militia can muster:

So the PCs make some solid Charisma checks and they rally two veterans and 10 tribal warriors to go giant-hunting. (If they’d rolled better, they could have perhaps gotten ALL the tribal warriors to go with them.)

On the way to the verbeeg lair, I frame a couple of scenes where they’re chatting with other members of the expedition. I use the opportunity to introduce Shandar Froth and Olivessa Untapoor (p. 78) and improvise a couple of new NPCs representing the common folk of Good Mead. During this small talk, the PCs also pick up two more rumors: Dwarves are having trouble bringing their goods to Bryn Shander due to yeti attacks. And Dougan’s Hole, down the road, is being plagued by a couple of dire wolves or awakened wolves or polar wolves or werewolves.

(Depends on who you ask and how tall the tale has gotten.)

Note: I’m deliberately inserting uncertainty and/or broader context into these rumors compared to the default text in the adventure. See also Random GM Tips: Surprising Scenario Hooks.

The giant-hunting expedition is a huge success. Maybe one of the group’s new NPC friends gets killed (a little pathos never hurt anybody). That’s another quest complete, so the PCs are now 3^rd level.

The PCs return to Good Mead. While everyone is celebrating (and mourning), Olivessa Untapoor approaches them: Good Mead needs a new Speaker. A strong speaker. Shandar Froth thinks he should do it, but he’s a jackass. People are asking Olivessa to run against him, but she really doesn’t want the position. She thinks that one of the PCs should stand for the election.

Now, if the PCs want, this could totally happen! They’re the heroes of the hour. They’ve got the support of a major civic leader. There’s a whole thing where Shandar, regardless of who he ends up running against, pulls some shenanigans during the elections (p.78), but the PC candidate can probably end up on top.

This would, of course, change the entire nature of the campaign! Which is great! As the DM you’d need to come up with some civic challenges for the new Speaker (and their closest advisors; i.e., the other PCs) to deal with. You don’t have to completely abandon the existing toybox while you’re doing this, though. For example, you can look at the existing rumor tables and think about how to re-contextualize them to the PCs’ current circumstances.

For example, this rumor:

In Lonelywood, beware the dreaded white moose! It attacks loggers and trappers on sight, and the town’s best hunters can’t seem to catch or kill the beast. They could probably use some help.

We could easily imagine Speaker Huddle of Lonelywood sending a diplomatic mission to the newly ensconced Speaker of Good Mead: Having heard the success they’ve had with the verbeeg raider, she’s hoping they can send help to Lonelywood. In exchange, she promises to give Good Mead a discounted rate on Lonelywood’s lumber.

Or maybe it’s not Speaker Huddle. Maybe local loggers have lost confidence with her and have sent their people to extend a similar offer to the PCs. Maybe the PCs end up conspiring with the loggers to oust Speaker Huddle, with another of the PCs taking her place. Desperate times call for strong men, right? So maybe this whole thing ends with one of the PCs rising to become the new King of Ten-Towns… but at what cost to their souls? Maybe this is what the devil supporting the Knights of the Black Sword wanted to happen all along!

ON THE ROAD AGAIN

However this might turn out, we’ve clearly moved away from the baseline structure of Rime of the Frostmaiden. So, for the sake of argument, let’s say this doesn’t happen: Maybe the PCs just aren’t interested. Or maybe one of the PCs becomes elected Speaker, but the others decide to continue adventuring (with the player of the Speaker starting a new character; maybe picking up one of the NPCs who fought by their side against the verbeeg).

In any case, they continue down the road to DOUGAN’S HOLE. Here we have a scenario hook in which the white wolves plaguing the town meet the PCs on the road and try to lure them back to their lair (p.54). But the PCs, having heard about them in Good Mead, know not to trust them. They kill one of the wolves and the other one runs away (as described in the book).

Reaching Dougan’s Hill they’re told people have been kidnapped by the wolves, so they track the wolf that escaped, rescue the prisoners, and are now 3^rd level. They also hear that there are adventurers in Targos planning an expedition to Kelvin’s Cairn.

But still no Sephek. And (I roll a 35) no chwingas, either. So they head back up towards the Eastway. They come back to the intersection and need to choose between Bryn Shander (where they’ve got a quest rumor) and Easthaven (where the killer has been known to operate).

