The archive · a Vultax Research series

Old forecasts, kept before they disappear.

Each edition takes a forecast record that was close to being lost, checks it against its source and publishes it with the original numbers, full credit to the people who made it, and its limits stated. Every edition can be read here and downloaded as a single file.

What each record covers, and when it reached us

No. 1 · What the ocean forecast remembers

Target months from January 1972 into 2001

Covers target months from january 1972 into 2001; published by Vultax on 25 October 2026.

No. 2 · What the market expected

October 2002 to June 2005

Covers october 2002 to june 2005; published by Vultax on 25 October 2026.
Years the forecasts cover Kept by its makers or a data library Published here

The editions

Two editions, in order

No. 1

Climate forecasting

What the Ocean Forecast Remembers: The ENSOFORECAST Archive

Explore historical ENSOFORECAST maps with original credits, traceable fields, and clear distinctions between hindcasts and unknown issue times.

4

lead times: 3, 6, 9 and 12 months

2

model versions, LDEO1 and LDEO2

The record
ENSOFORECAST: monthly maps of the tropical Pacific from the Zebiak and Cane El Niño model.
Made by
Lamont-Doherty Earth Observatory, Columbia University, with the model Stephen Zebiak and Mark Cane described in 1987
Kept by
The IRI Data Library at Columbia, which began winding down in October 2026
How it came here
Aaron Kaplan of IRI assembled the full directory and sent a copy, with a checksum, on 21 October 2026.
Read with care
The files carry no issue dates, and the early years are hindcasts made after the fact. The edition marks which is which.
Read the edition

No. 2

Economic forecasting

What the Market Expected: 153 Historical Economic-Event Auctions

Explore 153 historical economic-release auctions: probability ranges, announced outcomes, source lineage and credits to Gürkaynak and Wolfers.

153

economic releases, across four indicators

2,235

priced ranges

The record
Auctions that priced every range of answers before a US economic release, run by Goldman Sachs and Deutsche Bank.
Made by
Traders in the auctions; the results were assembled by Refet Gürkaynak and Justin Wolfers
Kept by
The authors' own research package, published with their paper
How it came here
Justin Wolfers gave permission to republish the auction data with credit in October 2026.
Read with care
A price is a probability only if traders were indifferent to risk. The edition shows the published numbers without rescaling.
Read the edition

How an edition is made

01

From the people who kept it

Each record comes from the people who made it or held it, and is checked against the copy they sent.

02

The numbers as they were

Values appear as published. Nothing is rescaled, smoothed or filled in, and a missing value stays missing.

03

Dates stated plainly

Each edition says what a forecast was for and, where the record allows, when it was made. Where an issue date is unknown, or a forecast was made after the fact, the edition says so.

04

Credit and a copy

The people who made the record and the people who kept it are named on the edition, and each edition is one file anyone can download and keep.

Why old forecasts

A forecast can be judged only against what was said before the outcome was known. Once the original files are gone, nobody can make them again: a forecast written today for 1997 already knows how 1997 turned out.

Records like these often survive on one server or one disk. ENSOFORECAST sat unchanged in one data library from 2000 until the library began to close in October 2026.

Do you hold a forecast record that could be lost?

A forecasting service, a market or a research group that is closing, or a dataset that lives on one person's disk: we can preserve it, name you on it and give it a permanent page. We take no personal data.

[email protected]