FIDE titles and EGF Go ratings
tderz: I do have the following questions:
Who of you is a tournament player in Chess and Go?
I would like you to compare your efforts in time & intensity of study for that moment where your Go and Chess ratings had the same value. (tderz)
1) How can I compare the Elo-ratings in Chess and Go w.r.t.
to their difficulty to achieve? [1]
Do you have FIDE and an EGF rating and could compare your personal efforts?
sagemerlin: I am a Brazilian Chess (1818 FIDE) player and arbiter and go player (14k Little golem, probablay about 5-3k in the Nihon ki-in Brazil). About 1 year ago I made a comparation of the Elo-kyu systems so I could compare my Chess strength to my Go's. So the first thing that i thouth was: "A Go profesional its easy to recgonize he is at least a 1pro-dan, what about a Chess player?" So, usually in Chess we call a player as professional when he is about 2200 Fide (Candidate Master). So 2200 Fide = to 1p. and then I made this way, 6 amateur-dan = 2100, 5d = 2000, 4d = 1900, 3d = 1800, 2d = 1700, 1d = 1600, 1k = 1550 (I belive that the diferance between kyu players is smaller then between the dan players) 2k = 1500 and so on. Whe you think up the ladder I am not sure wich way should take, 100 diff, or 50 diff, if I took 100 the strongest Go players would be about 3000 (probably the rigth way until now since this is the strength of the most strong computers, and in GO computer are still weak). If you took the 50 dif the strongest players will be about 2500... I think that its a bit low, any way, I think that my "table" works quite fine...
2) Who plays (played) Go and Chess and has (had) achieved the same
Elo-rating in Chess (FIDE) and Go (EGF-ELO).
e.g. FIDE 1500 & Go nkyu
FIDE master = 2300 Elo points; Go 3dan = 2300 Elo points etc.[2]
ilan: As far as I know, FIDE ratings start at 2200, except for women.
MRFvR: It was once like that, but they have been lowering their limit. Now Fide ratings starts at 1400 (but they have plans to move down up to 1000). This is a good thing, since allouds even patzers like me to be rated 8-)
3) How much time did you need for your getting your resp. Elo ratings (rank, dan)? [3]
4) What is your percentile (1),
percentile (2)
in the world or in your country in the respective sports (difficult to obtain figures (worldwide) + the country might be strong in Chess but weak in Go a.v.v.) [4]
- How many Go players are there? I remember an internet source (or the GW) speculating about 30 millions.
- How many Chess players are there?[11]
tderz: I realize now that the parameter percentile is quite un-interesting to know. It is - still difficult to establish - more wishful to know the number of players with a higher rating than yourself. Even for the EGF figures there will be a high number of players around 1, 2, 3 and 4 dan who are not listed with an EGF-rating. There will be an even bigger and more important dark number for the important countries China, Korea and Japan.
! Hence, the individual & personal responses of you w.r.t. questions 1 (effort), 2 (rating) and 3 (time) for that moment where your Go and Chess ratings had the same value are more important here.
5) I might not be correct with my following figures: i) If Chess uses a 200 Elo points interval for a winning probability (33%/67%; correct?) and Go a 100 Elo point interval per standard deviation; and ii) if Kasparov has now an Elo rating of 2804 (ref.: http://www.fide.com/ratings/advseek.phtml) and iii) Yi Chang-Ho, (9p) has 2960 Elo points (on 2004-12-10; source http://www.goweb.cz/progor/)
-> ii) then a lowly player of FIDE Chess Elo 2200 rating playing Gary (Elo difference = 600; => 3 standard deviations away => 99% probability for winning, resp. losing) had theoretically a (0.33)exp3= 3.6% winning chance vs. Kasparov (if Gary is not retracting to remis :-), or all remises are excluded), hence winning 3-4 out of 100 games.
Now in practice this must be highly unrealistic. Gary would win all simply by letting the player make the serious mistakes and not voluntarily taking own risks (I guess; I play some Go, but I do not play Schach, but schwach (German = weak)). Again, all remises - wished - by the weaker player would not be counted.
