Historical Market Prices: How to Read Wine Price Data

A historical wine price looks like one number. In practice it is a stack of choices: which sales count, how a 12-bottle lot becomes a bottle price, which venues are in the sample, and which year the sale belongs to. Change any of those and the history changes.

This guide uses the Wine Labs auction dataset as it stood on November 4, 2026 to show what wine price history really looks like. It covers 6,498,211 priced wine auction lots for 131,456 wines, sold between 2000 and November 2026, after the cleaning steps described below. For what each auction price field means (estimate, hammer, realised, buyer's premium), read our separate guide to auction wine prices. This article is about the time series.

How deep historical market prices go

Wine auction history is not evenly spread over time. Most of the evidence is recent.

20001 house2,559

20033 houses45,977

20067 houses92,137

20096 houses83,727

201211 houses109,753

201524 houses166,350

201840 houses276,114

202096 houses386,122

2022117 houses513,317

2024147 houses637,860

2025148 houses797,394

2026, to Nov 4161 houses824,503

Selected sale years, as in the table below. 2026 runs to November 4.

Sale year	Priced wine lots	Auction houses reporting
2000	2,559	1 house
2003	45,977	3 houses
2006	92,137	7 houses
2009	83,727	6 houses
2012	109,753	11 houses
2015	166,350	24 houses
2018	276,114	40 houses
2020	386,122	96 houses
2022	513,317	117 houses
2024	637,860	147 houses
2025	797,394	148 houses
2026, to Nov 4	824,503	161 houses

Two things follow from this table.

First, 2026 already has more priced wine lots than any full prior year, and more than 300 times the 2000 count. A chart that starts in 2000 starts on a thin base. Early points describe a few houses, not the market.

Second, the "median price per 750 ml" column is not a market index. It rose from $32 to $94 because the mix changed: from one house in 2000 to 161 in 2026, and from 450 wines to more than 56,000. A median across all lots measures what was sold, not how prices moved. The same effect shows up in published market reports. WineNews, reporting the 2025 iDealwine barometer, states that volume rose 18.5% and value rose only 9%, so the average price fell 8.2% to €137. Wines over 20 years old fell from 69% of sales in 2024 to 55% in 2025. The average fell partly because the mix got younger.

The fix is to hold the wine constant: one wine, one vintage, one format, followed over time.

One wine, one vintage, twenty years

The table below follows three wines with deep auction records. Each cell is the median realised price per 750 ml for that vintage in that sale year, with the number of lots in brackets. Years with fewer than five lots are left out.

The three histories tell three different stories.

Lafite 1982 is a boom and a long unwind. The median rose 4.3 times from 2006 to 2011, from $979 to $4,200. By 2025 it was $1,664, 60% below the 2011 peak. Anyone who reads "Lafite 1982 has gained 88% since 2006" and "Lafite 1982 has lost 56% since 2011" is reading the same series. The start date decides the headline.

Realised price per 750 ml. 2026 runs to November 4.

Sale year	Median per 750 ml	Lots
2006	$979	84 lots
2007	$2,200	88 lots
2008	$2,604	130 lots
2009	$2,602	124 lots
2010	$3,936	135 lots
2011	$4,200	229 lots
2012	$2,988	126 lots
2013	$2,924	108 lots
2014	$2,654	145 lots
2015	$2,227	133 lots
2016	$2,390	84 lots
2017	$2,629	124 lots
2018	$2,788	105 lots
2019	$2,390	142 lots
2020	$2,390	138 lots
2021	$2,739	185 lots
2022	$2,576	162 lots
2023	$2,188	175 lots
2024	$1,776	182 lots
2025	$1,664	197 lots
2026	$1,845	195 lots

Grange 1990 is close to flat. It moved from $398 in 2006 to a peak of $624 in 2011, then traded between $460 and $616 for fifteen years. Over twenty years it gained about 31% in nominal dollars, before storage, insurance and selling costs.

Dom Pérignon 1996 climbed slowly. It more than doubled, from $177 in 2006 to $392 in 2025 and 2026, with a high of $415 in 2022. Most of the gain came after 2020.

None of these is a forecast. They show why a valuation must name the wine, the vintage and the window. "Fine wine went up" is not a usable statement.

Single sales are noisy

A price history built from one or two sales a year will jump around for reasons that have nothing to do with the market. Our data shows how often that is the case.

We grouped every priced wine lot by wine, vintage and sale year. There are 1,664,731 such groups.

For most wines, a "price history" is a handful of points. Only about 4,500 wine-vintage pairs have a record that spans twenty sale years.

Even where the data is deep, prices for the same wine, vintage and year spread widely. Across the 125,004 groups with ten or more sales:

• The middle half of sales (25th to 75th percentile) spans 19.2% of the median.
• The 90th percentile price is 1.47 times the 10th percentile.
• The highest sale is 1.98 times the lowest.

So a single result can easily sit 25% away from the typical price without anything unusual about the market. Condition, fill, provenance, lot size, venue and timing all play a part.

The benchmark wines also show why every extreme record needs a second look. In 2011, the 229 priced lots of Lafite 1982 had a median of $4,200. One lot in the same year is recorded as a single 750 ml bottle at $60,000, fourteen times the median. In 2022, a general antiques sale listed "Dom Pérignon 1996, sealed box" and the record reads as 12 bottles at $13 each, against a median that year of $415. Records like these usually come from a lot description that does not state the true format or bottle count. Check the lot before you use it. A single extreme record should never set a value.

Venue matters too. A study of fine-wine auction prices from eight houses over 2000 to 2012 in the Journal of Wine Economics found significant premiums for Hong Kong and for particular auction companies, after controlling for the wine. The number of houses in our Lafite 1982 sample grew from 4 in 2006 to 26 in 2026. Part of any change in its median is a change in where it was sold.

