At 10:04:12 an AP push hits; by 10:04:45 a prediction contract reprices from 41% to 57%. That 35-second window is where game theory, microstructure , and information economics collide.
This chapter explains why news moves prices instantly, who wins the race, and how to interpret jumps without fooling yourself.
Information arrival as a shock
Treat price as filter of belief distribution. News is sufficient statistic shift for many traders—posterior probability jumps even if you have not processed the article yet.
Bayesian jump (conceptual):
- Prior 40%
- Likelihood ratio from headline large
- Posterior 55%
Market implements posterior via orders, not via survey.
The microstructure timeline
the microstructure module CLOB: lifts ask stack.
the microstructure module AMM: moves curve.
Who wins the jump
Winner’s curse for speed traders: fastest buy lifts ask; if news overreaction, they sell back down to slower crowd.
Jump size vs surprise
Surprise = news vs prior market-implied expectation.
- Expected debate gaffe → small move
- Surprise indictment → large move
Compare move to options-implied or poll-implied surprise metrics in finance; prediction markets analogous.
Partial information jumps
Headline: “Deal reached” without text. Traders split:
- Skeptics fade (maybe non-binding MOU)
- Believers buy (assume passage)
Wide spread after headline = uncertainty, not consensus.
Fake news and false jumps
False tweet → spike → revert in minutes when debunked.
Game:
- Attackers profit if they positioned pre-tweet
- Platforms halt trading or widen bands
- Oracles later ignore hoaxes for settlement
Trader rule: size inversely to source credibility lag.
Scheduled vs unscheduled
Scheduled (CPI, Fed, debate start):
- Makers widen spread pre-release
- Volatility auctions
- Less edge for slow retail
Unscheduled (scandal leak, health event):
- Larger jumps, wider spreads after
- Arb slower if rules ambiguous
Volume signature of “real” news
Healthy jump profile:
- Volume multiple of baseline across venues
- Correlated move in sister contracts
- Spread temporarily wide then tightens
Unhealthy (possible cascade/manipulation):
- One venue only
- Tiny notional
- Revert in <10 minutes without clarification
Expected value of speed
Investing $50k/yr in faster feed vs edge per jump × frequency:
- Political junkie retail: negative EV
- Multi-venue arb firm: positive EV at scale
Most forecasters should improve model not latency.
After the jump: drift vs revert
Drift — new level correct; information permanent.
Revert — overreaction; fade profitable.
Discriminate using:
- Subsequent official confirmations
- Poll shifts next cycle
- Structural arb still violated?
No formula—process beats reflex.
Media quoting jumps
“Markets swing to 70%” from one thin print harms public understanding. Better:
- Timestamp
- Composite mid
- Depth at move
Game with oracles
News near resolution may not move tradeable market—oracle rules freeze trading. Jump game ends; resolution risk begins.
When an alert fires, rank the source tier (wire services above random accounts), check whether the move was already priced on the chart, seek cross-venue confirmation, sanity-check sister contracts, and remember your edge is usually interpretation, not milliseconds.
What comes next
Next: Incentivized reporting and honest consensus at settlement.