What the crowd priced

On Monday 6 November 2006, the day before the US midterm elections, an Intrade contract that paid out if the Republican Party kept control of the Senate closed at 73.5. Intrade quoted its contracts on a scale from 0 to 100, where 100 was the full payout, and 4,033 of these changed hands that day. On election day itself 30,417 traded, and the day closed at 69.8. By the end of the next day it stood at 1.1: Democrats had won the Senate, with two independents who sat with them.

It is tempting to call that a miss. Read as a probability, 73.5 left about one chance in four for the other side, and one-in-four outcomes happen. One market cannot settle the question. This edition shows what the record holds, ten Congress markets and two presidential maps among it, so that a price like this one can be read against the rest.

The record

Intrade ran from Dublin and listed contracts on almost anything with a checkable answer: who would win an election, what a film would take at the box office, how much snow would fall on New York City, how busy the hurricane season would be. The last daily prices in the record are dated 10 March 2013, the day the exchange stopped trading.

Panos Ipeirotis, a professor at New York University's Stern School of Business, collected the contract list, the daily summaries and the trade files, and put them on GitHub on 2 April 2013. In October 2026 he rebuilt the archive: he checked the original files against their hashes, converted the records to research tables, documented the gaps, wrote a rights statement and deposited it all on Zenodo with a DOI.

The archive lists 17,295 contracts under 22 category labels. 10,815 of them have trade records, 2,129,385 trades in all, and the daily summaries add 2,035,482 rows. Figure 1 shows what Intrade listed.

Who would control Congress

For every federal election in the record, 2004 to 2012, Intrade listed markets on which party would control the House and the Senate. Figure 2 shows each one as the exchange recorded it: the close of each contract on every day it was listed, and how many contracts traded. Pick a year and a chamber, move along the line, then reveal who won.

Take the close on the day before each vote. In nine of the ten markets that close favoured one party, and in eight of those nine that party won control. The exception was the 2006 Senate. In the tenth, the 2010 Senate, Intrade listed a third contract, Neither, and on the eve the Democratic and Neither contracts both closed at 45, with the Republican one at 10. The rules that said what control meant are not in the record, so this edition does not say which of them paid.

Eight of nine is not a test of anything. The markets are not independent, since a national swing moves the House and the Senate together, and most of them were priced near certainty: in five of the nine, the winning party's own contract closed above 90 on the eve.

State by state

For the presidential elections of 2004 and 2008, Intrade listed a contract for every state and the District of Columbia. Figure 3 colours each one by the close of the Republican nominee's contract on the day before the vote: red above 50, blue below. Select a state for its numbers, then reveal the winners.

In 2004 the nominee priced above 50 on the eve won all 51. In 2008 he won 48 of 50, leaving out Nebraska. The two exceptions were Indiana, where John McCain's contract closed at 62.2 and Barack Obama won, and Missouri, where it closed at 48.0 and McCain won.

Many of those closes rest on very few trades. On the eve in 2008, nobody traded the Republican contract in 12 states, and in four nobody traded it in the whole final week. In 2004 the figure was four states on the eve. A close on a day without trades is the figure the exchange recorded for that day; the record does not say how it was set.

Twenty years on

On 3 November 2026, Democrats won the House and the Senate in one election, under a Republican president. The last time both chambers changed hands that way was 2006, and Intrade's Senate market from that year is in this archive.

On the eve of the 2006 vote, Intrade's contract on the Republicans keeping the Senate closed at 73.5. Democrats took the chamber, with two independents who sat with them. On the eve of the 2026 vote, Polymarket's market on Democratic control of the Senate stood at 75.5 cents, and Democrats took it: about the same confidence, on the other side, and this time the favourite won. The House was the easier call both times. Intrade's contract on the Republicans keeping the House closed at 21.6 in 2006, and Polymarket had a Democratic House at 94.5 cents in 2026.

The year before looked different. In the twelve months before the 2006 vote, the Republican Senate contract never closed below 63.8, its low on 17 October 2006, so the market never gave a Democratic Senate much more than a one-in-three chance. A year before the 2026 vote, Polymarket had a Democratic Senate under 30 cents, and it stood above 50 at every daily reading from 27 August 2026.

The nights differed most. In 2006 the Republican Senate contract first traded at 90 at 11:08 pm Eastern on election night. It fell below 10 at 1:25 am, was back at 30 at 6:58 am, and traded below 3 only at 8:50 pm the next evening, when the Virginia race that decided the chamber was called for Jim Webb. In 2026 Polymarket's hourly price on a Democratic Senate was 75.5 cents at 6 pm Eastern, 90.5 at 7 pm and 98.95 at 9 pm.

One pair of elections proves little. A favourite at 73.5 should lose about one time in four, and 2006 may simply have been that time. The 2026 prices are Polymarket's, not the archive's, and are marked as added.

Read with care

• A price is not quite a probability. Intrade charged fees, traders cared about risk as well as odds, and a thin market can carry the view of a few large accounts. This edition shows prices as recorded and calls them prices.
• Most days had no trades. 1,792,290 of the 2,035,482 daily rows, 88 percent, report a volume of zero. On 1,786,357 of them the source gives no open, high or low, only a close.
• The record has no outcomes. The spreadsheet does not say how contracts settled, and the archive leaves those fields empty. Every result in this edition was added from official records and is marked as added.
• The rules are missing. Each market was defined by Intrade's contract rules, which are not part of the record. The 2010 Senate market shows why that matters.
• Some files did not survive. Twelve trade files are missing, Barack Obama's and Mitt Romney's own 2012 contracts among them; for those twelve, daily closing prices survive in a supplemental file. 6,462 further contracts report no trades at all.
• Some rows repeat and some fall outside the scale. 180,382 trade rows share their time, price and quantity with another row, and 1,283 rows hold a price outside 0 to 100. The archive keeps every one and flags them.
• The original files are named NoFinancial. The record does not say which contracts that name left out.

Use, cite and correct

Cite the data as: Panos Ipeirotis, Intrade Historical Archive, dataset, Zenodo, 28 October 2026, https://doi.org/10.5281/zenodo.23740923. His README asks users to cite two papers as well: Nikolay Archak and Panagiotis G. Ipeirotis, Modeling Volatility in Prediction Markets, Proceedings of the 10th ACM Conference on Electronic Commerce, 2009, pages 275 to 284; and Modeling Dependency in Prediction Markets, NYU Center for Digital Economy Research working paper CeDER-10-05, December 2010.

Cite this edition as: Vultax Research, What the crowd priced: 17,295 Intrade contracts, 2002 to 2013, Tallinn, 13 November 2026, https://vultax.com/research/intrade-historical-archive. The record page at https://vultax.com/archive/intrade lists the archive's files with their checksums.

To search all 17,295 contracts and chart any of them, use Ipeirotis's own explorer. Corrections to this edition go to Vultax Research through vultax.com/contact and are dated in its change log. Corrections to the archive itself go to Ipeirotis, as the contributing guide of his repository explains.

Figure 2 data · CSV · Figure 3 data · CSV · Figure 1 data · CSV · Citation file · BibTeX · Checksums

Thank you

Panos Ipeirotis saved Intrade's data in April 2013, weeks after the exchange stopped trading, and has kept it public since. This autumn he gave it checksums, research tables, a rights statement and a DOI. When we asked about an edition, he gave us written permission to "use the data and repository as you wish". This edition, and the copies of his files on our record page, rest on that permission.

Vultax's part is this edition: its three figures, the text, the comparison with 2026, the election results added from official records, and the files behind the figures.