Skip to content
Data story · sample data

The half-life of news.

Some news moves a market for an hour. Some moves it for good. A look at how different kinds of news fade, built from sample data.

FForeseez ResearchOct 9, 20267 min read
01

Not every jump is the same jump

Watch any prediction market for a week and you will see sudden moves. A rate market jumps three points, an oil market four. Some of those moves are still there the next day. Others are gone by lunch.

The difference is mostly about what caused the move. A data release is a fact: once it is out, it does not get less true. A rumor is a claim: it is either confirmed or it fades. Markets seem to know this, at least on average.

FIGURE 1How much of a move survives, hour by hour
0%25%50%75%100% 0h1h3h6h12h24h48h hours after the move
Scheduled datahalf-life 30 min88%
Official statementhalf-life 2 h74%
Breaking headlinehalf-life 4 h63%
Unconfirmed reporthalf-life 1.5 h14%
Average share of the initial move still in the price, by what caused it. Hollow dots mark each half-life. Time axis is stretched near zero so the first hours are readable. Sample data.

In our sample, moves driven by scheduled data keep almost 90% of their size a day later. Moves driven by unconfirmed reports keep less than a tenth, and most of the fading happens in the first few hours.

02

Watch it spread

News rarely stays in one market. An oil headline hits oil first, then the markets that depend on oil: inflation, the Fed, the economy. Each step is weaker than the last, and each fades on its own clock. Play it with a rumor, then with confirmed news.

FIGURE 2 · SHOCKWAVEAn oil headline, rippling out
Brent $90HormuzOPEC+ cutInflationFed cutRecessionS&P 500Bitcoin Unconfirmed report
+0h
A rumor: loud, then goneBrent jumps 4.5 points on the report, then gives most of it back within a few hours. By the time the wave reaches the Fed market, there is almost nothing left to carry.
Closer markets are more tightly linked to oil. Numbers show the move each market is still holding, in points.
A schematic, not a live feed. Link strengths and retention rates are sample values.

The wave reaches the same markets either way. What differs is whether anything is still lit up when it gets there. A rumor's effect has mostly faded from oil before it ever reaches the Fed market. Confirmed news arrives at full strength and stays.

03

Big moves are not more permanent

You might expect larger moves to stick better, as if size were a sign of conviction. In our sample it is not. What predicts whether a move survives is its cause, not its size.

FIGURE 3Size of the jump versus what is left a day later
005510101515 kept all of it initial move, points
Scheduled data88%
Official statement70%
Breaking headline42%
Unconfirmed report8%
Share of the initial move still in the price after 24 hours, median by type.
120 sample moves, 30 of each type. Points on the dashed line kept their whole move. Sample data.

The rumor dots hug the bottom of the chart no matter how big the initial jump was. Some of the largest moves in the sample came from reports that were never confirmed. Size told you how surprised traders were. It did not tell you whether they were right to be.

04

What we do with this

This is why Foreseez alerts treat news differently by type. A data release triggers an alert immediately. A headline triggers one with a developing tag and a second look an hour later. A rumor never pushes to your phone on its own; it waits in your feed until a second source confirms it.

FIGURE 4One move, four kinds of alert
Wednesday10:42
FORESEEZDATAnowFed cuts in December moved 4ppCPI matched forecasts. All venues moved within 40 seconds.
0sAlert sent

A fact does not get less true. The alert goes out as soon as every venue has repriced, usually within a minute.

Alert policy as designed. Timings are defaults you can change in alert settings.

None of this means rumors are worthless. Sometimes they are early. But a system that pushes every jump to your phone teaches you to ignore it. Sorting news by how long it tends to matter is how alerts stay worth reading.

All moves, retention rates and links in this post are sample data, built to illustrate the pattern. Nothing here is investment advice.