Mamumamu list is a well-known rating list for professional go players. Most of players at least heard of it. But because it’s in Japanese it can be a little annoying to use. Even I haven’t really gotten past the main list until now.
Thus I decided to make an overview of the features. There’s quite a bit of information there.
Index page
http://mamumamu0413.web.fc2.com
The main part of the website are rating lists, one for Japanese players only and another for all professional players.
Japanese domestic rankings
Ratings for Japanese players from January 1951 until now. Domestic games between players are used as well as international games between Japanese players (including some informal games).
Newest rank list
https://mamumamu0413.web.fc2.com/rating/japan/ranking.html
The list itself is mostly self-explanatory. Association field differentiates between Kansai Kiin and Nihon Kiin, and different branches of Nihon Kiin too (Tokyo, Kansai, Chubu). Rating values go from around one to around ten, this might be a little unusual after the thousands of rating points we’re used to seeing. The table can be sorted by some of the fields.
If one scrolls all the way down, there’s a second table there that shows players whose rating deviation is too high to be included in the main table. Players like Catalin Taranu are there because they stopped playing, or like Cho Kosumi because they haven’t played enough games yet.
If one scrolls to the bottom, there’s a table showing how rating difference corresponds to win probabilities.
Past lists
https://mamumamu0413.web.fc2.com/rating/japan/archive1.html
This is simply a collection of past lists for each date. Japanese rankings start at 1951, until 1988 it’s updated every 3 months, and since then it’s every month.
Age breakdown
https://mamumamu0413.web.fc2.com/rating/japan/archive2.html
Age breakdown is a nice little feature of the list. In this one we aren’t comparing ratings at a specific date, but ratings at a specific age. Thus comparing ratings of players when they were that specific age. And you can see who, for example, was strongest 20-year-old player. The precision of age is down to a month.
Let’s see who is the strongest Japanese at 40 years old.
Apparently it’s Kato Masao in 1987, then Cho Chikun, then Yamashita Keigo who reached 40yo only recently and then Go Seigen.
I imagine it’s more like a toy because we have to accept that these ratings are comparable at different time points.
Player breakdown
https://mamumamu0413.web.fc2.com/rating/japan/archive3.html
These pages list rating of each player over time. The players grouped by birth year for easier navigation. There’s a whole group of players with unknown birth year which is curious.
№1 player over time
https://mamumamu0413.web.fc2.com/rating/japan/archive5.html
This is quite self-explanatory, a single table containing top player for all dates.
Rating evaluation: years
http://mamumamu0413.web.fc2.com/rating/japan/score/histogram2.html
This page compares the win percent and expected win percent for each rating difference, separate for 2018, 2019, 2020, 2021 years. In table form.
And a graph showing theoretical red lines and real values.
Rating evaluation: komi difference
http://mamumamu0413.web.fc2.com/rating/japan/score/histogram.html
The same table for different komi values (6.5, 5.5, 4.5, no komi).
To correct 5.5 komi black is added 0.13 rating points when calculating win probability. To put it into perspective, 0.15 rating difference corresponds to 54% win probability. So if two equally strong players played a game with 5.5 komi, black would be expected to win something like 54% of games.
To correct 4.5 komi black is added 0.24 rating points. 0.25 rating difference is 55.5% win probability.
To correct no komi game black is added 0.74 rating points. 0.75 rating difference corresponds to 70%.
There’s some more considerations in the about komi section.
World rankings
The same thing but for worldwide pros (Japan, China, Korea, Taiwan, no Euro pros). This is a separate rating, and it’s not comparable to Japanese-only values. The rating goes only until 1990 into the past.
List pages are mostly the same as in domestic rankings. Newest list, past archives, age breakdown, top player history. There’s no player-specific history, but there’s a separate worldwide female ranking.
Rating evaluation
Rating evaluation in this section evaluates international and foreign domestic matches.
http://mamumamu0413.web.fc2.com/rating/world/score/histogram2.html
On the first page it’s Japanese players against foreign players for each of 2017, 2018, 2019, 2020 years. Nothing interesting there except that since international matches are rarer, the graphs are less pretty.
http://mamumamu0413.web.fc2.com/rating/world/score/histogram.html
On the second page it’s the same but for total matches for each of Japan, Korea, Taiwan, China against everyone else. Plus for domestic matches for each of Korea, Taiwan, China.
I like how there’re so many Korean games the graph is very smooth. But I couldn’t help but notice the shape of experimental values doesn’t exactly match up, it looks more linear.
Rating system explanation
http://mamumamu0413.web.fc2.com/rating/explanation.html
This section outlines the rating system. Main points are:
- Rating system uses Glicko-2. All praise Glicko.
- Pure Glicko-2 is insufficient when it comes to the world rankings (presumably because there’re only few games so the rating pool splits into several parts). They apply some additional calculations to move the countries ratings against each other based on the international results.
- Japanese games published by Nihon Kiin, Kansai Kiin are used for calculation (including TV games omitted in the bulletins). For international ratings games published by federations are used although other information about games may also be included.
- Forfeits aren’t included in the ratings. Games are considered equal and not weighted by importance.
- Amateur games aren’t included but games of amateurs who later became pros may be included.
- The data about Kansai Kiin games is complete from around 2008. Before that the data is scarce except for games against Nihon Kiin pros.
- For TV games the dates of the games are unknown, and broadcast dates are used.
- Win probability is calculated as 1 / { 1 + 3^(rating difference) }
- Exact formula for komi correction is apparently: rating difference correction = 0.74 * ( 6.7 - komi ) / 6.7
There’s more but it’s a little technical.
Hikaru no Go characters strength
http://mamumamu0413.web.fc2.com/comic/hikarugo.html
A large section dedicated to the estimating the strength of the characters. This is the biggest section on the website, and maybe deserves a separate translation by someone who knows about it.
