Wine is rated out of 100.

The 100-point wine scale uses about eight points

Of 119,988 unique Wine Enthusiast reviews, none is published below 80 and 85.0% land between 84 and 92. Inside that band, each doubling of price buys about two points, vintage changes almost nothing, and the ten most prolific reviewers differ by 3.2 points on average.

Read the short version Code and data on GitHub

The lowest score in the dataset. The usable scale starts four fifths of the way up.

85.0%

Of wines score between 84 and 92, a band eight points wide.

r = 0.62

Between points and log price. Wines near $10 average 85.2 points; wines near $100 average 91.9.

3.2 pts

Between the most and least generous of the ten most prolific tasters (86.9 against 90.1).

At a glance

Question: How much of the 100-point scale do professional wine scores actually use?
Data: 119,988 unique Wine Enthusiast reviews (129,971 rows, 9,983 exact repeats dropped)
Method: Score distribution; Pearson r of points on log10 price ($4 to $500); linear fit of mean score on vintage, 1995 to 2016; means for the ten busiest tasters
Compared against: The scale's stated range of 0 to 100, and the price effect against the vintage trend and the spread between tasters
Result: Floor at 80; 85.0% score 84 to 92. Price: r = 0.62, 85.2 at $10 to 91.9 at $100. Vintage: 0.08 points per decade. Tasters: 3.2 points apart
Main limitation: One publication, one snapshot, no review dates: it shows a compressed scale, not scores creeping up over time

The question

What does 92 points mean?

A shelf tag that says 92 reads like a grade out of 100. That reading assumes the scale has a bottom half in use and that 92 sits well above the middle of it. Each review in the dataset has a score, a price, a title (which usually carries the vintage) and the taster's name, which is enough to check both assumptions.

Set the score and price from a bottle you have seen. The readings come from the full set of reviews; the charts further down move with them.

Read a shelf tag

against 119,988 reviews

Score on the tag

Price

$25

Reviewed wines scoring lower
Scoring the same or higher
Typical score at this price

The shape of the scale

A floor at 80 and a hump in the high 80s

8 pts

The width of the band that holds 85.0% of all wines.

The scores form a narrow hump that starts at a cliff. Nothing below 80 is published, which is the magazine's editorial choice and says nothing about the wines. The mean is 88.44 and the median is 88. A scale can be inflated in this structural sense without any year-on-year creep: the bottom four fifths simply never get printed.

Histogram of wine scores with a hard left edge at 80, a tall peak between 86 and 90, and a thin tail toward 100.

Reviews at each score, from tables/chart_data.json. The 84 to 92 band is shaded and the score on your tag is in olive. Hover a bar for its count.

What money buys

Each doubling of price buys about two points

Against the logarithm of price the correlation is r = 0.62, strong for data this noisy. On a log axis the curve over twenty equal-count price bins is close to a straight line: $10 to $20 adds 2.6 points, and $50 to about $115 adds 2.0. In dollars that is a steep diminishing return. A tenfold step from about $10 to about $100 moves the average from 85.2 to 91.9, and the first ten dollars of that step buy more points than the last fifty.

Average points by price bin on a logarithmic price axis, rising from about 85 at ten dollars to about 92 at one hundred dollars.

Mean points in each of 20 equal-count price bins, $4 to $500, plotted at the bin's median price on a log axis. The top bin runs from $85 to $500, so the chart says nothing about how scores move above $100. Hover for a bin.

Vintage and taster

Vintage barely moves the score. The taster moves it more.

Average score by vintage, 1995 to 2016, is close to flat: a fitted slope of 0.08 points per decade. Good and bad years exist for a given region, but they don't show in the mean of a catalog this broad.

Among the ten tasters with the most reviews, average scores run from 86.9 to 90.1. That 3.2-point spread is about half of what a tenfold price increase buys, and more than a third of the band where most wines sit, from a change of author alone.

By vintage

0.08 points per decade

Mean score per vintage year, with the fitted trend dashed. Early vintages rest on few reviews (50 for 1995); hover for counts.

By taster

86.9 to 90.1

Mean score of the ten tasters with the most reviews. Hover a dot for the taster's review count.

Limitations

One magazine, one snapshot

All of this comes from a single snapshot of one publication. The dataset has no review date, so it cannot show scores creeping upward over time, which is what most people mean by grade inflation. What it can show is the structural version: an 80 floor, a compressed band, and a stated range that is mostly unused.

Vintage is pulled from each wine's title with a regular expression. Titles without a year drop out of the vintage chart and stay in every other view. The taster comparison is a raw mean, so a taster who covers more expensive regions will look more generous for that reason alone. The price views leave out reviews with no price.

Correction, 2026-10-29: the source file repeats 9,983 reviews word for word, and earlier versions of this page counted each of them twice. With the repeats removed, the share scoring 84 to 92 went from 85.7% to 85.0%, r from 0.61 to 0.62, the $10 average from 85.3 to 85.2, the mean from 88.45 to 88.44, the vintage trend from 0.09 to 0.08 points per decade, and the top taster mean from 90.0 to 90.1 (spread 3.1 to 3.2). The 80 floor, the median of 88 and the $100 average of 91.9 did not change.

How it was built

Where the numbers come from

The scale, by the glass

from tables/summary.csv

Reviews scored: 119,988
Lowest score published: 80
Mean score: 88.44
Share scoring 84 to 92: 85.0%
Average at about $10: 85.2
Average at about $100: 91.9
Points against log price: r = 0.62
Drift per decade of vintage: 0.08
Taster means, low to high: 86.9 to 90.1

Data: The public winemag-data-130k-v2 Wine Enthusiast dataset, downloaded from a GitHub mirror by fetch_data.py
Sample: 119,988 unique reviews with score, price, title and taster; the source file has 129,971 rows and repeats 9,983 of them, which analyze.py drops
Price model: Pearson correlation of points against log10 price, $4 to $500; averages at $9 to $11 and $90 to $110
Vintage: Year parsed from the title; linear fit of mean score on vintage, 1995 to 2016
Tasters: Mean score for the ten tasters with the most reviews
Charts: chart_data.py writes the score counts, price bins, vintage and taster means to tables/chart_data.json with the same filters as analyze.py; the page draws them as plain SVG
Built with: Python, pandas, SciPy, matplotlib
Related: The same reviews are the retrieval index in Wine Sommelier RAG and the search tool inside the Wine Pairing Agent