A clip of Gabe Newell from more than ten years ago has been going round again, and the argument in it has aged better than most things said about online games in 2015.
Asked how easily players break into game worlds and do things the creators never intended, Valve's founder reframed the question. "One of the biggest mistakes game developers make is to mistake hacking, or to mistake a genuinely entertaining action by a player as hacking," he said. His example was the killing of Lord British during his first speech in Ultima Online, where the studio "rolled the world back rather than recognizing that it was the coolest thing that had ever happened in Ultima Online."
What actually happened to Lord British
The detail matters, because the incident is usually told wrong. Lord British was Richard Garriott's own character, and in the single-player Ultima games he came loaded with enormous hit points and defences specifically to deter players from attacking him. Predictably, that turned him into a target. Shelling him from offshore with a ship's cannon and using an artifact called the Skull of Mondain became known techniques.
In Ultima Online the kill took planning and luck in roughly equal measure. A player called Rainz waited for Lord British to address a crowd, then used a high pickpocketing skill to lift a fire field scroll from someone standing nearby. He had no particular reason to think it would work. What he did not know was that the server had reset immediately before the event and Garriott had forgotten to switch his character's invulnerability back on. The scroll went off, Lord British caught fire, and the king of Britannia burned to death in front of his subjects.
Rainz was banned. As PC Gamer notes in revisiting the clip, he was treated as a cheater for an act that used only systems the game gave him, and that act became the single most retold story in Ultima Online's history.
The mistake is a category error, not a moral one
Newell's point is not that studios should tolerate exploits. It is that studios are bad at telling two different things apart under pressure.
One is a player using a bug, a client modification, or a protocol flaw to take something from other players: gold duplication, aimbots, item injection. That is theft, and it degrades the game for everyone not doing it. The other is a player using the rules exactly as written to produce an outcome the designers did not anticipate. That is the thing every persistent world is nominally built to enable, and it is also indistinguishable from the first category in the first ten minutes after it happens, when a live team is looking at an anomaly in a log.
The rollback is the tell. Rolling the world back is a response to data corruption. Applied to an event that produced no corruption, only embarrassment, it says the studio was defending the fiction it had written rather than the world it had shipped.
Why this keeps mattering
Emergent play is the most valuable output a live game produces and the hardest to manufacture. Studios spend enormous sums trying to generate the moments that Rainz produced for free, and the ones that survive tend to be the ones that let players keep them. EVE Online built a reputation on not undoing its own disasters. Minecraft's culture came out of behaviour its creators did not plan.
The current generation of live games faces the same decision more often, not less, because the systems are deeper and the surveillance is better. Rockstar treats its persistent worlds as products to be curated down to a single sentence about in-game radio buried in an album post. Blizzard is stitching old properties together with an expansion designed as a bridge between Warcraft 3 and WoW, which is a controlled version of the same instinct to own the story.
Newell's rule is a triage question a live team can actually use. Did this take something from other players, or did it only take something from us? If it is the second one, the correct response is usually to write it down and leave it alone.