Life In 19x19

uPWarrior wrote:I decided to gather some data so that we can answer this question in the way that, for me, makes the most sense.

I started by downloading the FIDE list of all players standard ratings (243252 players, found at http://ratings.fide.com/download.phtml) and the EGF list of all players who have participated in any tournament (38928 players, found at http://www.europeangodatabase.eu/EGD/EG ... system.php).

The question now is: someone who is better than 50% of the chess players, what EGF rank should they reach in order to be better than 50% of the go players as well?

The next figure shows the FIDE vs EGF same-percentile isoline. It shows the rating that players need to have in both systems in order to achieve the same relative position in the respective populations.

go_vs_chess_ratings.png

According to this comparison, an EGF 10k (rating 1100) is equivalent to a FIDE 1800 player, a EGF 1d (rating 2100) is equivalent to a FIDE 2170 (expert), a FIDE master (rating 2300) corresponds to a EGF 4d and a FIDE grandmaster (2500) corresponds to an EGF rating of 2800 (EGF 7d is at 2700).

The EGF 10k = FIDE 1800 might be the most surprising comparison, the rest seems acceptable to me. We might be skewed to believe that 10k is easy to reach, or maybe this says something about the data sources.

(I had to remove everyone with a EGF rating of 100, as 5500 players clustered at this exact rating)

Could you somehow compare this data to goratings.org site?

No, because goratings only rates professionals. I would need an equivalent chess source, comparing with the FIDE general table would not work.
I could eventually remove every rating below grandmaster, but the threshold is quite arbitrary.


The reason that 10k shows up so high is that ~16000/33000 players on EGF have a rating lower than that.
Yes, almost 50% of the EGF players have a rating lower than 10k (after removing the 100 rating cluster, otherwise it would be more than 50%). The 50% cut-off in FIDE is at ~1800.
The more I look at the data, the more I believe that we underestimate how difficult it is to get to 10k.


I would be interested in doing a similar analysis with online servers data. If you have access to a chess ratings database from an online server, please let me know.

uPWarrior wrote:The EGF 10k = FIDE 1800 might be the most surprising comparison, the rest seems acceptable to me. We might be skewed to believe that 10k is easy to reach, or maybe this says something about the data sources.

Yeah, that is surprising. I think of 10k as being the level where you begin to play a game that is recognizably Go. I would peg the equivalent FIDE level, where you are actually playing chess rather than just shuffling pieces around, at maybe 1200.

I think the lesson is that it is harder to get to the point where you are playing a "real game of Go" than it is to get to the point where you are playing a "real game of chess". I definitely think there is a much larger gap between learning the rules and playing reasonably in Go than in chess.

dfan wrote:

uPWarrior wrote:The EGF 10k = FIDE 1800 might be the most surprising comparison, the rest seems acceptable to me. We might be skewed to believe that 10k is easy to reach, or maybe this says something about the data sources.

Yeah, that is surprising. I think of 10k as being the level where you begin to play a game that is recognizably Go. I would peg the equivalent FIDE level, where you are actually playing chess rather than just shuffling pieces around, at maybe 1200.

I think the lesson is that it is harder to get to the point where you are playing a "real game of Go" than it is to get to the point where you are playing a "real game of chess". I definitely think there is a much larger gap between learning the rules and playing reasonably in Go than in chess.

This may be an artifact relating to sample size and the number of people who pick up go for a short time and then give it up as compared with chess. I also think beginners are encouraged to enter tournaments in go much sooner than they are in chess because there's less competition otherwise, which may lead to this as well. I'd be interested in seeing a comparison for a country with a more active go population like, say, Japan, or perhaps a comparison between two popular online servers.

For those of us more familiar with AGA (or KGS) ranks, an EGF 10k is about the same as an AGA 6k. Using that more familiar scale, it doesn't really surprise me that a 6k is near the 50th percentile mark of tournament players. (Okay, maybe I'm a little surprised.) I can play chess at around a 1200 level with minimal practice or study. It's taken me a couple of years of consistent effort to get a stable 5k rank on KGS.

(Also note that the minimum FIDE rating you can start with is 1000, so this limits the bottom range of the scale.)

It seems reasonable to me to use the ranks/rating histograms to make an approximate equivalence. Go and chess are similar activities so it seems reasonable that the distribution of ranks over the two populations would be the same. That is people whose go rating is in the nth percentile play in the go population the way someone whose chess rating is in the same percentile would play in the chess population. That says nothing about how an individual player with a certain go rating would play chess.

A more meaningful question might ask for comparative ratings of people who play both games as much and about as well.

It might not be of any interest but I play go at AGA 5d level. I took a 50 year hiatus from playing chess, that is I hadn't played even one game for a 50 year time period. I had no knowledge of chess openings or tactics, even didn't know who the world's top players were. Then my brother's daughter's husband challenged me to play chess. Our first game was a draw, I lost the second, and won the next three games. His chess rating is 1700. We played "correspondence" games, with three days time for each move. This benefited me because I could take a lot of time reading the position before moving. I also found that my go focus on making my stones/pieces work together efficiently helps in chess. He now wants to learn go.