Relation of (long) speculation 5) with question 4 [11]:
A 2200 Elo points Chess player would be (theoretically and only if my figures are not flawed) approx. 3 standard deviations away from the best, highest-ranked chess player (Kasparow). However a 2200 Elo (EGF) Go player (= EGF 2 dan) would be 6 standard deviations away from todays best Go players (e.g. Yi ChangHo with guessed 2900 Elo). This difference in standard deviations (3 for Chess vs. 6 for Go) for probability of winning against top players should express itself in percentiles (to which %-age of the best players do you belong?).
-> iii) For Go it is not much different. Due to the different nature of the game, Yi could always shelter ( :-) ) behind 100% sure 0.5 point wins and avoid the tempting fair 60-100 point whipping. ( http://en.wikipedia.org/wiki/Go_ranks_and_ratings#Winning_chances_are_more_than_cumulative) In other words, by moderating the risk exposure and steering for a lower win in points (not a 50% chance for 140 points), the better player increases the probability for this win. This behaviour increases the chance up to 100%. [5]
6) (I do not indulge in nor understand the mathematical difference to the ratings here http://holigor.fcpages.com/baduk.htm, 1SD = 200 Elo points for Go) [6]
I do hope that several Chess&Go players here on Senseis could report on their experiences of how they acquired their resp. ratings (time, effort etc.); Many thanks in advance, Thomas
Discussion
ilan: I am starting to find this page somewhat annoying due to the errors and mistakes which the author tderz continues making despite being corrected. When someone asks for information from people who know what is going on, that information should be incorporated, otherwise the whole exercise is just a waste of time and effort. In particular,
1. The term "ELO" is incorrect and must be abandoned.
tderz: Ilan, I am terribly Sorry if writing ELO instead of elo was my foremost and gravest error. It just slips my attention and many other linguistic errors occur as well. I would like to behave from now on, but are still confused: the very [7] ("see also this intro on ELO" (sic) http://gobase.org/studying/articles/elo/, which I, of course, checked before) which you cite below, writes - as you do - ELO. Yet I see your point that elo (or Elo?) makes mores sense. All the information as FIDE elo-rating of e.g. 1600 points, FIDE master and expert I got from the very FIDE-websites. I thank you for your continuous effort and contribution here on this site. I can feel from your writing that you wanted to express your annoyance and you were not the anonymous 10.53.72.156, who modified the somewhat vague version 2 (below) on 11 February 2005 - 15:37.
2. There is a complete misconception about FIDE rankings which are limited to games played in relatively infrequent round robin international at the highest levels. In particular, only the best chess players have FIDE rankings, and, apart from dedicated international players, even these play a minority of their tournament games in FIDE tournaments.
3. FIDE rankings start at 2200 and there is no such thing as 1500 FIDE which is as ridiculous as saying that there is a 20K professional. (women's ratings start at 2000 though, so some ratings are < 2000).
ferdi: That was the case for many years, but since then the FIDE has lowered the starting point in general, for both men and women, from 2200 to 2000 to 1800 to 1600 points right now. Elo 1500 is still impossible, though ;-)
-- it has now been lowered even further. Although rare, I have seen FIDE 1500 players.
4. There is no such thing as a FIDE expert.
[1]
A) Use a histogram approach
tderz: Please give me your figures. How much time/effort did you need to achieve the same Elo rating points in Chess as in Go.
What does effort have to do with it - comapre the percentages of people with top grades (in bands obviously) to find some comparison. Presumably you should get some sort of Bell curvish figure?? Afterwards I could create any diagram I wish. Indeed a histogram chart as subgroup of colum diagrams might be appropriate once I have much numerical data (ratings versus times). First the data please.
Dear A), do you understand now, what I would like to know?
C)
[2] A) ...
C) what relevance is this?
tderz: We'll see, first I would like to hear from your experience.
[3] A) ... C) Time is dependent on a persons inate ability, access to knowledge, time available to learn, age, etc
tderz: I know, please, how much is it in your case? Do you play/ed Chess and Go on the same level and got ELO ratings?
