Go Rankings Survey (Feb 2024)

It’s been a while since there’s been a survey of the different server and Go association ranks (as far as I’m aware), so I thought I’d start one.

I’ve included the following servers/associations in the survey:

  • OGS
  • KGS
  • Fox
  • Tygem
  • Pandanet
  • EGF
  • AGA
  • Other national association (if applicable)

The survey is anonymous. All fields are optional, so please skip any that aren’t applicable or where you think your rank is inaccurate.

Form link

I’m planning on running the survey for the next week and will post the results once it concludes.

I originally posted this on reddit, but thought I’d post here as well to try to get more entries.

8 Likes

Hi!

Thank you for the survey.

A little criticism:

  • BadukPop provides both a “rank” in kyu,dan, and an “automatch rating” which is a four-digit number. The rank is more an achievement thing that reflects the user’s activity on the app, including live games, correspondence games, life and death problems and yose problems. This “rank” is a bit bullshit and can be very, very far from the go player’s usual rank in both directions (ie I could easily be OGS 3dan and BadukPop 15kyu, or OGS 3dan and BadukPop 8dan, depending on how active I’ve been on BadukPop). The “automatch rating” is probably a much better descriptor of the user’s “go rank”, but it’s not in your survey.
  • GoQuest is a pretty popular go server which is not in your survey (GoQuest also has separate “rank” and “rating”, with rating probably being more accurate)
5 Likes

Hi, the survey is also missing Dragon Go Server (DGS) if it’s of interest to add. ^^

3 Likes

I suggest to also plot histogram like that: Unofficial OGS rank histogram (and graphs) 2022 - #26 by stone.defender (but by using result of survey only)
don’t only compare rank between servers. Information of which rank got the most users on each server may be useful for understanding size of collected data and for comparing servers.

3 Likes

Thanks for the feedback. I’ve added both rank and rating for BadukPop and GoQuest (and rank for DGS, @fuseki3), but most likely there won’t be many entries for those this round since they were added so late, so I’ll probably not include results for them when I do the analysis. I’ll copy this format for future surveys so they’ll be included then (looking to do it every 6-12 months).

4 Likes

That’s a good suggestion. I’ll be releasing the full data, so people can generate whatever they want from it.

3 Likes

Results are in, though I wasn’t able to do as deep of an analysis as I had wanted to.

You can view and download the raw data here: Go Rankings Survey Feb 2024 Raw Data

To compare two servers/associations you can do the following:

  • Click “Hide fields” then at the bottom click “Hide all”. Enable just the two fields you care about.
  • Click “Filter” → “Add condition” and filter to only include filled results for both fields (e.g. Where “OGS rank” is not empty and “Fox rank” is not empty)
  • Click “Sort by” and select the first field included in the comparison

Graphs are available here: Go Rankings Survey Feb 2024 Results

Some high-level stats:

  • 119 entries
  • OGS was the most popular server with 105 entries, next was Fox at 69 (totals per server can be seen at the bottom of the raw data view columns)

I’ve included histograms for each of the servers/associations (excluding those with low number of entries). I wanted to run a regression analysis to generate a comparison table similar to the survey ran in 2018, but I couldn’t get scipy working on my machine. If someone has the know-how to do this sort of analysis it would be much appreciated (I think the python notebook from the previous survey analysis should work with only minor changes).

If anyone has any other data visualizations they’d like to make and share I’d be happy to add them to the graphs link.

8 Likes

OGS to Fox:


OGS to KGS:


OGS to Tygem:


OGS to Pandanet:


OGS to AGA:


OGS to EGF:

12 Likes

That’s a really nice visualization. I’ve generated similar ones for all server/association pairs and uploaded them to the results table.

5 Likes

lol, of course almost no one is in America and in Europe simultaneously
But there still certainly enough information to do some more comparison.


+

=
EGF to AGA

OGS is in the middle


then if we remove central column and simplify ranges

EGF to AGA


EGF to AGA

Fox is in the middle


EGF to AGA
image
KGS is in the middle


EGF to AGA

Pandanet is in the middle


EGF to AGA

Tygem is in the middle


TOTAL:

EGF to AGA:


AGA to EGF:


tables like last two (total) possible to do for every pair of servers)

6 Likes

I conclude that in the high kyu and low dan range, AGA rank = EGF rank +2.

IIRC when OGS implemented the glicko2 rating system few years ago, there was some data on AGA ranks being roughly 1.5 ranks higher than EGF ranks at 1d level, and ogs sytem was designed to sit about half-way between the two. So based on this survey, it looks like we’re still pretty close to that goal ^___^

3 Likes

Not sure if this is already referred to (just too lazy to check).
It is maybe a bit outdated.

https://www.reddit.com/r/baduk/comments/1axzu6k/comment/krwyesy/

Andeol57:

I should be able to do some linear regressions on that. But it’s not a lot of data, so the results will need to be taken with a big grain of salt.

All right. I finished my data analysis. Here is an easier to read rank comparison table:

As mentionned earlier, it’s to interpret with caution, considering that we don’t have much data. In particular when you’re close to the extremes.

OGS KGS Fox Tygem Pandanet EGF AGA
20k 16k 19k 18k 20k 20k 20k
19k 15k 18k 17k 20k 20k 20k
18k 14k 17k 16k 19k 20k 19k
17k 14k 16k 15k 18k 19k 18k
16k 13k 15k 14k 17k 18k 17k
15k 12k 14k 13k 16k 17k 15k
14k 11k 13k 11k 15k 16k 14k
13k 10k 11k 10k 14k 15k 13k
12k 9k 10k 9k 13k 14k 12k
11k 8k 9k 8k 12k 13k 11k
10k 8k 8k 7k 10k 12k 10k
9k 7k 7k 6k 9k 11k 9k
8k 6k 6k 5k 8k 9k 8k
7k 5k 5k 4k 7k 8k 7k
6k 4k 3k 3k 6k 7k 6k
5k 3k 2k 2k 5k 6k 5k
4k 3k 1k 1k 4k 5k 4k
3k 2k 1d 1d 3k 4k 3k
2k 1k 2d 2d 2k 3k 2k
1k 1d 3d 3d 1k 2k 1d
1d 2d 4d 4d 2d 1k 2d
2d 3d 5d 5d 3d 2d 3d
3d 4d 7d 6d 4d 3d 4d
4d 4d 8d 7d 5d 4d 5d
5d 5d 9d 8d 6d 5d 6d
6d 6d 9d 9d 7d 6d 7d
7d 7d 9d 9d 8d 7d 8d
8d 8d 9d 9d 9d 8d 9d
3 Likes

What about training NNs on games by specific ranks on specific servers, optimizing not for winning, but for guessing the next move of that level on that server, and then play them against eachother to derive relative ranks on the servers?

1 Like

it would be easier to just use Kata
https://forums.online-go.com/t/server-comparison-ranks/50428/7
with more data, like 100 games from each OGS rank it would be possible to properly measure angle of border between ranks in (number of mistakes)X(size of mistakes) space

then we could plot perpendicular line to that border

make projection of these dots on this perpendicular line

average coordinates of these dots to compress them into 1 dot each

Then we would get unique dot for EACH ogs rank, they wouldn’t be too close to each other

Then we could do all this again for each rank of each other server and paint resulted dots on lines that are parallel to OGS line
Then we would get picture that perfectly compares all servers

2 Likes

Good point, I hadn’t thought of the connection to that thread