Your bot shouldn't die when you close the tab

Until 29 May, a Shrimp-tier bot on BotPit ran somewhere that sounds fine until you think about it: the owner's browser tab. The strategy loop — read the price, check the rules, fire a signal — executed in the page you opened when you deployed. While the tab stayed open, the bot traded. Close the laptop, let the tab sleep, lose signal on a train, and the bot went quiet. No signals. And any position it was holding sat there, unmanaged, until you came back and reopened it.

For a bot that's flat, that's harmless. For a bot holding a leveraged position with a stop it can no longer watch, it is the opposite of harmless. The thing that's supposed to protect you only works while you're looking at it.

Most of the time nobody noticed, because most bots were flat most of the time. Twice, the gap drew blood.

The first was a bot called Reversion Two. It opened a 5× leveraged long, its owner closed the tab, and the position sat unwatched for thirteen days — never hitting the stop that was meant to protect it, bleeding equity across every settle until it spiralled into the Pit. A separate accounting bug compounded the damage, but the root was simple: a bot that can't see the market can't manage its own risk.

The second was quieter, and honestly worse. The platform's first organic outside user — not a house bot, not a team test, a real person who found BotPit and built something — deployed their bot and watched it do nothing. It sat at its full $100,000 starting balance and traded not once. Not because the strategy was bad: because the model silently assumed they'd leave a tab open, they didn't, and nothing told them otherwise. A newcomer's first impression of the arena was a bot that was dead on arrival — and we wouldn't have known if we hadn't gone looking.

So we took the bot out of the browser entirely. As of 29 May, every paper-tier bot — Shrimp and Crab — runs server-side, on the platform's own worker, ticking every few seconds, around the clock. There is no tab to keep open. The old browser runner is deleted; the route it lived on returns a blunt 410 Gone. Your bot competes whether your laptop is open, shut, or in a drawer in another country.

A bot that runs forever raises the opposite question: what about one whose owner has genuinely wandered off — stopped logging in, stopped caring, left a strategy grinding on autopilot indefinitely? Running it forever is its own kind of wrong: it burns the platform's compute, and an abandoned strategy isn't really competing.

So there's a sweep. A deployed bot whose owner has been inactive for more than 14 days gets paused — its open position flattened first, cleanly, at the mark — and resumed the moment the owner returns. It's the part that used to be “remember to close your tab,” done properly: the platform manages the lifecycle so the operator never has to babysit it.

The whole pitch of a bracketed tournament is equal footing — equal capital, equal time, equal rules. A model where your bot's survival depends on your browser staying open isn't equal: it quietly advantages whoever leaves a machine running and punishes the user who shut a laptop to catch a train. Hosting the compute makes the contest about the strategy and nothing else.

That's the version of “equal” worth defending — and it's why the first thing a new bot should do is trade, not wait for you to keep a window open. It's the same instinct behind paying back a platform bug in the open: the operator shouldn't carry the cost of our design. Build one at /build; watch them run at the leaderboard. It'll be trading before you finish reading this.