Board Game Arena

Hello fellow board gamers on BGA!

I've been exploring the ELO distributions across games on BGA and have put together a statistical snapshot that I thought could be of interest to the community. It's a visual representation of where our ELO scores land on the wider scale.

Featured below are the ELO distributions for a selection of games, highlighting specific percentiles to delineate the spread of player rankings.

Update: I've added a few games, the up to date list of graphs can be found here:
https://github.com/RemiFabre/board_game ... elo_parser

Update: As requested I created a comparison between plots, in the case Azul vs Yatzy:



Here are the ELO distributions for Azul, Race For The Galaxy, Quoridor, and Yatzy:








Ever pondered what it really means to be "Strong" (300 ELO) or "Expert" (500 ELO)? For instance, in Azul, being "Strong" puts you in approximately the top 10%, while "Expert" status catapults you to the top 1%, and "Master" is an exclusive echelon reserved for the top 0.06%!

I was also impressed by the large numbers: there are currently 90 000 active Azul players!

If there's interest in seeing this analysis for other games, I'm open to requests and will try to accommodate them.

This post is meant to enrich our community discussions with a data-driven perspective. If there are any concerns about this content or its suitability here, please let me know—my intention is to share, not to disrupt.

Feel free to share your insights or any curiosities you might have about this.

Best,

WOW! totally awesome!!

Is that a lot of work? Could you make the same for 7 Wonders Duel and Catan ?? O.o

Interesting snapshot. Is it possible to use the same horizontal scaling on all of them? It makes it easier to visually inspect them to get an overall feel for the shapes of the distributions. At first glance something may look really different, then you realize 100 ELO on one graph is lining up with 150 ELO on another.

silentProtest wrote: ↑13 January 2024, 22:33 WOW! totally awesome!!

Is that a lot of work? Could you make the same for 7 Wonders Duel and Catan ?? O.o

I'll add them to the list tomorrow hopefully

Meeplelowda wrote: ↑13 January 2024, 23:25 Interesting snapshot. Is it possible to use the same horizontal scaling on all of them? It makes it easier to visually inspect them to get an overall feel for the shapes of the distributions. At first glance something may look really different, then you realize 100 ELO on one graph is lining up with 150 ELO on another.

That's an interesting input. I could make a single graph with several games inside to compare them

The sleeper datapoint here to me is how Azul's top player has almost double the ELO of the threshold to put you in the top 0.1%.

I would only include ELO >= 100 in the statistics and graphs.

Jellby wrote: ↑14 January 2024, 09:27 I would only include ELO >= 100 in the statistics and graphs.

Don't do that as then we would lose all the information about the pre-100 peak the charts are currently showing.

Well done, this is genuinely interesting.

Anything pre-100 is only about players who haven't played enough to have an even remotely reliable elo estimation. Sure, it is interesting to know how many players are below 100, but I don't think the distribution there is very interesting. The peak around 30 in Quoridor, for instance, is probably just players who have won a single game and that's it.

Also keep in mind that the "all time" ranking is only for players who have played at least one game in the last few months (2 months, I believe), so if take the data from there it wont be as exhaustive as it may look.

I also wondered about the 100 limit. So I plotted both versions: starting at 0 and starting at 100. You can browse the images here:
https://github.com/RemiFabre/board_game ... elo_parser