The overwhelming majority of arguments about roleplaying can be traced to this question. I think much of roleplaying's current market situation can too. Once we decide it, answers to numerous other questions become obvious. Why? Because different kinds of worlds suit these two play styles. For a sandbox game, a world in which the party can respond to countless details, each launching an adventure, is ideal. But if you lean toward safety rails and a fixed adventure module? All those things may lead them astray. Even where a middle ground is possible because characters are confined to a small environment, the differences between the approaches soon emerge.

If players have developed their characters' personalities in detail and always know where those characters would go, every time the path leads elsewhere disappoints them in the game's freedom. If we promise precisely that freedom to new players already coming from sandbox experiences, a few such disappointments become a serious problem. But if players are unfamiliar with sandbox play and you base the game on it anyway? Without a clear guiding thread, they may feel lost and bewildered, becoming bored without ready-made requests.

RoselandMoreover, distinguishing the two GMing styles is often harder for players than we think. A group plays Shadowrun with every character a robbery detective in a particular Lone Star precinct. A GM thinking in terms of safety rails considers how the adventure can be solved and sees one or more ideal paths for the characters. A sandbox GM? Folks, this case may not be the one that raises your clearance rate above 20%. But the game master knows what happened and how, and can use an appropriate simulation of the world to determine clues, witnesses, and so forth. Whether the party abandons the investigation to work on a colleague's home-renovation project instead, or pins six additional robberies on the perpetrator, is up to them. The GM cannot predict it. 

Without discussing sandbox or safety rails beforehand, players do not know whether solving that robbery is the adventure, or whether the party writes the story while the game master merely narrates the world. Stories can arise from house renovations, family drama, weekend drinking, or events surrounding a concert the children attended, as well as investigations. Not every story necessarily has a good solution or happy ending. And yes, sometimes the feds will catch the culprit just as the players are closing in. Life does not always deliver happy endings, and the world does not always favor situations appealing to the party. 

Is the executioner good or evil?

Sandbox enthusiasts dislike the “good/evil” simplification: varied personalities and countries offer countless adventure possibilities. Those preferring fixed adventures often need clear good and evil—fewer sides make it easier to write an adventure suiting everyone—and often enjoy illogical but atmospheric writing, which adds much atmosphere without its illogicality derailing the adventure.

There is a reason our class design does not aim at an “even more overpowered head-hunter,” but considers an executioner, prompted by a simple passing remark. Such a divisive character can initiate adventures through arguments within the party. Without reason for party members to slaughter one another, that can be good. Whether someone opens a book on medieval torture and interrogation instruments at the “pear of anguish,” Judas cradle, or breast ripper makes no difference: a knight who may agree with the goal but not necessarily the method will argue. Debate about what can be achieved and how can become an adventure. 

Simulating the world includes such details. And, as you can see, nowhere in discussing simulation did we mention combat modifiers supposedly pursuing “realism” while turning a fast-moving battle into “chess.” That is not the simulation needed. Attribute frequencies matter. Roughly realistic chances of success matter. But if the system answers more than “success or failure,” letting you easily and realistically describe many details after a roll, with the character learning from them and the next two adventures arising from that? Better still. 

Enthusiasts of these two approaches will not agree on what Gorvik should be like in M.A.G.U.S. Nor on how important it is to carry an adventure's consequences forward. Nor on what a priestess of Ellana, with combat values far better than those of the described soldiers and orks, can do. In a sandbox, you do not raise attackers' levels for her sake; she defeats many and may become a hero. With safety rails? Certain gentlemen authors apparently favor a more impotent approach for everyone except their own heroes, deciding Ellana's priestesses do not learn to fight, while good combat values and extra skills are incidental. Run the game faithfully to their Ynev, and raise every opponent above the priestess's combat values: the story is ready to go. 

In a sandbox, half a sentence in the wizard's description defines the characters' field of research, producing the adventure. Since they use magic for research, whether they cast spells is never in question: even the most conservative interpretation lets them cast for research. With safety rails? The same half-sentence and philosophizing over it take time away from the adventure and may obstruct it. You make a wizard to cast spells, and play out why they are like that. Anyone trying to confine you to research is an idiot. What guides a sandbox may obstruct safety rails. What is a little novelty or surprise in a safety-rails adventure is, to sandbox enthusiasts, a disturbing inconsistency leaving the character less at home in the world. 

All right. Some say there is powergaming sandbox, life-path sandbox, and safety-rails adventures...Miraculously, the description now has three poles, almost like GNS theory. Yet here the three directions really seem difficult to reconcile if an interesting life path does not give the powergamer the optimal reward. Someone fond of fixed, guided adventures can easily be content when the adventure is guided not by the GM, but by a sandbox-savvy player's life path. The greater problem comes when two people following sandbox principles would go in very different directions because the party lacks a strong shared concept. 

Many say sandbox GMing does not necessarily require writing an adventure, assuming it therefore requires less preparation. In reality, I have seen plenty of improvised adventures firmly belonging to safety rails. Good sandbox GMing requires preparing the story's location and numerous events, so I think it takes more preparation, even if preparing player-facing materials is difficult for this style. Many other things differ too, including player preparation. If guided through an adventure, a simple background suffices; malicious tongues say many manage without one. But without background, motivations, and goals, a player cannot even get started in many sandbox games, and every detail is worth its weight in gold. 

And yes: accusing me of firmly favoring sandbox play is justified. But we must also accept that most roleplaying sites favor safety rails, simpler in many respects. Adventures will appear here too, of course, because some ideas deserve writing down, and sometimes we too can use them with a new group. Our preference nevertheless remains.