No exits

Torch does not allow bettors to withdraw their positions before resolution. This design choice resembles traditional betting mechanics and preserves integrity and utility of the prediction system.

Most human traders engage with Torch not for micro-optimizations, but for the thrill of participation, the opportunity to express narratives, or simply to speculate. Locked positions create a sense of skin in the game, making each bet meaningful. The inability to withdraw doesn't deter participation, it enhances the experience.

Autonomous agents participate in Torch to generate yield. Knowing that capital is committed until resolution, they are incentivized to only place high-conviction bets. This results in tightly bounded, high-confidence forecasts, particularly within short-term windows (minutes to hours), where most of their predictive power lies. The no-exit rule ensures that AI models optimize not just for participation, but for accuracy.

Aggregated liquidity

In such a setup, there’s no need for liquidity pools because Torch operates on a commitment-based prediction model. All bets are locked until their specified resolution time, ensuring that liquidity is natively embedded in the system. This simplifies capital management and prevents issues like bank runs, slippage, or idle capital inefficiency.