Influence of ownership and comments on the average rating of all games

Ratings on games at BoardGameGeek add up to an overall rating, but are given under individual circumstances. The game can be rated by owners or non-owners. Also, a comment can be made in addition to the rating. This analysis explores the impact of these variables on the entirety of all games.

About the dataset

The dataset used consists of reviews submitted for games on BoardGameGeek. Included are ratings that were submitted before 2023/01/01. The dataset only includes games that have received at least 2000 ratings at the specified time. This restriction is intended to prevent games with few ratings from distorting the analysis.

Rating analysis

All games with at least 2000 ratings as of January 1, 2023 have collected a total of 15,542,331 ratings. On average, users gave these games a rating of 7.28. Statistics about all given ratings and rating subsets are displayed in chart 1.

Influence of the ownership on the rating

The first analysis is about the influence of ownership on the rating of a game. It is immediately noticeable that the average ratings of the groups of owners and non-owners differ significantly from the average rating of the entirety of all raters.

If a game is owned, the average rating increases to 7.68, which is a percentage increase of about 5.2% compared to the overall rating. The owner of a game thus rates it better than the average of all players.
This finding is not unexpected. There are several conscious and unconscious factors that bring about this influence. One conscious factor is that gamers are more likely to buy games they think they will like. If you already know you won't like a game, you won't make a purchase decision in the first place. Also, games are often bought after they have already been played with others and the game is therefore already known.
But unconscious factors also play a role. People tend to assign a higher value to things they own than they would to the same thing in someone else's hands. The often high costs that have gone into a purchase also have an effect. Thus, if you give a game a bad rating, you are admitting to having made a bad purchase. It's surprising that over 60% of all reviews are created by owners. Considering that board games are usually played together and only one person needs to own the game, this value seems high.

Unlike owning a game, not owning it has a negative impact on the rating. The average rating drops to 6.68, which is a percentage decrease of 8.2% compared to the general average. The difference is even more pronounced when compared to owners' ratings. Owners rate their games about 14.8% better than people who do not own the game. This corresponds to just under a full point on BGG's rating scale. This result is also not unexpected. Due to the lack of commitment through ownership, ratings can be given more objectively. Likewise, a game that could not make a positive impression while playing will not be purchased afterwards.

A special group in the entirety of all ratings of non-owners are those ratings that were given by people who have owned the game in the past. People who have removed games from their collection due to dislike, lack of space, or other reasons rate the games lower on average than the aggregate of all non-owners. This group has an average rating of 6.56, a decrease of about 10.0% compared to the general average. Owners rate their games about 17.1% better than people who previously owned the game.

Influence of a comment on the rating

Whether the rating is done with or without a commentary has a minor influence on the rating. In general, games without commentary are rated better. In just under 15% of all ratings, a comment is provided alongside the rating. Ratings are about 2.2% better when no comment is provided.

Distribution of the average ratings across individual games

From the overall view of all submitted ratings, some insights could already be gained. However, the results cannot be applied to every game. Instead, the ownership situation and the submission of a comment have an individual effect on games.

An overview of the specific average ratings of the individual games is provided by chart 2. There, all 1845 games per subset are shown in a violin plot. The wider the violin at a point, the more games have this average rating for the corresponding subset of games. In addition to the violin plot, a box-whisker plot is also included. Here, the box in the middle corresponds to the area where 50% of the data lies. Thus, the outer boundaries of the box represent the upper and lower quartiles. In the middle of the box, the median can be seen as a line, representing the average rating of the game, which lies exactly in the middle of the data set (rank 923). With the help of the whiskers, which are represented by the two small vertical lines, outliers are made clear. Everything that lies outside the whiskers is considered an outlier. The whiskers are at most 1.5 times as long as the interquartile range (median - quartile).

It can be seen from the graph that the average ratings of all analyzed games in all categories are relatively normally distributed. This suggests that the ratings of the individual games do not influence each other.
Depending on the subset, one can see a different density in the distributions. The less wide a distribution is, the more similar the scores are in the corresponding subset. The smallest width can be found in the subsets of ratings of non-owners and former owners. This can also be seen from the distances between the whiskers in these subsets, as they have the smallest distances between each other. In contrast, the subset of owners has the greatest width of the distribution of ratings. Thus, the whiskers of the Box-Whisker plot are the widest apart here.

In general, a relationship can be derived between the average rating of a subset and the width of the distribution. The better the average rating, the wider the width of the subset's normal distribution. From this follows that worse average ratings result in a narrower width of the normal distribution. This means that all ratings of a subset are closer together, the worse a subset is rated overall. The better a subset is rated, on the other hand, the wider the individual ratings are distributed.

Another consideration addresses the outliers, which are the values that lie outside the whiskers of the box-whisker plot. From the graph, it is difficult to identify how the number of outliers varies between categories, but it is possible to look at the distances of the maximum outliers from the ends of the whiskers.
When it comes to the positive outliers, one thing stands out: while the weakest positive outliers are seen in the ratings of non-owner games, the largest outliers seem to be in the games of former owners. This is particularly noteworthy in the sense that former owners are a subset of all non-owners. This observation suggests that there are some games that are rated particularly highly by former owners, even though they are no longer owned. In contrast, there are no games that are particularly singled out by all non-owners. When looking at the negative outliers, no distinctive conclusions can be drawn.

After this analysis of all ratings, it is worth taking a look at how the factors considered affect individual games. A more in-depth analysis of individual games can be found in this article: Influence of ownership and comments on the average rating of individual games.