Then you are qualified to answer. I am not, because I do not have a chess rating.
[4] A) Go and look it up in the rating lists ;-p
tderz: I do not have your name - do you play/ed Chess and Go and are listed under the same name? Do you have an EGF-ELO-rating? Furthermore your experience of time & effort for achieving your resp. levels are not listed there. You are most probably the only one qualified to tell me your experience.
Thank you for participating, A).
[5] A) What was the question?
tderz: You got me here. If you are Chess and Go player, please tell me about your feelings and imagined winning chances when playing someone with 100, 200, ... 600 difference in ELO points, up to the strongest players.
[6]
ilan: The first thing to understand about the Elo system is that it was named for the actual person who invented it, Arpad Elo, so the spelling "ELO" is misleading, since all capitals is reserved for acronyms, in English, that is. [7]
As far as I can tell, the rough correspondence 2000 <=> 1 dan with each dan being a 100 point difference works fairly well, and also in comparing Go and Chess ratings, at least for levels above 5K. I don't think the comparison hold up well for weaker than about 5K, in particular, in chess, I don't believe that there is a strength level corresponding to a rating of less than 1000 apart from complete beginner, though I have seen US Chess rating lists with negative ratings (which would theoretically correspond to 20K in Go). [8]
However, the rating comparison only goes so far. There have been 2700 grandmasters who lost tournament games to 1800 players (Bent Larsen in the Canadian Open in the 1970's)[9], but it is hard to imagine a 7d player losing an even tournament game to a 2K, if they were ever to be paired up, that is. I should mention that I did see a 8d lose an even 9x9 game (.5 komi) to a 10K computer on KGS http://files.gokgs.com/games/2004/12/8/Arwen-GnuGoBot.sgf This example is quite telling, as 9x9 is more similar to chess in terms of rank difference, since you cannot always recover from one error or risky move, so you can always lose to someone much weaker than you. [10]
tderz: Ilanpi, thanks for replying. I thought of you when I created this page and checked your info. I may have contacted you individually if I had found rating info on your home page.
I would like to stay focussed on my specific questions first: Do you have FIDE and an EGF rating and could compare your personal efforts?
ilan: To answer your main question, I would definitely say that, up to first order, my progression in chess has been about the same as my progression in go. You have to take in the fact that I am now 30 years older, which means that I have more and less mental abilities. My analytical skill was completely undeveloped when I started chess, and completely developed when I started go. On the other hand, I had total recall for useless facts back then, and could concentrate on many things at a time, which I cannot do now. More generally, I was discovering the world, and things made more of an impression on me. On the other hand, I know a lot more about myself and the world now. All in all, the plusses and minuses seem to have evened out.
Otherwise, I am continually amazed by the number of levels of awareness one has in go, that is, how one has to completely change one's outlook on the game in order to improve. That was not so true in chess, where you could improve by simply studying and knowing more openings, endgames, and tactics. Maybe it is for this reason that there are more opportunities for one's improvement to stop in go. In chess, I certainly knew why most people stopped improving -- it was because they like the social aspect of learning openings, you could be a "fan" of your favourite player by copying his opening, the equivalent of wearing the jersey of your favourite football player, and were turned off by the austere nature of the endgame and its unsocial aspect -- you can't play your favourite players endgame, because they will all play it the same correct way. In fact, my advice to most chess players who stop improving is to burn their opening books (OK, just store them somewhere out of reach), which they of course ignore because it is too harsh, just like The Truth they are busy avoiding. Nothing like that in go, it looks like...
I also recently took up roller skating after many years of bike racing, and I find a good analogy between cycling versus skating and chess versus go. Cyling and chess will both give you the power base you need to compete, but I find that both go and skating need something extra, which is a sense of balance. In particular, miles and miles of training on the road will not achieve as much results in speedskating as short technical sessions indoors where you learn to shift your weight to develop power. On the other hand, road cycling is well known to be the last refuge of uncoordinated athletes with lots of power (I know a number of world class cyclists who cannot ride no hands). I see the analogy in go, for example, when there is a "tactical" player who always tries for interesting long combinations. He will be able to make life for a hopeless group, or win some fight in a corner, but will lose because he overlooked the "vague" moves which didn't have any direct tactical significance. Anyway, I certainly play like that... However, in chess, if you win the tactics, you will usually win the games.
