Browse the complete wines index →
Live · 69,234 wines
Wines
Every wine with at least 3 z-qualifying Vivino reviewers, ranked on the mean per-reviewer z-score and dual-scored (z-normalized + raw average).
Top of the rankings
Highest z-normalized percentile
Where the methodology pays off
Biggest inflation gaps
Wines where z-normalized and raw-average percentiles diverge most.
Score distribution
z-normalized · all 69,234 wines
Source coverage at launch
One source at launch — Vivino. Every qualifying reviewer is normalized onto their own z-scale and then counted equally: no reviewer and no source counts for more than any other. Professional-critic integration (Wine Advocate, Wine Spectator, Jancis Robinson, Decanter, Vinous, James Suckling, Jeb Dunnuck, CellarTracker) is on the roadmap, and when a second source joins its reviewers enter the same normalization on the same equal footing.
| Vivino (live) | 3.24M reviewers, 14.3M reviews ingested. 1.17M qualified for z-normalization (DB1, stddev>0); 1.37M qualified for raw-average track (DB2, review_count≥2). Catalog filter: ≥3 z-normalizable reviewers per wine. |
|---|---|
| Wine Advocate (planned) | Oldest professional 100-pt scale |
| Wine Spectator (planned) | Institutional panel reviews |
| Jancis Robinson MW (planned) | 20-pt scale with less inflation pressure |
| Decanter, Vinous (planned) | International panel coverage |
| James Suckling, Jeb Dunnuck (planned) | Prolific professional critics |
| CellarTracker (planned) | Enthusiast crowd, deep on specific wines |
Global, cohort and adjusted percentiles
Every wine publishes three figures, and all three are built from the same reviewer-normalized z-score — each reviewer re-centered on their own mean and divided by their own standard deviation, then counted equally.
- Global percentile — headline number on the product card. The unweighted rank of that z-score against every wine we rank. A 90 means top 10% of all wines.
- In-cohort percentile — the same z-score, simply re-ranked inside a smaller peer set: wines of the same type from the same country. A re-ranking, not a second computation.
- AI-adjusted percentile — the same reviewer-normalized mean, corrected for how much confidence the sample actually supports: mean z × n / (n + 53), so a wine with a thin reviewer base is pulled toward the corpus average until its coverage earns the distance back.
Each product page publishes all three numbers alongside the reviewer counts behind them — so you can audit any percentile against the published formulas and constants on our methodology page.
What we'll cover at launch
Year-one target: 150-200 wines reviewed end-to-end with full normalization math visible. Cadence: 2-4 wines per week. Coverage priorities:
- Under-$15 value picks across common varietals
- Napa Cabernet 2019-2022 vintages
- Red and white Burgundy 2020-2022
- Champagne + sparkling value picks
- Italian reds (Barolo, Brunello, Chianti Classico)
- Spanish Rioja + Ribera del Duero
- Australian Shiraz + Cabernet
Expansion from there driven by search demand and seasonal relevance. See the full methodology for the math behind every score.
Frequently asked questions
Why wines first and not books or electronics?+
How often will wine scores update?+
Will you cover natural wines / obscure producers?+
What retailers do you link to?+
Status: wines is live with 69,234 dual-scored wines. Browse the catalog at /wines/catalog, or view the full methodology for how each score is computed.