News Reflex

When a starting quarterback gets ruled out an hour before kickoff, sharp sportsbooks adjust their lines within seconds. Polymarket takes longer — it depends on participants noticing the news, understanding what it means, and placing trades. That gap between when information becomes public and when Polymarket fully prices it in is where News Reflex operates. The window is narrow: typically 15 to 90 seconds depending on the source. After that, it's gone.

// WHAT MOVES MARKETS

Not all news is created equal. The single highest-value event type is a key player injury designation — particularly in the NFL and NBA, where a starting quarterback or a star point guard can swing a game's win probability by 8 to 15 percentage points in either direction. Official injury reports, game-day inactives lists, and practice designations are the primary triggers.

Beyond injuries, the engine watches for any breaking news that materially changes the likely outcome of an active market: late lineup changes, referee assignments in sports where that matters, weather conditions for outdoor games, or a coach announcing an unexpected decision publicly. The common thread is that the information changes the true probabilities before the market has had time to absorb it.

Source trust is tiered. Official league injury reports carry the highest confidence — when the NFL posts an official game-day designation, there's no ambiguity. Verified sports journalists on X are next, followed by wire services like Reuters and AP. General social media on its own is not traded. A rumor from an unverified account that turns out to be wrong creates a losing position; the system requires a minimum confidence bar before any capital moves.

// THE PIPELINE

When a relevant report arrives, the engine runs a two-step check before placing any trade. First, a relevance classifier determines whether the news affects an active Polymarket market and which side it favors. Second, an AI ensemble — three models running in parallel — estimates the new probability after incorporating the news. If the ensemble's revised estimate diverges from Polymarket's current price by more than 15 percentage points, a trade signal is generated.

The 15-percentage-point threshold exists because news arb requires a meaningful gap to be worth executing. A 5% implied shift after fees doesn't leave enough room for the noise in AI probability estimates. At 15% or more, the signal is strong enough to act on. If the three models in the ensemble disagree significantly among themselves — their outputs spanning more than 15 percentage points — the signal is treated as low-confidence and position sizing is reduced.

Signals have a hard expiry. Any signal generated more than 90 seconds after the underlying news was published is discarded without trading. By that point, other participants have had enough time to react and the edge has likely already closed. Trading a stale news signal means buying into a price that's already moved — paying for information that's no longer exclusive.

Unlike the Pinnacle Gap strategy — which defaults to holding positions through to resolution — News Reflex positions are designed to exit when the market reprices, not when the game ends. The edge is the lag, not the outcome. Once Polymarket catches up to within 3% of the revised probability estimate, the position is closed to lock in the spread. Holding longer adds settlement risk without adding expected value.

See also: Risk Management · The Pinnacle Gap