[7]: see also this intro on ELO http://gobase.org/studying/articles/elo/
[8]: EGF-Info: http://gemma.ujf.cas.cz/~cieply/GO/gor.html
http://gemma.ujf.cas.cz/~cieply/GO/gor.html
[9]: To make up his reputation, wasn't Bent Larsen also the only one with a positive score (he won two games) against Fischer, before he was crushed as Taimanov and Petrosjan?
- No he wasn't players from Vik Pupols to Boris Spassky to Efim Geller to Gligoric all had positive scores over Fischer.
[10] I agree with you.
PurpleHaze: A better testament to Larsen's strength is The USSR vs The World match, when Fischer demanded first board Larsen objected and Fischer uncharacteristically backed down quickly conceding that he no reason to believe that he was stronger.
To suggest that Larsen in the 70's was a 2700 player is also inaccurate. Spassky and Petrosian as World Champions were about 2690, and he was less than that. Further a rating of 1800 CFC (Chess Federation of Canda - Federation Canadien d'Echec) would be above 1950 on the world scale (in about 1979 we all received a boost of 154 points to bring us closer).
Current FIDE ratings are nearly 100 points higher than 1970's era ratings.
The best example I know of would be Bruce Amos who is a chess IM and a 6d. Who during his active periods in the respective games would have been in the Canadian top 5 for more than a decade. Shame he is playing bridge now.
ilan: Well, I disagree with you regarding Canadian ratings vs FIDE ratings during the 1970's. That was the exact time and place when I was active in chess, and I have no reason to believe that Canadian ratings were lower than FIDE ratings. As I recall, most Canadian players of the time had CFC ratings which were not lower than their FIDE ratings (of those who had both).
Otherwise, there is irony in the Larsen losses since his defeats by computers were often cited by 1970's computer chess advocates who would then fail to note his losses to human A players (< 2000 rating).
ferdi: To give at least one answer to at least one of tderz questions: Fide around 2250, egf around 2250 (I don't know exactly :-) FIDE 2242, EGF 2254 (Feb2005), playing both games for more than 15 years.
tderz: Many thanks, ferdi. You are the first player here to respond to my questions 1+2+3. I guess that you did the same effort for both games - is that correct? If yes, would you then say that your FIDE and EGF ELOs of 2250 are directly comparable? That was the cause for all my questions. Could you try to give me your opinion on questions 4 and 5?
Many thanks in advance, Thomas
ferdi: I wasted the same amount of time, yes :-) But since I don't think Elo or EGF ratings are good for anything, I don't feel qualified to answer questions like "Are Elo and EGF ratings directly comparable?". And I can't answer your questions 4 and 5 because I don't understand them :-(
tderz: Questions 4 & 5 are meant for you to judge your very, very ... theoretical chances indeed: In 100 even games with each Kasparov (Chess) and any of the top 9 dans Yi ChangHo, Gu Li or Cho U (etc.) - where would you think your winning chances are higher? Ok, ok , I would think that you'd lose all (excluded Chess remises). Yet if you count the levels of players between you and the resp. top players with each next-higher level having a 2/3 winning probability against the lower level. How many more levels would that be? I think 3 in Chess vs. 6 in Go (for about your rating). So you did the the same effort in both games and I would conclude from there that Go is the deeper game. Does a Go 3dan (EGF) equal a Chess 1.600 FIDE rating (or ca. 2250)?
Second, theoretical question: What do you think is easier for you now: Going to EGF-ELO 2350 (3-4dan) or FIDE 2350? What title would that be? Does your present chess rating have a title (name)?
An easier, similar question: coming from FIDE and EGF-ELOs 2150 and promoting to 2250, what felt easier?
dnerra: A fried of mine is ELO 2430, and has just become 3d. I think he has invested a lot more in chess than in go.
tderz: Adding to dnerra's report: a Go aquaintance of mine in Berlin, A.K., has played in the German Chess Bundesliga (ELO 24xx, I forgot the exact value) and learned Go later, now 3dan.
