Which board games keep or even improve their rating over time — and which were just hyped at launch? A rule-based classification of rating trajectories plus cohort medians per release year.
About the dataset
The used dataset is based on ratings that were submitted for games on BoardGameGeek before Oct. 16, 2026.
It includes all ratings given to games with at least 2,000 ratings.
This criterion was met by 305 games (excluding expansions).
A total of 1,814,921,092 ratings were collected for these games, resulting in an average rating of 6.82.
The reason for introducing the minimum rating number restriction is to reduce potential bias,
but also to increase statistical relevance and focus on well-established titles with a large number of users.
Ratings Over Time: Hype vs. Lasting Quality
Not every board game that starts strong stays strong. Using monthly rating time series anchored at each game's availability date, this analysis asks which games hold or improve their rating over the years — and which were mostly hyped in their first year and then declined.
For every game we compute rating anchors (the cumulative average rating 1, 3, 5, 10
and 20 years after availability, plus the current value rating_end).
All further groupings are derived in the browser from these persisted values —
the central derived figure is delta_end, the difference between the
current rating and the "first-year verdict" (rating_1y).
Games younger than 12 months have no first-year verdict yet, and games with fewer than 100 ratings are not shown. Note that only games from the top rankings are included — the analysis describes how successful games evolve, not how many games fail.
Each bubble is one game: x = rating change since the first year (delta_end), y = current cumulative rating (rating_end), bubble size = number of ratings. Color encodes the rating change (red = declined, green = improved), opacity encodes the current rating. Hover for details.
The same per-game data as in Figure 1, shown as the distribution of the rating change since the first year (delta_end). The violin outlines the kernel density, the box-whisker shows minimum, first quartile (Q1), median, third quartile (Q3) and maximum; the median, Q1 and Q3 values are labeled next to each box and the whisker ends mark the 2.5% and 97.5% percentiles (95% interval) with their values labeled at the tips; the dashed line marks delta_end = 0 (no change). The release-year filter is synchronized with Figure 1: when a year is selected there (or above), the view starts zoomed into that year's cohort range while the axis always spans the full all-years distribution — zoom out (mouse wheel or the slider on the right) to see the complete all-years violin.
Rating Trend × Attention: Four Longevity Archetypes
The rating-delta figures above look only at how games are rated — not at
whether they are still being rated at all. Combining both dimensions
yields four archetypes derived purely from the persisted metrics
(delta_end and activity_retention, split at 0 and 1.0):
- Evergreen — rating holds or improves and the game is still actively rated.
- Sleeping Classic — rating holds or improves, but hardly any new ratings arrive (activity retention < 1.0): respected, but no longer on the table.
- Contested — rating declined, yet the game is still heavily rated: a large community keeps re-evaluating it.
- Fading Hype — rating declined and rating activity collapsed: the classic "hyped once, then forgotten" pattern.
This view reveals what the pure rating-delta classes cannot: a "Stable" game whose rating never moved can still be practically dead, while a game whose rating slipped may still be one of the most actively rated titles on the platform.
Each bubble is one game: x = rating change since the first year (delta_end), y = activity retention (1.0 = rating activity holds since release; the y axis is capped at the 95th percentile so outliers do not compress the view), bubble size = number of ratings, color = archetype. Dashed lines mark the quadrant boundaries. The release-year filter is shared with Figures 1 and 2.
Modern vs. Old: Cohort Medians by Release Year
Comparing cohorts (games grouped by official release year) requires caution: older games were released when BoardGameGeek had far fewer active raters, so their early months are noisier. The count-weighted smoothing of the underlying monthly windows dampens this effect, but cross-cohort comparisons should still be interpreted with care.
Median cumulative rating (first year vs. today) and median smoothed rating over the last year (smooth_mean_recent — works for recent games too, where cumulative anchors are not yet reached) per release year, with the number of games per cohort. Games released in 2001 or earlier are combined into a single bucket, since BGG itself only went online in 2000.
Table 1 lists the key figures for every individual game: anchors, deltas and
activity measures (activity_retention, share of ratings received
early). Use the search box, the archetype filter and sortable columns to explore.
Searchable, filterable and sortable table of all games with their rating anchors and the client-side derived longevity archetype.
| Game | Year | Archetype | Rating 1y | Rating now | Delta | Share 1y | Retention % | Ratings |
|---|