It’s still a toss-up, in my opinion. Players could rationalize either choice pretty easily. (They might also head all the way back to Caer-Konig to see if there’s any chwingas there, but this seems like a marginal possibility to me at this point.) Rolling randomly, it looks like this hypothetical group is heading to BRYN SHANDER.

As they enter town, they’re approached by three dwarves who would like their help recovering a sled shipment of iron that they had to abandon during a yeti attack (p. 34).

The PCs do that, completing their fourth quest. They also pick up two more rumors at the Northlook Inn: Kobolds have invaded the gem mines of Termalaine. And people are also talking about how no one has seen the town speaker of Caer-Dineval for weeks now…

Huh. That’s weird, actually. The people in the castle were very nice, but now that you mention it, we never actually saw the Speaker did we?

CONCLUSION

At this point, I’m not sure what our hypothetical group will do next. Lots of options, though:

• Maybe they’re running low on coin and decide rescuing a gem mine from kobolds in Termalaine sounds profitable.
• They might double back to Caer-Dineval to check out what’s really going on with their “allies.”
• They might try to backtrack the goblins who stole the dwarves’ iron.
• Before leaving Bryn Shander they might stop by the local shrine to Amaunator and speak to a gnome tinkerer who asks them to check in on his friend who lives in an abandoned cabin north of Lonelywood (p. 33).
• Or they could just continue down the road to Targos, searching for fairies and serial killers.

There are a couple key things to note here:

First, looking over these events, you can see how easy it would be to end up with a completely different campaign. A different starting town; a different decision by the PCs; a different random die roll; a different moment of creative inspiration and everything is transformed. This is not just interesting and exciting, it is also empowering. The players can feel the difference, and it’s intoxicating.

Second, the level of emergent complexity that we see here — the event horizon beyond which you can have no clear vision of what the campaign will be — is inherent to true sandbox play. Do the PCs become political leaders? Run a tavern? Become security guards for a logging consortium? Start a trading company? Mount archaeological expeditions to explore giant ruins? There’s no way to know and only one to find out!

(This is also why Rime of the Frostmaiden collapsing the sandbox of early levels into a more-or-less linear plot at the middle levels is rather disappointing. At the very point that the limitless potential of the sandbox begins to truly explode, the book instead says, “Okay. That was nice, but let’s lock it down.”)

If nothing else, I hope you’ve seen here that there’s nothing magical or even particularly difficult about running a sandbox campaign: After the sandbox has been filled with a selection of simple toys (some NPCs, some dungeons, some bad guys), all you have to do is observe a fairly simple procedure and follow the players’ lead, responding to their actions by picking up the appropriate (or most convenient) toys and actively playing with them.

Go to Icewind Dale Index

Go to Icewind Dale Index

The opening chapters of Icewind Dale: Rime of the Frostmaiden are designed as a sandbox.

And I mean a true sandbox. The term is often abused to just mean any non-linear scenario, but Icewind Dale is the real deal: A campaign in which the players are empowered to either choose or define what their next scenario is going to be.

As such, it’s going to be the first sandbox experience for a lot of new Dungeon Masters and players. And that’s great. Unlike the “hexcrawl” in Tomb of Annihilation — which was so dysfunctional in trying to force a linear plot into a non-linear structure that new DMs running it seem to frequently know less about how a hexcrawl works than they did before they started — I think Icewind Dale will introduce sandbox play to a lot of new players in a really positive way.

Now, I do have some quibbles:

• It’s a little half-assed, collapsing into a linear plot at higher levels.
• The advice given on how to run the campaign is inconsistent and, in some places, simply bad.
• There are a number of places where I think very small adjustments can create very large improvements in the experience at the table.

But the Icewind Dale sandbox fundamentally works for as long as it lasts.

As such, it also offers a great opportunity to see what a sandbox campaign looks like in actual play. And, because the book could be a lot clearer about some of this stuff, I’ve seen a fair number of DMs on social media expressing confusion. So let’s take a closer look.

SPOILERS FOR RIME OF THE FROSTMAIDEN!