Dispersion also differs by wine. In 2025, the middle half of Lafite 1982 sales spanned 36% of the median. For Dom Pérignon 1996 it was 22%. For Grange 1990 it was 6%. Recent Grange 1990 lots are very uniform: only 1 of 330 in the last two years was not a 750 ml bottle. Lafite 1982 sells in many formats and lot sizes, across more houses.

What normalisation removes, and what it changes

Before any of the numbers above can be compared, raw auction records go through a set of rules. Each one changes the sample. This is the funnel for the November 4 snapshot.

The largest cut is the price step. 8.2 million completed lots have no positive realised price: they were unsold, withdrawn, or the house did not publish a result. A history that counts every lot offered, not every lot sold, overstates the evidence by a wide margin.

Two further rules do not remove lots. They change the price.

• Lot size. 3,246,156 of the kept wine lots hold more than one bottle. Their lot price is divided by the bottle count.
• Format. 484,324 kept wine lots are not 750 ml bottles: halves, magnums, jeroboams and larger. Their price is restated per 750 ml equivalent, so a magnum counts as two bottles.

Here is what those rules do to the three benchmark wines over the last two years (November 5, 2024 to November 4, 2026):

If you skip the per-bottle step, Lafite 1982 looks 31% more expensive than it is. If you compare a hammer price with a realised price, you mix in the buyer's premium: for these three wines, realised per bottle runs 18% to 20% above hammer per bottle. Under Christie's New York wine sale conditions, for example, the buyer pays a 25% premium on the hammer price before taxes. The same rule applies to headline sale totals. Wine Spectator's report on Sotheby's September 2017 New York sale gives $2.7 million including a 22.5% premium, which is about $2.2 million at hammer.

Retail versus auction today

Auction history is long. Retail history in our data is short. Wine Labs has stored daily retail price aggregates since May 25, 2025, so the retail series covers about 17 months. Before that date, we do not show retail history.

Today, the same three wines cost more at retail than their recent auction median. Retail here is the lowest 750 ml bottle price at each online merchant, then the median across merchants. Auction is the two-year median realised price per 750 ml.

$0$1,000$2,000$3,000

Lafite Rothschild 1982$1,750$2,995

Penfolds Grange 1990$485$677

Dom Pérignon 1996$387$491

Per 750 ml. Retail: online merchants, November 4, 2026. Auction: last two years, buyer's premium included.

Wine	Retail median	Auction median
Lafite Rothschild 1982	$2,995	$1,750
Penfolds Grange 1990	$677	$485
Dom Pérignon 1996	$491	$387

The gap is not an arbitrage. Auction prices include the buyer's premium but exclude tax, shipping and storage. Retail prices often include sales tax or VAT, depend on the merchant's market, and are asking prices, not trades. The gap does tell you which side of the market you are pricing from. An insurance replacement value and a sale value can differ by a quarter or more for the same bottle.

Retail history has its own mix problem. Look at Lafite 1982 by quarter:

Retail jumped 23% from Q3 to Q4 2025. In the same quarter, the number of tracked offers almost quadrupled, from 30 to 113, and the auction median moved 4%. Most of that retail jump is new merchants entering the sample, not a price rise. Read retail history together with its offer count.

How to read a historical price chart

Use this checklist before you trust a line on a chart.

1. Is it one wine and one vintage? If it is a basket or a category average, ask how the mix changed over the period.
2. Is the price per 750 ml? Multi-bottle lots and large formats must be restated before they are comparable.
3. Hammer or realised? Do not mix them. Add fees only once.
4. How many sales sit behind each point? In our data, half of all wine-vintage-year groups have a single sale. One sale is a data point, not a price.
5. Which venues? A series that adds new houses over time will move with the venue mix.
6. Which start date? Lafite 1982 is up 88% or down 56%, depending on whether you start in 2006 or 2011.
7. Which side of the market? Retail asking prices sit above auction results. Name the side you need: buy, sell, or insure.

Researchers handle the mix problem with constant-quality or hedonic models that separate the value of producer, vintage and other attributes. Research in the Journal of Wine Economics using a worldwide auction database from 2003 to 2014 found that most of those attribute values were not stable over time. That is a further reason to value a bottle from recent, close comparables, not from a long-run average.

The bottom line

Historical market prices for wine are useful when the series is narrow and transparent: one wine, one vintage, per 750 ml, sold lots only, with the sale count shown. Broad averages mostly measure who was selling what. Long histories exist for a few thousand wine-vintage pairs. For the rest, a valuation is a range built from a small number of recent comparables, and it should say so.

See per-wine histories in Wine Labs auction data, value a full cellar with Cellar Pricer, or explore fine wine market data.

Data methodology: Wine Labs auction dataset (current_auction_cleaned) and retail datasets, snapshot taken November 4, 2026. Sale date is the auction date, or the sold date when no auction date exists. Auction figures use lots with a sale date on or before November 4, 2026, a positive realised price, a canonical product match, product type wine, and no outlier flag; no lots in this snapshot carried a mixed-lot flag, so that filter removed none. Prices are realised prices in USD including buyer's premium, restated per 750 ml equivalent. Hammer per bottle is hammer price divided by bottle count. Year cells with fewer than five lots are omitted. Dispersion and depth statistics exclude non-vintage wines. Retail today is the lowest 750 ml bottle price per merchant in current merchant offers, then the median across merchants. Retail history is the quarterly median of daily world off-premise aggregates, available from May 25, 2025. Counts change as houses update records and matching improves.