The project goal is to make game recommendations for the customer. By comparing the customer’s collection of games to other collections I hope to find out where the customer ranks within the gaming community and if there are any clear preferences in game types or differences in preferences from the community. (All data pulled from www.boardgamegeek.com)
How does the collection size of the customer compare to other gamers? Over 158,000 gamers had at least one game in common with the customer. I selected 1000 of those gamers at random and loaded their collections.
## [1] "The customer has 160 games in their collection, and has rated 62."
## [1] "Of 1000 random gamers, they have 280248 games collected, and have rated 135621, for 1385920 in total."
These charts show that ratings from the customer and other gamers both follow a normal distribution, slightly skewed left.
Next, we learn a little more about what kinds of games the customer likes. Games have mechanics, like card-drafting, dice-rolling, and tile-laying. They also have categories like science-fiction, co-operative, and family.
First we look at which mechanics are most common in the customer’s collection.
This will be interesting later: Line drawing is to the bottom of the list, with 2 games.
For each mechanic, this chart shows the difference between the customer’s rating and the rating overall.
Since some mechanics are more common in games than others, we also look at the customer’s rating of games with each mechanic. I removed games rated under 6 from the chart. Some observations:
Mechanics at the bottom right of the table are in games the customer enjoyed but does not have very many of, and might be a good place to look for fresh recommendations.
Mechanics at the top right are in games the customer enjoyed and has purchased more often than other gamers, so they might be safe bets for games that will be liked.
Which game categories are common in the customer’s collection?
This chart shows categories of games that are more popular with the customer, than they are popular overall. This should be tempered by the number of games the customer owns of that category. Overall, gamers rate science fiction games higher than the customer does (shown on the chart as -0.5), but the customer owns a high proportion of these.
Here we show the number of games the customer owns with each category, and the customer’s rating of games with each category I removed games rated under 5.5. Some observations:
Next steps might be to identify groups of categories and mechanics based on rating and ownership and games.
This might be an opportunity to recommend games based on attributes as well as by ratings; a hybrid system.