SANDBOX STRUCTURE

Here’s how Icewind Dale: Rime of the Frostmaiden works:

• There are ten towns in Ten-Towns. Each one has a quest associated with it, and these are almost universally proactively triggered — i.e., if the PCs go to the town, they’re supposed to get the quest associated with that town.
• There is a Stage 1 rumor table (p. 18). Each rumor on this table points at one of the town-associated quests.
• There is also a Stage 2 rumor table (referred to as “tall tales” here, p. 102) which is unlocked at either 3^rd or 4^th level (depending on which page you’re looking at). These rumors point to higher level quests located in the wilderness around Ten-Towns.
• Finally, there are two Starting Quests. These are to be given to the PCs at the beginning of the campaign, and each is designed to motivate the PCs to travel to other towns.

For example, the “Cold-Hearted Killer” starting quest instructs the DM to:

1. Randomly generate a town. That’s the starting town for the campaign and will also be where the PCs get the “Cold-Hearted Killer” scenario hook (to track down a serial killer targeting victims throughout Ten-Towns).
2. Randomly generate a different town. This is the town where the serial killer will be found.

Although not explicitly stated in the book (which seems to have caused some confusion), the intention here is fairly clear: In following the starting quest, the PCs will go to another town and investigate to see if, for example, the serial killer is there. In that town they’ll scoop up the quest associated with that town and, in many cases, additional rumors pointing to quests in other towns.

This essentially becomes the default action of the campaign: If in doubt, go to a town and look for your starting quest objective.

The final element here is how the campaign handles milestone advancement:

• The characters advance to 2^nd level after completing their first quest.
• They advance to 3^rd level after completing three Stage 1 quests.
• They advance to 4^th level after completing five Stage 1 quests.

(Advancement gets fuzzier after the PCs reach the Stage 2 quests.)

QUIBBLES & ADVICE

Now, I do have a couple of quibbles here, and I’ll take a moment to call them out:

First, the book instructs the DM to only use one of the starting quests. I disagree. You should use both starting quests. Partly this is logistically superior: Each starting quest ends essentially at a random town. (The other starting quest, for example, has the PCs hunting chwingas — a type of fairy — and there’s a 25% chance that any town they visit has chwingas in it.) Having both quests in play makes it less likely for the PCs to exhaust their starting quests after just one town.

But it’s also just a superior experience: Having both quests in hand will deepen the default interactions with each community. It also breaks the players of the expectation that they’ll be doing a linear set of assigned tasks.

My recommendation is to give one starting quest in the first town visited and the other starting quest in the second town visited. (This creates a motivation for the PCs to potentially backtrack to the first town — either immediately or at some later time — to investigate the second starting quest there.)

My second quibble is how the book handles the rumor tables. Its advice is at best inconsistent (in some places suggesting that rumors should be doled at one at a time; in other places saying that specific taverns can deliver two rumors at a time).

Here’s what it boils down to: The point of a rumor table in a sandbox is generally to give the group LOTS of rumors (i.e., scenario hooks) that they can then choose between and prioritize. (See Juggling Scenario Hooks in a Sandbox.) There’s a much longer discussion to be had here, but briefly I would recommend:

• Delivering 1-2 rumors whenever the PCs investigate a town for their starting quest(s); i.e., tie the rumors into the default action of the campaign.
• Deliver 1-2 rumors any time they’re hanging out in a tavern. (This may already be covered by how they’re investigating their starting quest.)
• Opportunistically drop rumors as part of miscellaneous conversation with NPCs. People gossip about current events, so rumor tables almost always make for good topics of casual chat.

The other key thing here is to make sure that the rumors come from actions the PCs are taking (i.e., chatting in a tavern, canvassing the town for information, etc.). You shouldn’t just arbitrarily say, “Lo! Thou hast heard a thing!”

I’d probably also discard the distinction between Stage 1 and Stage 2 rumors. I don’t think it’s necessary, particularly for players experienced with sandbox play. If your players are conditioned to linear play, though — and might therefore heedlessly plunge into danger they’re not ready to handle — you might want to:

• Keep the proscribed unlock condition.
• Make sure the first 3-5 rumors they get are from the Stage 1 table (which will bias them towards completing those tasks and leveling up first).
• Start by dropping non-actionable versions of the Stage 2 rumors.

For example, one of the Stage 2 rumors points the PCs to the Jarlmoot — a hill where the frost giant leaders once met to settle disputes. What you could do at lower levels is just have an NPC mention the old jarlmoots of the frost giants (without telling the PCs where they can find that hill). Maybe the PCs go hunting specifically for that information, or maybe the foreshadowing just pays off later when they learn the location through another rumor drop. Either way, the depth of the experience makes the game world feel more real and meaningful to the players. (See, also, Random GM Tips: Getting the Players to Care.)

