Ramon Oliveira Vidal
Germany
Göttingen
flag msg tools
mbmbmbmbmb
By extracting information from BoardGameGeek, using statistics and machine learning approaches, I am trying to find “hidden” information in the bgg database.
I started by asking the question: “What is so special at the top ranked games on BGG?”
WARNING: These results are not conclusive, I know there are some problems with the method, but I will try to fix in the future.


As top ranked games, I selected all the top 200 games (in 21 January 2016). As low ranked games, I selected games positioned between the 1000 and 1400. I know this selection was arbitrary, but I couldn’t select much more games, otherwise my program would never stop computing. I also want to select games that are not too bad evaluated because we all know the reasons. On the future I will play more with this.

As game features I used:
Mechanic
Subdomain
Year published
Category
Minimum and maximum play time
Family
Minimum and maximum number of players
Artists
Designers

The classifier had a true positive ratio of more than 80% for the low ranked games and only 53% for the top ranked. This means that it is easier to predict the lower ranked games than the top ranked ones.
Moreover, none of these features are highly predictive per se. The combination of features is the key to understand what makes a game high or low ranked on BGG.

There are very few differences between the top ranked games and low ones. Some of them are pretty obvious if you are an experienced gamer.

1. Strategy. The most important factor to be high ranked on bgg. Games that require decision tree analysis are on the top. This is also a good and easy example of the feature combination. This rule cannot be applied for strategic games published before 2009.
2. MaxPlayTime. Gameplay lower than 70 min is not a good sign. When I check the results I was surprised by a game ranked as 126 on BGG with maxplaytime of 12000 hours.
3. Family games. They are not at the top and this is more or less common sense. Family games are also less strategic and faster to play, what explains the previous conclusions.
4. Tableau Building.
5. Card Drafting, Worker Placement, Variable Player Powers and Hand Management (except if it is a crowdfunding game) are all less represented in low ranked games.
6. Artist. The art in the game is very important and there are two artists that make your game going to the top: Henning Ludvigsen and Dennis Lohausen.
7. Colonial Theme (The European colonial period). This theme is present in 10 of the top 200 games and is not present in the lower ranked ones.
8. Corey Konieczka. This designer is responsible for many of the top 200 games on bgg.

More details about the methods I used you can check in my blog post: http://meeplezone.com/2016/01/21/what-makes-a-high-ranked-ga...
Let me know if you have suggestion and/or comments.
4 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Russ Williams
Poland
Wrocław
Dolny Śląsk
flag msg tools
designer
badge
Avatar
mbmbmbmbmb
ramonvidal wrote:
2. MaxPlayTime. Gameplay lower than 70 min is not a good sign. When I check the results I was surprised by a game ranked as 126 on BGG with maxplaytime of 12000 hours.

The game I see currently ranked as 126 is Dungeon Lords, with playtime 90 minutes. What game are you talking about?
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Rob Foley
United States
Maine
flag msg tools
Avatar
mbmbmbmbmb
russ wrote:
ramonvidal wrote:
2. MaxPlayTime. Gameplay lower than 70 min is not a good sign. When I check the results I was surprised by a game ranked as 126 on BGG with maxplaytime of 12000 hours.

The game I see currently ranked as 126 is Dungeon Lords, with playtime 90 minutes. What game are you talking about?


Probably a typo and meant 136 for Advanced Squad Leader which has 12000 minutes as the high end for playtime.
3 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Travis Dean
United States
Escondido
CA
flag msg tools
The Golden Camel UberBadge is a mystery. Nobody knows where it came from or how it got there.
badge
No one expects the Spanish penguin!
Avatar
mbmbmbmbmb
Reric wrote:
russ wrote:
ramonvidal wrote:
2. MaxPlayTime. Gameplay lower than 70 min is not a good sign. When I check the results I was surprised by a game ranked as 126 on BGG with maxplaytime of 12000 hours.

The game I see currently ranked as 126 is Dungeon Lords, with playtime 90 minutes. What game are you talking about?


Probably a typo and meant 136 for Advanced Squad Leader which has 12000 minutes as the high end for playtime.


You're probably right. Two typos then. Meant 136 instead of 126, and minutes instead of hours. That's a magnitude of difference there... Only 200 hours.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
maf man
United States
Portage
Wisconsin
flag msg tools
mbmbmbmbmb
you say "these results are not conclusive" but it sure as hell reads like conclusions.
These types of "findings" make me queasy. Too many connections drawn between things that aren't defined enough.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ramon Oliveira Vidal
Germany
Göttingen
flag msg tools
mbmbmbmbmb
These results come from machine learning approaches. I am just reporting them.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ramon Oliveira Vidal
Germany
Göttingen
flag msg tools
mbmbmbmbmb
Dolus wrote:
Reric wrote:
russ wrote:
ramonvidal wrote:
2. MaxPlayTime. Gameplay lower than 70 min is not a good sign. When I check the results I was surprised by a game ranked as 126 on BGG with maxplaytime of 12000 hours.

