Most of you should have heard about Tetris. Similar to Go, Tetris has simple rules and high emergent complexity, and a wide skill gap between strong and weak players. Tetrio (one of the online Tetris servers) uses Glicko-2, which allows for comparison with Go ratings.
Inspired by the go-chess comparison threads, I have attempted to make an overall go-tetris rank comparison table:
Go rank
Go rating
Tetrio glicko
Tetrio rank
Shin Jinseo
3700
Ke Jie
3600
Top 10 pros
3500
4200
Best player
Top 25 pros
3400
4079
X
Top 50 pros
3300
3959
X
Top 100 pros
3200
3838
X
Top 200 pros
3100
3718
X
9 dan professional
3000
3597
X
7 dan professional
2900
3477
X
4 dan professional
2800
3356
X
1 dan professional
2700
3235
X
6 dan
2600
3115
X
5 dan
2500
2994
X
4 dan
2400
2874
X
3 dan
2300
2753
X
2 dan
2200
2632
X
1 dan
2100
2512
X
1 kyu
2000
2391
U
2 kyu
1900
2271
U
3 kyu
1800
2150
SS
4 kyu
1700
2029
SS
5 kyu
1600
1909
S+
6 kyu
1500
1788
S
7 kyu
1400
1668
S-
8 kyu
1300
1547
A
9 kyu
1200
1426
A-
10 kyu
1100
1306
B+
11 kyu
1000
1185
B-
12 kyu
900
1065
C+
13 kyu
800
944
C
14 kyu
700
824
C
15 kyu
600
703
C-
16 kyu
500
582
D+
17 kyu
400
462
D
18 kyu
300
341
D
19 kyu
200
221
D
20 kyu
100
100
D
*Go rating as on goratings.eu, or approximately goratings.org elo minus 100
I agree it would be rather uncorrelated, this is a comparison of the amount of combined time, effort, skill and talent required for each level, and most people won’t be equally devoted to or have equal talent in these two games.
Isn’t Glicko2 supposed to be used in one-to-one matches situations?
Is tetris so?
I’ve seen arcade games where you could play tetris against an opponent, but I always thought that Tetris way of working was related only to one person’s skills. I don’t see how your score could be affected by your opponent’s score.
This is an example of 2-player Tetris, from the grand finals of a recent tournament. To win you need to do difficult clears (Tetrises, T-spins, all clears and combos) to send garbage (the grey blocks), which will kill your opponent if it pushes them up to the ceiling.
Those amateur go ratings look like EGF ratings, which are not Elo ratings (nor Glicko ratings), because they are defined/calculated differently. It’s mostly coincidence that the lower pro ratings from goratings.org line up (more or less) with the higher amateur (~7d) EGF ratings.
Is it a coincidence? I believe the EGF ratings use a variation of the Elo formula, and have been parametrised specifically to fit kyu/dan ranks. So it looks more a design decision and less a coincidence?
Yes, those are design decisions. But some of those decisions are arbitrary, so I consider a rough match between EGF and WHR (goratings.org) around 2700 rating mostly coincidental.
Rating gaps have a clear meaning, both in EGF ratings and WHR. But the meaning of a rating gap is not the same in both systems. In EGF it means handicap for 50% winrate, in WHR it means winrate in even games (as in OGS rating system and regular Elo rating systems).
Absolute ratings are determined by the rating anchor chosen by the creators of the system. In the EGF rating system, Ales Cieply decided around 1995 to define absolute EGF rating = 2000 + dan * 100. So 1k = 2000, 1d = 2100 and 7d = 2700. This is fairly arbitrary. He could just as well have chosen 3000 or 1000 for 1k. I think he chose 2000 to have ratings in a similar range as chess Elo ratings for “fairly competent club player level”.
Ales’ decision to fix 1 rank gap to 100 EGF rating gap was probably influenced by the previously known observation that around EGF 1d, 1 rank gap corresponds to about 2:1 odds to win an even game, which corresponds nicely to an Elo rating gap of about 100 points. I think Ales just exploited this coincidence (serendipity?) in making EGF ratings similar to chess Elo ratings around 1d level.
Chess rating systems often use 1500 as the initial Elo rating for players new to the system. Because of that, 1500 is the rating anchor for most chess rating systems (at least it used to be). It corresponds to the typical chess player (intermediate level). OGS does a similar thing, and currently OGS rating 1500 corresponds to aboutk 6k, which would be a typical OGS player.
On the one hand, this is a funny example of people not understanding the relative nature of a ranking system. On the other hand, if they were trying to compare the numbers to EGF ratings, it’s no longer a total coincidence that there is now a point where the systems line up.