Go to Part 2: A Simulated Campaign

There’s a particularly prevalent — but completely incorrect — belief wandering around that sandboxes don’t have scenario hooks.

To the contrary: A good sandbox has scenario hooks hanging all over the place. The successful sandbox will not only be festooned with scenario hooks, it will also feature some form of default action that can be used to deliver more hooks if the players find themselves bereft of interesting options.

For example, a typical hexcrawl sandbox features a rumor table (which serves up some arbitrary number of scenario hooks to the PCs) and a default action if none of those rumors sound appealing (wandering around the map until you find something interesting).

A megadungeon sandbox similarly features a rumor table and a default action (go explore some unknown part of the dungeon).

Prepping this plethora of scenario hooks can be daunting for a GM who believes that every scenario hook needs to be linked to a distinct, unique plot. The trick to a sandbox is that you don’t prep plots: You prep situations. And for the sandbox you’ll be able to hang countless hooks off of every situation. You’ll also discover how sandbox situations “stay alive” even after the PCs have interacted with them (instead of being completely chewed up and discarded).

For example, let’s say you’ve got a dungeon a fair distance outside of town that’s the remains of a Neo-Norskan temple complex. It’s currently being occupied by a Bandit King who has forged together an alliance of humans, goblins, and ogres. He’s also renting skeletons off a nearby necromancer.

In terms of scenario hooks, there’s all kinds of stuff you can hang on this situation: Bandit raids are terrorizing local villages. A powerful magical artifact was stolen from a local caravan. There are old legends about the Neo-Norskan temple and what it contains. Because of the skeletons, there are false rumors that the necromancer lives there. Or that the necromancer has allied with the Bandit King. (And you can salt these scenario hooks into the campaign in any number of ways: Rumor tables. Lore recovered from other locations. Allies of the PCs who are now in need. Et cetera.)

So one day the PCs grab one of these hooks and they go off and they kill the Bandit King and they take the magical artifact he was carrying.

Over and done with, right? Only not really, because the guy who originally owned the magical artifact still wants it, so now the PCs are getting attacked by bounty hunters attempting to recover the artifact. Meanwhile, they didn’t wipe out all the bandits and the remaining goblins are renewing their raids under the leadership of the One-Eyed Ogre.

So the PCs go back to the Neo-Norskan temple and this time they wipe out all the bandits, permanently ending their threat to the region. Except now the Necromancer sees a big, open dungeon complex filled with the discarded corpses the PCs have left in their wake, and so he moves in and animates the corpses as a skeletal army.

Which all sounds like a lot of work, but because you prepped the whole thing as a situation to begin with you haven’t needed to spend more than about 5 minutes “refreshing” this content between sessions: You’re reusing the same maps and stat blocks over and over again. You spent a little time putting together new stat blocks for the bounty hunters when they showed up. And there was probably some light re-keying necessary for the changes the Necromancer made when he took over the complex.

You didn’t have to buy a whole new set of tools every single time. You just occasionally added a new tool when necessary. (And occasionally removed a hammer that the PCs had broken.)

This can be easier to visualize with a location (which is why I use it as an example), but the same basic process holds true for, say, factions in an urban campaign. Create a gang that’s, for example, manufacturing and marketing a drug derived from blood that’s been harvested from vampires and you should be able to use that toolkit to generate dozens of sessions of play.

The other thing that happens in a sandbox campaign is synergy between the different elements of the sandbox: By holding onto the artifact that was stolen from them, the PCs make enemies of House Nobuzo. This unexpectedly earns them a patron in the form of House Erskine, unleashing a flurry of scenario hooks from the “feuding noble houses” toolkit you designed. As the PCs get drawn into that world, they’re approached by a minor house named Tannar: They’re currently allied to House Nobuzo, but their daughter has been murdered by the Necromancer who has now stolen her body in order to transform her into his Corpse Bride. If the PCs can rescue their daughter from a fate literally worse than death, they’ll break their alliance with House Nobuzo and pledge for House Erskine.

After that scenario has resolved itself, you might find that the players are now actively looking for minor houses that they can endear to their political causes by doing favors for them. (Which would organically create a new default action for delivering scenario hooks.)

In any case, once your sandbox toolkits start interacting with each other like this, you’ll quickly find that the sandbox is basically running itself.