He told me that he invested way more time in Chess than in Go (also because he started younger).
[11] The difficulty with both figures would be the cut-off point for beginners. On the other hand both, FIDE and EGF-lists are finite. I guess one could take the FIDE list as representative for the whole population of Chess players.
However the EGF list is not representative for all Go players, because the high dan region is not as dense as in China/Korea/Japan.
Does an International Rating list exist which lists all players from all countries, where, e.g. the ratings are converted to one standard?
An idea would be to take players who are listed in several lists (e.g. AGA and EGF and ...) as anchors for conversion (I guess one still misses out on Asia here).
Question: Do rating lists based on ELO statistics exist for Asian amateurs?
tderz: I had the thought is a 2300 ELO-Go rating comparable with FIDE-"expert" or, if not, with what FIDE rating is it comparable. Perhaps I should take standard deviations (6 s.d. to top Go) and deduct 6 times 200 ELO points (=1.200) from top Chess ratings. This would result in 2800 - 1.200 = 1.600.
Does a Go 3dan (EGF) equal a Chess 1.600 FIDE rating?
HolIgor: Comparing to the old Soviet chess or modern Russian ranks: 1 dan (about 2000 EGF) is like a 1st grade player, 3 dan ( 2200+ EGF) is like a candidate, 5 dan ( 2400 +) is similar to a master, 7 dan is like a grandmaster. As far as I know 1600 in chess is the level of a casual player, perhaps, something like 5 kyu in go.
larsen I played chess at a level somewhere between 1600 (Polish rating) and 1800 (English active chess rating about 145 BCF). In Go, I'm 6 kyu, yet I feel my chess play even now is less erratic.
Koblentz?: Today 6/13/07 I found the following link with a pretty good discussion http://en.wikipedia.org/wiki/Go_ranks_and_ratings
However, it is hard to agree with their table at the high end which equates a 1 dan professional with FIDE 2660 in chess and a 9p with FIDE 2900, a rating no one has ever achieved, not even Kasparov. Clearly the Chess World Champions - Kasparov, Karpov, Fischer, were all 9p equivalents, as were Alekhine, Capablanca, Lasker and probably Botvinnik, Tal and Spassky, Petrosian, Smyslov at their peaks. Euwe, Steinitz, Kramnik as well as a few current day players and several through history who barely missed winning the world championship (notably Bronstein, Keres, Korchnoi and maybe Schlechter) may be considered 8p.How many 8p and 9p players have there been in the history of go? If it is roughly a dozen of each then this mapping seems good.But even if you granted that 9p = 2900, then 1p =2660 means that the gradation between professional ranks is only 30 Elo points. This seems doubtful that the gradation for all amateur ranks is 100 points then suddenly changes to 30 points for professionals. Also, chess players well below the 2660 level are considered chess professionals.
The 30 Elo point difference between professional ranks is correct. Between amateurs, a 1 rank (kyu/dan) difference corresponds (roughly) to a 1 stone handicap (eg players 400 pts apart should be able to play a reasonably even game at 4 handicap). The rank difference between 9p and 1p is about 2 to 3 stones, so professional grades are much more finely grained. It should further be noted that the Elo formula used for Go is slightly different from that of chess. The k-factor and standard deviation used are variable, see http://gemma.ujf.cas.cz/~cieply/GO/gor.html (point (3) of the system description has a table). This is done to make 100 Elo points roughly equal to one kyu/dan grade. --Herman Hiddema
Also, note that the table given on the wikipedia page does not try to compare Go and Chess ratings. The numbers given are the actual Elo ratings of pro and amateur players as calculated by the EGF (see link above). The range of distinguisable skill levels in Go is much larger than in chess (roughly twice as large, I think) --Herman Hiddema
Herman Hiddema: Regarding the comparison between Chess and Go level. I would make the following comparison:
- Those players that have achieved an Elo rating of over 2800 (Kasparov, Kramnik, Topalov, Anand) could probably be compared to those that are dubbed Kisei (Go Sage), or those that have achieved other remarkable results, eg Dosaku, Jowa, Shusaku, Go Seigen, Lee Changho, etc.
