Point-in-time data: how often public gas and power data changes after publication
Published ; figures from Voltstack capture to
Point-in-time data keeps each published value with the moment it was seen, so you can ask what a series said at 07:00 on a past morning, not only what it says now. It matters because public gas and power data is corrected after publication, and often: in Voltstack capture from 23 Sep to 28 Oct 2026, 31 of 39 gas days of EU LNG send-out and 14 of 36 gas days of EU gas storage changed after first publication, as did 66% and 67% of quarter-hours of Spain's two imbalance prices and 49% of Austria's surplus price. Every count is a minimum, because each series was re-read for only a limited time after it appeared.
For a backtest, that is the difference between testing on the numbers you would have had and testing on numbers that arrived later. For a desk, it is the difference between showing what the morning view was built on and reconstructing it from memory.
Keep the value you saw, with the time you saw it
Voltstack records these series as GIE and ENTSO-E publish them and keeps each value it reads with the time it read it, from the start dates listed further down this page. It cannot fill in anything from before those dates.
Start a 14-day free trialRequest early access →The button opens a 14-day free trial sign-up. Sign-up needs a work email.
The button opens the early-access request on our home page.
Corrections per series, 23 Sep 2026 to 28 Oct 2026
Voltstack fetches each of these series on a schedule and stores a new row only when a value differs from the one it already holds, with the time of the read. A value that changed after we first saw it counts as corrected. Changes below a small rounding threshold are left out. The window ran to 28 Oct 2026, 14:02 UTC, and the counts were published on 29 Oct 2026 as a working paper (doi:10.5281/zenodo.23749147) with the change records behind every figure (doi:10.5281/zenodo.23749094). Both are ours: the author runs Voltstack, and the capture is Voltstack's.
| Series | Interval | First seen | Corrected | Share, a minimum | More than once | Median hours to last correction | Why it is a minimum |
|---|---|---|---|---|---|---|---|
| EU LNG send-out GIE ALSI · eu_lng_sendout | gas days | 39 | 31 | 79% | 17 | 40 | The five newest gas days were re-read; later corrections are not counted. |
| EU gas storage volume GIE AGSI+ · eu_gas_storage_volume | gas days | 36 | 14 | 39% | 0 | not shown | Each gas day was re-read only until the next one was published; later corrections are not counted. All 14 corrections were upward. |
| Spain imbalance price, surplus (A04) ENTSO-E Transparency Platform · es_imbalance_price_surplus | quarter-hours | 3,674 | 2,423 | 66% | 515 | 35 | Re-read for about three days after delivery; later corrections are not counted. |
| Spain imbalance price, deficit (A05) ENTSO-E Transparency Platform · es_imbalance_price_deficit | quarter-hours | 3,677 | 2,456 | 67% | 537 | 34.5 | Re-read for about three days after delivery; later corrections are not counted. |
| Austria imbalance price, surplus (A04) ENTSO-E Transparency Platform · at_imbalance_price_surplus | quarter-hours | 3,274 | 1,606 | 49% | 1,248 | 38 | Re-read for about three days after delivery; later corrections are not counted. |
- Share is corrected divided by first seen, rounded to a whole per cent. Each figure covers 23 Sep 2026 to 28 Oct 2026 only.
- First seen means Voltstack's first capture. GIE was read three times a day and ENTSO-E every 30 minutes, so a value may have been public for a while before we saw it, and a change inside that time is not counted.
- No median time is shown for EU gas storage. All 14 corrections were caught at the last read our capture made before the next gas day appeared, so the timing describes our schedule, not GIE's.
- Corrections changed the sign of the price in 142 Spanish surplus, 140 Spanish deficit and 27 Austrian quarter-hours. The Spanish and Austrian series include some quarter-hours from 20 to 22 Sep, because the first imbalance reads on 23 Sep reached three days back.
Sources: GIE AGSI+ (Gas Infrastructure Europe, https://agsi.gie.eu/) and GIE ALSI (Gas Infrastructure Europe, https://alsi.gie.eu/); ENTSO-E Transparency Platform (https://transparency.entsoe.eu/), CC BY 4.0, aggregated and changed by Voltstack.
Why every count is a minimum
- Imbalance corrections made more than about three days after delivery are not captured. Imbalance prices were re-read for about three days, and that cut-off binds: for delivery on 21 Sep, 96 of the 102 Spanish quarter-hours we had recorded as corrected had changed again by 28 Oct, judged against what ENTSO-E publishes now. Publishers settle later than three days, and ENTSO-E's imbalance price documents carry a status of intermediate or final.
- The gas series were re-read for a short time too. Each AGSI+ gas day was re-read only until the next gas day was published, and ALSI only for its five newest gas days. Corrections made later are not in the counts.
- First capture is not first publication. See the note under the table.
- Five weeks, one season. The window covers late September and October, the end of the gas injection season. It does not tell you what happens every winter, or always.
What point-in-time (as-of) data is
Most data stores keep one value per gas day or quarter-hour and overwrite it when the publisher changes it. A point-in-time store keeps each value it has seen for that gas day, stamped with when it was seen. Every row then carries two dates: the day the value describes, and the moment it became known. Data engineers call this bitemporal; economists call each published version a vintage; analysts usually just say as-of.
