Blog / Fifth edition rules
How to play D&D solo (and what changes when you do)
You can play D&D 5e on your own, and the whole problem is one thing. The sheet works with one person. The dice work with one person. The rules work with one person. What does not work is that you are now the only source of surprise in the room, and a roleplaying game is mostly the feeling of finding out. So every method of solo play is an answer to the same question: what, other than you, gets to decide what happens next?
There are five serious answers and they are genuinely different from each other. Here is how each one works, and what each one is bad at.
Method one: an oracle
An oracle is a randomiser you ask questions of. The core move: phrase what you want to know as a yes or no question, decide how likely a yes is, roll, and read the answer off a table.
The best known is the Mythic Game Master Emulator. You ask "is the gate guarded?", judge it likely, roll on its Fate Chart, and get back yes, no, or an emphatic version of either. It also fires random events on certain rolls, which is the part that matters more than the yes or no. A random event drops something into the scene you did not plan, and it is that interruption, not the die, which stops a solo game from becoming a story you already knew.
Good at: working with any system, including D&D 5e exactly as written. You bolt an oracle onto the game you already play. It also produces the strangest and most memorable material of any method here, because it has no taste and no plan.
Bad at: everything is interpretation. The oracle says "yes, and", and you still have to decide what that means, in character, without letting what you want to be true leak in. It is slow, it is a lot of table lookups, and it gives no help with pacing - a bad oracle session is a long sequence of true facts that never becomes a scene.
Method two: a game designed for solo play from the start
Some games build the oracle in rather than bolting it on. Ironsworn is the standard example: it ships with a solo mode as a first-class way to play, alongside co-op and traditional GM-led play, and its rules are free. You roll an action die against two challenge dice for a strong hit, weak hit or miss, you swear vows and track progress toward them, and the book carries oracle tables for names, places and plot turns.
The difference from bolting Mythic onto fifth edition is that the mechanics themselves generate story: a weak hit is not "you fail", it is "you get it, but here is a cost", and the cost is the next scene. The game constantly hands you complications you did not author.
Good at: momentum, and the absence of conversion work. Nothing in it assumes a party or a GM, so nothing has to be adjusted.
Bad at: it is not D&D 5e. If what you want is your fifth edition character, your subclass and your spell list, a different game is a different game, however good it is. Solo-first designs are also generally lighter on character build, which is precisely the part some people came for.
Method three: run a published module solo
Buy an adventure, read only the section you have physically walked into, and play it straight. A dungeon is well suited to this, because a dungeon is a decision tree somebody else already built: the map decides what is behind the door, so you do not have to.
This is no longer an unsupported way to play. The official 2025 anthology Dragon Delves flags several of its adventures as runnable for one player rather than the usual party, and includes an optional supernatural gift, the Blessing of the Lone Champion, that hands a lone character temporary hit points and more frequent Heroic Inspiration. That is a published acknowledgement that one player is a format worth writing rules for.
Good at: structure and content. You get maps, statted encounters and a plot, none of which you have to invent while playing.
Bad at: you know the secrets. You read the boxed text, so the twist is spent before you reach it. Social scenes are the weakest part, because a negotiation with an NPC whose motive you wrote down ten minutes ago is not a negotiation. And the encounters assume four characters, which is a problem this post comes back to.
Method four: play the whole party yourself
Roll up three or four characters and run all of them. This is the oldest solution and it is still the most reliable for combat.
Good at: fifth edition works as designed. The action economy is intact, the party covers its own weaknesses, and a fight is a real tactical puzzle rather than a sequence of saving throws you have to survive.
Bad at: surprise, again, and bookkeeping. Four sheets is four times the arithmetic. Worse, four characters you control tend to converge into one voice with four stat blocks. Most people who try this end up pairing it with an oracle so that at least the world is not also them.
Method five: an AI Game Master
Software that narrates: you write what your character does, in your own words, and it describes what happens. This is the only method on the list where the outside voice actually answers you in prose rather than in a table result you then have to interpret.
Good at: the interpretation load, which is the thing that ends most solo campaigns. There is no lookup, no conversion, and no moment where you have to decide what "no, and" means. It also plays NPCs who have not read your notes, which no oracle and no party you control yourself can do.
Bad at: two things, and both are worth knowing before you start. Tactical positioning is the honest gap - grid movement, cover, opportunity attacks and flanking assume a map, and text is a poor map. Rolegend publishes exactly which fifth edition rules it enforces, half enforces and does not enforce at /rules-coverage. The second gap is that no AI Game Master will plant something in session two that pays off in session twenty because it decided you specifically would like that.
What genuinely changes when you play alone
Whichever method you pick, four things are different, and they are different in ways that will end a campaign early if you ignore them.
You have no one to cover your weaknesses, so survivability matters more than damage. In a party, the wizard is allowed to have eighteen hit points because three other people are standing in front of them. Alone, every gap in your build is a gap nobody fills. Favour the durable end of the roster, take the defensive option when it is close, and treat a healing potion as standard equipment rather than a luxury. Starting a level or two above the adventure's recommendation is a legitimate fix and nobody will know.
You have nobody to think out loud at, so write it down. This is the single most useful habit in solo play. Keep a journal in your character's voice, or just a plain log of what happened and what you intend to do next. The point is not the record. It is that articulating a plan in writing does the job saying it to another player does: it commits you to a reading of the situation before the dice resolve it, which stops you deciding afterwards that you meant something else.
Encounter balance assumes four characters. A "medium" encounter for a party of four is a coin flip for one character, and not in an exciting way: the action economy is three to one against you. There are two honest fixes. Cut the number of enemies rather than only their hit points, because four attacks a round is the thing that kills you, not the size of any one of them. Or give yourself a companion: a sidekick NPC, a hireling, a familiar with actual utility. A second body on your side restores most of what the maths assumes.
You are both the player and the person deciding what happens, and that is a discipline problem. Nothing stops you rerolling. Nothing stops you deciding the guard was distracted after seeing the roll. The reliable defence is to make the decision before you can see whether you like it: set the difficulty out loud, or in writing, before you roll. Say what a failure will cost before you find out whether you failed. If you catch yourself asking the oracle the same question a second time, that is the tell.
Where an AI Game Master fits in that list
Honestly: it is the method that removes the most friction, and it is not the method that produces the most interesting worlds. An oracle will hand you a stranger campaign than any software will, if you have the patience to interpret it.
What it does solve is the two failure modes that kill solo games - the interpretation workload, and being your own referee. On Rolegend those are split deliberately. Keystone Engine enforces the rules: it rolls every die on the server, applies your modifiers, and keeps hit points, conditions, spell slots and inventory, so a number cannot quietly become what the story needed. ROWAN narrates the world and is handed the result afterwards, with no say in it and no way to roll again. You cannot lean on the die, because it happened before there was any narration to lean on.
That is not the same as a good human Game Master running a game for you, and it is worth saying so before you decide. It is patient, consistent, available at two in the morning and it never cancels. The best table you ever sat at was better than that. If you have one, keep it. If you do not, this is not a compromise version of a thing you already have.
If you want the trade itself argued out rather than the methods listed, what solo play costs you and what it gives back is the longer version of the section above.