The game I see currently ranked as 126 is Dungeon Lords, with playtime 90 minutes. What game are you talking about?


Probably a typo and meant 136 for Advanced Squad Leader which has 12000 minutes as the high end for playtime.


You're probably right. Two typos then. Meant 136 instead of 126, and minutes instead of hours. That's a magnitude of difference there... Only 200 hours.


thanks for the correction
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ramon Oliveira Vidal
Germany
Göttingen
flag msg tools
mbmbmbmbmb
mafman6 wrote:
you say "these results are not conclusive" but it sure as hell reads like conclusions.
These types of "findings" make me queasy. Too many connections drawn between things that aren't defined enough.


These results come from a datamining algorithm. As I mentioned on the post, I fed the machine with the BGG database information and, as output, I got these as main features. I am showing the math behind the database. The results could vary if I use the games ranked between 2000-2400 instead of 1000-1400 (this is what I mean by not conclusive concerning my initial question).
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Dan Wojciechowski
United States
Aurora
Illinois
flag msg tools
Avatar
mbmbmbmbmb
Given the size of the BGG database and the rate the average rating declines, I am not sure that games between 1000 and 1400 justify the label "low ranked". May I suggest that your retry your analysis using games 1 to 400 as the "high ranked", and 5000 to 5400 as "low ranked"?
2 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ramon Oliveira Vidal
Germany
Göttingen
flag msg tools
mbmbmbmbmb
danwojciechowski wrote:
Given the size of the BGG database and the rate the average rating declines, I am not sure that games between 1000 and 1400 justify the label "low ranked". May I suggest that your retry your analysis using games 1 to 400 as the "high ranked", and 5000 to 5400 as "low ranked"?

Thanks for the feedback, I will try the same approach using these two groups this weekend (I will include here the findings).
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ramon Oliveira Vidal
Germany
Göttingen
flag msg tools
mbmbmbmbmb
danwojciechowski wrote:
Given the size of the BGG database and the rate the average rating declines, I am not sure that games between 1000 and 1400 justify the label "low ranked". May I suggest that your retry your analysis using games 1 to 400 as the "high ranked", and 5000 to 5400 as "low ranked"?


I have already some results from the comparison between the top 400 games and the low 5000-5400. It shows the Strategy still as the main difference between the top and down. The other factors are: "Press Your Luck", high "maxPlayers" and Set collection games as positive factors for low ranked games.
2 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Dann Albright
United States
Denver
Colorado
flag msg tools
Avatar
mb
Interesting stuff! Sure, there are issues with various things in the statistical analysis, but it's still an exercise worth undertaking. Of course these results aren't going to be conclusive, but it gives you something to think about. Even on a personal level–what is it that really makes a game appeal to me? The idea of analyzing games specifically on BGG is also very interesting because you're looking at a specific subset of gamers, and this can tell us as much about the type of people that hang out on BGG as it does the games themselves.

Thanks for sharing, and keep us updated on future analyses!
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ramon Oliveira Vidal
Germany
Göttingen
flag msg tools
mbmbmbmbmb
dralbright wrote:
Interesting stuff! Sure, there are issues with various things in the statistical analysis, but it's still an exercise worth undertaking. Of course these results aren't going to be conclusive, but it gives you something to think about. Even on a personal level–what is it that really makes a game appeal to me? The idea of analyzing games specifically on BGG is also very interesting because you're looking at a specific subset of gamers, and this can tell us as much about the type of people that hang out on BGG as it does the games themselves.

Thanks for sharing, and keep us updated on future analyses!


Thanks for like it Dann, I made another quick data analysis using combined machanics, I just published on my blog and I will soon publish also here on BGG: http://meeplezone.com/2016/02/02/dual-mechanics/
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Daniel Reed
msg tools
More different analyses and more in-depth consideration of different game qualities would make for a great scholarly paper! It would be informative and could potentially be an optional guide--not hard-and-fast rules, mind you--for people trying to develop a strong game.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Front Page | Welcome | Contact | Privacy Policy | Terms of Service | Advertise | Support BGG | Feeds RSS
Geekdo, BoardGameGeek, the Geekdo logo, and the BoardGameGeek logo are trademarks of BoardGameGeek, LLC.