- Players in the range 2750-2800 might equal those historically called Meijin and those professional players that have won many mayor tournaments (eg Sakata, Kobayashi Koichi, Rin Kaiho, etc.
- Players in the range 2700-2750 would then constitute remaining 'strong' professionals (7p-9p, mayor tournament winners).
- Players in the range 2600-2700 would be medium level professionals (4p-6p).
- Players in the range 2500-2600 would be the weakest professionals (1p-3p), strong Insei (pro students) and 8 dan amateurs.
- Players in the range 2400-2500 would equal top amateurs (6 dan and 7 dan, winners of European or US Championships) and weaker Insei.
- From here I would equal 50 chess Elo points to about 1 go kyu/dan grade, so: 2350-2400 = 5d, 2300-2350 = 4d, 2250-2300 = 3d, 2200-2250 = 2d, 2150-2200 = 1d, 2100-2150 = 1k, etc.
- At the end of the EGF scale, this puts a 20kyu (beginner) as comparable to a chess player with rating 1150-1200.
- Go grades do not stop at 20 kyu, and some childrens tournaments use grades in the range 30 or 50 kyu, which would equal chess rating 0-1000
Koblentz?: I stumbled across this page accidentally but now would like to add my two cents worth. I have been a rated USCF player for 36 years and achieved my present rank of Expert 30 years ago (current rating 2054, top rating ~ 2100). While working at Bell Labs I started playing Go and found it absolutely fascinating. I bought books, studdied and was coached by several players who were in the range 1 kyu to 6 dan amateur. I played intensely for one year and played in the 1983 US Go Championship as a 4 Kyu, obtaining an even score 3-3. Of course, my progress was helped by my previous chess experience, but it was clear to me that 4 kyu was significantly below USCF expert in "equivalent" terms, i.e. I clearly had much less experience and had done much less work in go than in chess. I was curious about the same question you raise here so I took the US Go Association rating list (nearly 1000 players at the time) and recorded the frequency of rankings of all players in groups, and then equated the frequency of chess groups. What I found (this part is obvious) is that there is more granularity in the Go rankings that in chess classes, e.g. a single chess class spanning 200 USCF or Elo points might equate roughly to 2 or 3 Go rankings. I also took 9 dan professional to equate to the super grandmaster class in chess (FIDE ~ 2750 to 2800 or so with recent FIDE inflation), 8 dan professional = world championship candidate, 7p = grandmaster, 6p = "weak" grandmaster, 5p strong intermnational master ... 1p or 2p ~ FIDE Master (FIDE 2300-2350). If you do this then the USGA amateur ranking distribution lines up pretty well with chess ratings below 2300. As I recall (I'm going from memory here) my 4 kyu go ranking was roughly eequivalent to a Class B chess player, 1 kyu ~ Class A, 2 dan amateur ~ Expert, 5 dan amateur ~ National Master. Again, with recent inflation (since 1983!) the 4 Kyu might be in the low to mid Class A chess category, and extrapolate from there. If someone has the time and interest, I think it would be useful to take my basic approach but with more recent data. Regards,
Dr. Michael Koblentz
P.S. as an additional data point, one of my playing partners (Art O'Leary) was in 1983 a strong A player in chess, nearly Expert, and the last I checked he had moved from 1 kyu to 1 dan amateur in go. His ratio of go effort / chess effort was higher than mine for longer duration so this may be an argument supporting USCF 1900-1950 ~ 1 dan amateur in go.
Michael (Dr. Koblentz), thanks for your input! (tderz)
The curve, resp. rank distribution comparison is an interesting thought.
ilan: Yet another difference between chess and go: When I was learning chess, when I got to about 1800 I was fairly confident that I could beat any person who had never studied the game, that is, who had progressed only by playing (I now realise that this was probably true, except for a handful of exceptions). On the other hand, I don't yet see any equivalent strength in go, maybe 5 dan?
tderz: How should I understand this who had never studied the game and progressed only by playing (to where?)?
