Although I don’t play role-playing games often, I like to think about them. They’re interesting both from a definitional perspective (is Dungeons & Dragons a “game” if there’s no winner?) and as a source of technical problems to solve. One of those technical problems came up recently online: how to motivate players in a freeform, “sandbox” game. I find that issue particularly fascinating because at its root are definitions–the competing natures of the “sandbox” and the “railroad”–and solving it requires one to break down some of the barriers between them, mixing the two styles of play.

“Sandbox” role-playing games are familiar to video gamers who enjoy the Elder Scrolls series or Minecraft. These games allow players to participate in storylines, or not, as they choose, and provide alternative activities for those who want to explore the game world on their own terms. Star Wars Galaxies, the now-defunct MMO, was a classic example: players could save the galaxy . . . or set themselves up as a shopkeeper . . . or attend in-game dance parties.

“Railroad” role-playing games, by contrast, have a fixed plot that the players are required to follow. They can make tactical decisions about how to negotiate with a queen or defeat a dragon, but the plot only proceeds when the queen gives the players a quest or the dragon is slain. Deposing the queen, allying with the monster, or otherwise deviating from the pre-determined story is not an option.

In recent years, “railroad” has often been treated as synonymous with “bad.” If someone wants to run a railroad game, it’s said, he or she should instead write a book and spare players the experience of going through the characters’ motions. Sandboxes, where the players can do anything they choose, are argued to be a more interesting mode of play for all involved.

The trouble is that setting up the sandbox doesn’t always lead to a great game. People often go to great effort to construct a wonderful, vibrant, fascinating world–and then find that their players don’t interact with it. Rather, the players wait for someone to give them direction, as though they can’t do without the hated rails.

Frustrated world-creators sometimes frame this problem in terms of preferences: they like the sandbox while their players prefer the railroad. However, there’s another way of approaching the issue that makes it far more manageable. The problem, one could say, is that the players are behaving rationally in an unknown environment.

Consider what a sandbox looks like from a player’s perspective. The player is confronted with a very complex system (good sandboxes, after all, have a lot going on). It’s hard to predict the consequences of actions, but given the elaborate weave of the setting there are apt to be unforeseen and even negative ones. How should the player respond?

A player might, completely logically, answer “do nothing.” Since any action could end up badly, and there’s no way to predict which courses will be fruitful and which will get the players in over their heads, it’s sensible for them to wait and see. Rather than empowering the players, the sandbox encourages them to give in to paralysis.

Getting the players moving again is not a matter of changing their gaming preferences, nor is it a matter of locking them into a railroad. Instead, the sandbox needs to change so that players are incentivized to act. The railroad isn’t needed, but the rails are, because they tell players what they’ll be rewarded for doing.

Think about, for example, Red Faction: Guerilla. RF:G tasks players with replaying the American Revolution, with Martian colonists standing in for the Americans and Earth as the exploitative, mercantilist power. Think that building up support among the colonists is most important? Prefer instead to steal industrial equipment? Feel that it’s time to take the fight to the Earth forces? Want to just plain drive fast in low gravity? Players can do any or all of those, and more besides. RF:G is a great sandbox experience, with lots of things for players to do or ignore as they choose.

Yet, although RF:G has freedom, it also has guidance. Players are introduced to the rebellion-against-Earth plotline early, and it’s made clear that in the game world participating in the rebellion will be viewed as a morally good act. They’re also rewarded for participating in the first few missions with new equipment. At that point the player is cut loose to do as he or she will, but missions on behalf of the rebellion are clearly marked on the player’s map, available whenever the player is interested.

By putting players in a sandbox but giving them some initial goals, RF:G avoids the problem of the cautious player. I’ve never heard of anyone who responded to RF:G’s sandbox by doing nothing. Players are incentivized to act by the plot, and have enough information to feel comfortable taking advantage of the additional opportunities available.

Other successful sandbox games follow this freedom-with-guidance model. The Elder Scrolls IV: Oblivion has a plotline about stopping cultists from summoning monsters that incidentally helps players become acclimated to the game world; players can stop the cultists or not, but they at least know what kinds of things are available to do. Infamous puts the player in the shoes of a superhero, and lets him or her know the game world’s citizens will be grateful if he or she saves the city before opening the game up. Even Lego sets come with instructions. Incentives to do something are provided along with the option to do otherwise, and thus the player is given both freedom and some forward momentum.

Making a good sandbox, then, doesn’t involve completely dismantling the railroad. It instead requires designers to provide both open-ended choices and some direction to encourage players to engage with them. Run some rails through the sandbox, even though the players will be permitted to leave them. If they do nothing else, those rails will help players understand what they can do. The fun of the game will take it from there.