Here is one gas day of EU LNG send-out, GIE ALSI, from the published change records:
| Gas day | Seen by Voltstack | Value (GWh/d) |
|---|---|---|
| 23 Oct 2026 | 3,601.8 | |
| 23 Oct 2026 | 3,775.2 | |
| 23 Oct 2026 | 3,777.8 |
A latest-value table holds 3,777.8 GWh/d and nothing else. A point-in-time table holds all three rows, so “what did ALSI say for 23 Oct as of 08:00 UTC on 26 Oct?” has an answer: 3,601.8 GWh/d, the only value published by then. The as-of query is simple: for each gas day, take the newest row seen at or before the moment you ask about.
First release or latest value?
Both are correct; they answer different questions. Use the value as it stood at a given moment to test a rule or a model you would have run at that moment. Use the latest value to study what physically happened. The mistake is to mix them up without knowing it, which is what a latest-value table does by default.
Why it matters for a backtest
A backtest asks how a rule would have done if you had run it in the past. If it is fed corrected history, it uses numbers that were not public when the trade would have been placed. That is look-ahead bias, and it flatters the result. Macroeconomics met the same problem decades ago and answered it with real-time datasets that keep each published version of a series (Croushore and Stark, 2001).
The fix is to feed the backtest the as-of value for each decision time. Two limits come with it:
- It works only from the day someone started keeping values. Voltstack's record of the series in the table starts on 23 Sep 2026; we have no record of what they first showed before that, and nothing on this page says what was public on earlier dates.
- It is only as good as the re-read window. A store that stops re-reading after three days holds the first round of corrections, not the settled value.
Why it matters for checking a desk's morning numbers
A gas desk starts the day from a few public series: storage, LNG send-out, flows. Later someone asks what the morning note was based on. If the desk keeps only the latest value, it cannot show what the number said when it acted. If it keeps only the first value, it carries figures the publisher has since changed. Neither copy is wrong to keep. What is missing in most setups is the pair, with the time each was seen.
Storage gives an example. One gas day of EU gas storage volume, GIE AGSI+:
| Gas day | Seen by Voltstack | Value (TWh) |
|---|---|---|
| 14 Oct 2026 | 837.84 | |
| 14 Oct 2026 | 839.92 |
The correction was 2.09 TWh. Every storage correction in the window went the same way: all 14 were upward, so on those days a view of European storage built on the first figure was running low. Most were far smaller than this one.
When Voltstack's as-of record starts, by series
Voltstack captures forward only and cannot back-fill, so each series has its own start date:
- Forecast feeds (NESO and RTE): from 31 Jul 2026.
- JAO: from 5 Aug 2026.
- Open fundamentals (ENTSO-E imbalance prices and cross-border flows, GIE AGSI+ and ALSI): from 23 Sep 2026. Gassco follows from 24 Sep and NVE from 21 Oct 2026.
Other series are served as currently published. Imbalance corrections made more than about three days after delivery are not captured.
How to keep your own point-in-time record
- Store each read with the time you made it, and never update a row in place. Key it on the series, the interval it describes and the time you saw it.
- Write a row only when the value differs from the last one you hold. That keeps the table small; it is how the change records above were built.
- Keep re-reading after first publication. GIE publishes AGSI+ and ALSI daily at 19:30 CET, with a second run at 23:00 CET for late reporters, and each record carries a status and a last-updated time. For imbalance prices, re-read for weeks rather than days.
- Store the publisher's own status and timestamp beside yours, and set a threshold so rounding noise is not counted as a correction.
Frequently asked questions
What is point-in-time data?
Why does point-in-time data matter for backtesting?
How often is European gas storage data corrected?
How often is EU LNG send-out data revised?
Are imbalance prices on the ENTSO-E Transparency Platform revised?
Work from the record, not from memory
If a desk keeps only the latest value, it cannot later show what a number said when it acted. Voltstack keeps each value it reads with the time it read it, for the series and from the dates listed above.
The button opens a 14-day free trial sign-up. Sign-up needs a work email.
The button opens the early-access request on our home page.
Would your desk rather get the overnight changes by email? You can request the Revision Record morning note. It is a pilot, and we add each desk to the note by hand: write to research@voltstack.energy with the public series your desk works from.
More on these series
Disclaimer: This reference is produced by Voltstack Intelligence for information only and is not investment advice. The counts describe Voltstack's own capture from 23 Sep 2026 to 28 Oct 2026; they say nothing about why publishers correct their figures or what a correction cost anyone. Corrections to research@voltstack.energy.
Sources: GIE AGSI+ (Gas Infrastructure Europe, https://agsi.gie.eu/); GIE ALSI (Gas Infrastructure Europe, https://alsi.gie.eu/); ENTSO-E Transparency Platform (https://transparency.entsoe.eu/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/), aggregated and changed by Voltstack. Corrections measured from Voltstack's capture, 23 Sep 2026 to 28 Oct 2026: working paper doi:10.5281/zenodo.23749147 and dataset doi:10.5281/zenodo.23749094, both CC BY 4.0.