Saving a table’s time means deciding what the table no longer gets to decide.
A computer remembers every modifier. It tracks every spell slot.
It makes the attack roll, compares it against armour class, applies the result, and moves on before anybody has found the right die under the chair.
That’s not a small virtue.
It’s the difference between a rule being available and a rule becoming a pause in the evening.
But the same machine needs its answer before the question arrives.
Keep resolving. Keep displaying. Keep advancing the state.
The contradiction is easiest to see in the earliest licensed adaptations.
An engine turned turn-based combat, hit points, saving throws and character tracking into software.
It did that well.
It did not automate the referee’s judgment.
It replaced that judgment with fixed, pre-written scenarios.
This is automation as subtraction.
A combat-results table has its answer waiting inside it.
A player who decides to bargain with the monster does not.
Nor does one who talks past a suspicious guard using an odd combination of props, or turns an overlooked room into the whole plan for the night.
The fixed scenario gains a great deal in exchange.
It’s quick, repeatable and testable.
It gives every group the same encounter without requiring one person to invent what happens next.
It gives up the thing that made the encounter a table rather than a board.
Somebody at it can decide the rulebook didn’t exhaust the situation.
Modern adaptations make the bargain more fluent without making it disappear.
Baldur’s Gate 3 runs its rules loop faithfully.
Rolls against armour class. A second roll for advantage and disadvantage. Spell slots tracked. Multiclassing at the right thresholds.
These aren’t gestures toward tabletop play.
They’re rules made legible at speed.
Then the adaptation edits them, and the edits are where the interesting part lives.
Drinking a potion became a bonus action rather than a full one, a change the tabletop rules team later adopted.
The default dice are quietly weighted away from long runs of very bad or very good luck.
They don’t sample each roll as pure independent randomness.
That isn’t failed fidelity.
It’s a claim that repetition becomes a different experience when a program has to display it, pace it, and make it bearable.
Nobody sits across the table to decide the evening has had enough of the same misfortune.
A ruling becomes a rule by becoming a branch, a condition, or a prohibition.
The program accepts what it was prepared to accept.
Everything else has to become impossible, decorative, or somebody else’s problem.
Interactive narrative has been inside this accounting problem for a very long time.
Typed input made the invitation feel enormous.
The state underneath still had to recognize whatever arrived, and mostly it didn’t.
The gamebook made the same bargain in the open.
A two- or three-option fork feels like freedom, because each choice is presented in the grammar of a decision.
The production record is less sentimental about it.
Early books in the big branching series reached more than forty endings.
Later ones fell below ten.
A tree looks generous from the page where it splits.
It looks expensive from the page where every path needs another room, another consequence, and another ending that can’t borrow the furniture from its neighbour.
So the tree becomes a graph.
Branch-and-bottleneck design lets choice diverge, then brings it back to a fixed plot point.
It doesn’t make choice fake.
It makes choice payable.
The scale involved is genuinely strange.
A modern branching game can run to thousands of script pages, with a comparable volume of notes existing purely to track which page can follow which.
One writer-director compared the structure to a Rubik’s Cube.
Which is generous. A Rubik’s Cube can be solved.
The branch isn’t a shortcut around authorship.
It’s authorship multiplied, then carefully made to meet itself again.
That is the price of carrying a referee’s possible answer into a fixed work.
A player-authored exception becomes a pre-authored option.
A human response becomes a flag.
A conversation becomes a state machine with better lighting.
There are ways to resist the reduction, and the honest ones all reintroduce a person.
Some adaptations shipped a referee client. A live human ran the session on top of the automated combat and character engine.
The software processes the known rules.
The human keeps the right to interpret the unknown.
That arrangement gives up a different kind of scale.
Someone has to be present.
Someone has to prepare, notice, improvise, and occasionally decide that a clever plan is clever enough to work.
The machine can host the table.
It can’t turn the table into a product that runs unattended, and unattended is what a product mostly means.
Experimental work has widened the invitation without closing the gap.
A system can accept unscripted natural language and let a player intervene in a dramatic scene however they like.
That remains striking, precisely because it’s wider than a dialogue wheel.
But accepting unanticipated words is not the same as interpreting unanticipated intentions.
No amount of parsing has closed that distance.
Simulation offers another escape hatch and another limit.
In a deep enough simulation, an intoxicated creature starts a brawl and causes a death nobody scripted.
The incident is real, and it’s genuinely new.
Elsewhere a director system watches performance and adjusts later spawns, producing pacing rather than incident.
One generates events. One generates rhythm.
Neither is a referee hearing a proposal and deciding what it means here, now, for these people.
Licensing made the distinction commercial long before it was aesthetic.
The right to automate a rules system was never merely permission to type the rules into code.
It was permission to decide which parts of the table would count as the game.
Automation gains speed, consistency, accessible state, and an evening that runs without appointing one player as unpaid civil service.
It gives up the odd plan that becomes memorable precisely because nobody had written it down.
Human judgment gains responsiveness, surprise, and a story authored by the people present.
It gives up repeatability, immediate legibility, and any guarantee that the same rules carry cleanly from one table to the next.
A fixed branch can make a system shareable with strangers.
A live referee can make it alive in a way no branch anticipates.
Both are adaptations.
Both leave something on the floor, and only one of them has to admit it in a manual.
Which judgments does a tabletop system become more itself by losing, and which losses make it a different kind of game?