The Attention Gap

The finding

Prediction markets do not reprice on the news you would expect. Put 618 news sources on one scoreboard, does the price move at all after a story lands, and the four biggest wires come out under the line with three niche specialists above it. What that measures is not the newsroom. It is where the market’s attention already sits.

Across 476,000 scored pairs the price had moved a cent or more an hour later 15.2% of the time, the 1.0x line. Score each source against the same markets the others wrote into and 15% of the separation is left, with the order broken. What sorts these names is mostly which markets they write into.

The market reprices where its attention already is, not where the pecking order says it should be.

The Attention Gap · Vera Research

The method

Fifty-six days of Polymarket data ending 25 June 2026: 86,060 stories matched to 3,434 markets, 476,000 scored pairs from 618 sources. Each pair records one bit. An hour after the story landed, had the price moved by a cent or more? A source’s score is its share of moved pairs over the all-source share of 15.2%, with thin records pulled toward that share. No magnitude and no direction: the population is every scored pair whichever way the price went.

The sources do not cover the same markets

The obvious objection is that this measures the markets, not the sources, and it partly does. How often a market reprices depends enormously on its subject: the price moved after 21.8% of geopolitics pairs and 6.2% of technology pairs, and across the fourteen categories the range runs from 0.3% to 25.4%. The sources do not write into the same mix: @IranIntl_En is 93% geopolitics, while no single category is even a third of any wire’s output. So we rescored every source twice more, against its own category mix and against the same individual markets other sources wrote into.

The ladder on page 1 reads off this table. Category mix leaves 51% of the gap, and every specialist still sits above every wire once it is out. The same-market control leaves about 15%, and the order does not survive it: @IranIntl_En lands below two of the four wires. What sorts these names is mostly which markets they write into.

What this does and does not show

What is left after the controls is small, and the measure cannot say which of two mechanisms produced it. A market is far likelier to reprice when several stories land on it in the same hour, 29.7% of pairs against 10.7% in a quiet hour, and the specialists arrive into those busy hours more often than the wires do. That fits a market repricing where attention is already gathering. It fits equally a wire’s story being largely in the price by the time it lands. Counting arrivals and price changes cannot tell the two apart. Raise the threshold to five cents, or move the horizon to ten minutes or two hours, and the blocks keep the same order. Three of the four wires hold still across both halves of the window. The Financial Times does not: 1.20x in the first half, 0.83x in the second, and it is the one wire that crosses the line, to 1.09x, once category mix is out.

The takeaway

1. A prediction market reprices where its attention already sits, not where the newsroom pecking order says it should. Measured over 476,000 scored story-market pairs from 618 news sources, 56 days on Polymarket ending 25 June 2026.
2. The all-source rate is 15.2% of pairs, and that rate is the 1.0x line. All four of the biggest financial wires sit under it: Bloomberg 0.70x, Reuters 0.74x, Wall Street Journal 0.75x, Financial Times 0.90x.
3. Three niche specialist accounts sit above it, all between 1.5x and 1.6x, on the same measure: @financialjuice 1.55x, @IranIntl_En 1.55x, @solidintel_x 1.60x.
4. The sources do not write about the same markets, and that is most of the story. Category mix leaves 51% of the gap and moves the Financial Times to 1.09x, at the all-source rate. The same-market control leaves 15%, and the order breaks.
5. The system had started tier-1 wires at a 1.3x weight because they are tier-1 names. Measured on this scoreboard, the Wall Street Journal comes out at 0.75x.

The multiplier records one thing a market can be seen to do after a source publishes: how often its price moved by a cent or more within the hour. It records neither how much the price moved nor which way it went. It is a record of past market behavior, not a forecast of any outcome.

Method and data

Built on the Vera dataset: 476,000 scored story-market pairs from 618 sources, 56 days on Polymarket ending 25 June 2026. The 1.0x line is the all-source rate, 15.2% of pairs. The multiplier records how often a price moved by a cent or more in the hour after a source published, never how much and never which way.

Odds shown are each market’s own price at the time Vera flagged the story, stated as fact, not advice. Nothing here is a recommendation to buy, sell, or trade any market or asset. Vera and Crypto Briefing are not registered investment advisers.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.