The difference is that there are a lot of lower level chess players have no FIDE Elo because low lewel tournaments are often only ranked in the national ranking systems. On the other hand, there are no national Go Ratings in Europe, and even very small Beginner Tournaments are EGF-ranked. Clearly this makes for a "too high" median chess rating compared to the Go Ratings. I know that in Germany, the median of the german chess rating(DWZ, usually 50-100 points lower than Elo on amateur level) is between 1500 and 1600.

dfan wrote:I think of 10k as being the level where you begin to play a game that is recognizably Go.

I'm not sure I agree with that. I found some of my old 21-22k games on KGS from 2005, and I think they're recognizably Go. The reading is atrocious, and there's a lot of tunnel vision, but I wouldn't call it "not recognizably go". Take one example, where I get slaughtered: http://eidogo.com/#url:http://files.gok ... rBlanc.sgf.

I think there's at least two questions you can have here:
1 (Easy): What is the range of skill that a player can have? If you wanted to be scientific, you could take people who have never played either game, spend a day teaching them one game, then measure an ELO style rating. It seems that Go is a bit wider than Chess here. On a scale where Fan Hui is 2800, there are players whose ELO is negative. This isn't particularly helpful for comparing a chess 2000 ELO player with a Go 2000 ELO player, unfortunately.
2 (Hard) Try to come up with some kind of way to compare the effort or place in the distribution, so that you can say that someone with a 1500 Chess is ELO is as good/better/worse than a person with a 1500 Go Elo.

Comparing specific rating numbers is meaningless. If we added 2000 points to everyone's ratings, people's relative playing strength would not change. Any comparison has to be based on players' percentiles in the cumulative histograms.

gowan wrote:Comparing specific rating numbers is meaningless. If we added 2000 points to everyone's ratings, people's relative playing strength would not change.

I agree, but maybe I should make it clearer that I am not talking about doing that. What I am talking about is the distance between the highest and lowest ratings players actually achieve. And that is meaningful.

Consider Tic-Tac-Toe: there's basically three classes of players: 1) those who play badly, 2) those who play sensibly, but didn't quite internalize the trick, 3) those who never lose to anyone. If you did ELO ratings for Tic-Tac-Toe players, the gap between someone in 3 and someone in 1 would be relatively small. Less than 1000 points, I would guess. So there's a real sense in which the skill range in Chess is bigger than Tic-Tac-Toe (stay tuned for other novel claims ). This is true whether you call the worst Tic-Tac-Toe player 0, 1000 or -1000.[1]

You can do the same comparison with Chess and Go. When I mentioned that Fan Hui was 2800, my point isn't "is that less or more than Magnus Carlsen?" It's that in Go, Fan Hui's 2800 is on a scale that includes players all the way down to 100, and Chess only goes down to 1000.[2] The range of skills in Go is somewhat broader than in Chess.

Any comparison has to be based on players' percentiles in the cumulative histograms.

This is also a weak comparison. As others point out, it's heavily dependent on who acquires ratings, and it seems clear that in the US and Europe, Go players are much more likely to get rated when they are beginners than Chess players are.

[1] This somewhat understates how much deeper Chess is than Tic-Tac-Toe.
[2] I'm not going to be precise about this. There are actual Chess players below 1000 ELO, they just don't have ratings. There are actual Go players below 100 ELO, they just don't have ELO ratings in Europe (but there are established 30 kyu players in the US!).

hyperpape wrote: You can do the same comparison with Chess and Go. When I mentioned that Fan Hui was 2800, my point isn't "is that less or more than Magnus Carlsen?" It's that in Go, Fan Hui's 2800 is on a scale that includes players all the way down to 100, and Chess only goes down to 1000.[2] The range of skills in Go is somewhat broader than in Chess.

Board size matters. My guess is that the range of skill at "Western" chess is broader than the range of skill at 9x9 go. And there was a form of shogi played on a 25x25 board! How broad was the range of skill for that game!

It makes sense to try comparisons between rated go players versus rated chess players. Certainly there are a lot of chess players who, if they were rated, would have ratings below 1000. Similarly there must be go players who, if rated, would have ratings below 100. On the other end, there are many go players who are not rated but if they did get a rating it would be dan level. I base this on the number of apparent USA online players with strong playing levels but who are not rated by the AGA. And there are quite a few strong go players in the USA who are immigrants from Korea or China but who do not participate in rated tournaments. It would of course affect the histogram of player strengths if it were possible to include all these players. Of course that is not possible in practice, hence my statement that comparisons should be made based on populations of rated players.

uPWarrior wrote: (I had to remove everyone with a EGF rating of 100, as 5500 players clustered at this exact rating)

If you remove the bottom 14% of go players, you should also remove the bottom 14% of chess players, IMO, to get a fair comparison. The EGF system has an artificial bottom, most of those rating 100 players are somewhere between 20k and 35k, and should have negative ratings.

Yes, but the bottom chessplayers(even more than 14%) are already removed by looking at fide rating as
1. there is a minimal Elo(I think it is now 1000, but not so long ago it was quite a bit higher), if your first tournament is weeker than that, you dont get one.
2. Many tournaments for players at weeker levels are only rated in national rating, so most week players dont have an Elo(despite having played in tournaments), while almost all strong players do