If the ratings are the same (Chess 1800, Go 2k or 5dan) it should not matter for the winning probability which style you apply or have. If the homegrown, most probably stubborn peasant's fighting style + willpower results in a rating X (e.g. 5dan) without theoretical opening finesses then it should be equal to the well-founded player with theory knowledge. Or did you mean, never lose to a beginner anymore? Of course beating any person means in Chess, confidently being beyond remis chances ("Remisbreite ", even against beginners). See e.g. "Nimzowitsch kam gegen den Amateur nie aus der Remisbreite... " that Amateur was no beginner ...
koblentz? This is a very interesting discussion! I have a couple of additional comments and a question, Comment (1): Yes, chess ratings go below 1000, all the way down to a theoretical low of 100 USCF. There are actually a lot of players below 1000, most of them young "scholastic" players. The average of all USCF players is about 1450 (last time i checked). This is the average of players serious enough to buy books and pay entry fees in tournaments - the average strength of all people who have learned the rules and played casual games would be much lower than 1450. I'm using USCF instead of FIDE because FIDE has been used by professional and near-professional players only, not the much wider range of "club" players. I think the equivalent go average is around 6 to 9 kyu (again, just based on the USGA ranking list in 1983 - before Elo system was used in go) and the average of casual, non-tournament go players may be something like 20-25+ kyu. Comment(2): Since Kasparov is the top rated chess player in history at 2851 and studies have been done to normalize the ratings of past champions (adjusting for rating inflation) to levels slightly below this, it seems to me that you could peg the 9 dan prefessional go level to something like 2800-2850 FIDE, but not 2900 and certainly not 3000, which would make all comparisons meaningless.
Question: Is the 10 dan professional ranking an honorary title or does it mean that a 10p is actually a little bit stronger than a 9p? (I thought honorary, but not sure).
xela: Yes, it's honorary. See Judan.
Herman Hiddema: Hi Michael, good to see you're back. Here's my take on it:
It may be good to note that a 9p player may not neccessarily be stronger than an 8p player, it's a bit more tricky than that. I'll explain why:
- There are at least 5 organisations that award professional grades, see ProfessionalGoAssociations. Levels between these organisations vary.
- Promotion rules may not be quick enough (players that get stronger very fast may have a certain playing level, but may not yet have received the appropriate dan grade).
- Players are never (AFAIK) demoted. So players whose level declines (eg because of age) retain their grade.
As I mentioned earlier, Go has a wider range of distinguishable skill levels. In chess, the winning probability against someone 100 rating points stronger is constant (at 31%, see the Elo table). In go, it is varied (from 39% at 20kyu to 19% at 7d). This is done so that 100 rating points matches 1 kyu/dan grade as much as possible. As kyu/dan grades are not based on winning probabilities but on handicap stones required to even out the odds, the basic Elo system did not match.
The current Elo rating system used by the EGF starts with 20 kyu at Elo Rating 100. This is considered to be the lowest kyu grade that is regularly seen at tournaments. Weaker players than this do sometimes play in tournaments, but the EGF ignores such results (or considers such players as 20 kyu for rating purposes). Using this scale, top european players end up in the 2700+ range (currently 9 players, one of whom has a 2800+ rating). As some of the top european players have played against pro players, it is possible to extend the EGF rating to include pro players. This has been done, and the best rating ever achieved was 3035 (By Yi Ch'ang-ho)
Here is a list of top go ratings achieved: http://64.233.183.104/search?q=cache:m3Wc9dE_EqsJ:www.goweb.cz/progor/maxim.html+http://www.goweb.cz/progor/maxim.html&hl=en&ct=clnk&cd=1
Now compare this list: http://en.wikipedia.org/wiki/Comparing_top_chess_players_throughout_history
Suppose we try to match these lists? I'll substract 184 points from the ratings of the go players. That matches Yi Ch'ang-ho's 3035 with Kasparovs 2851. Both these players have quite a huge gap compared to the next player!
With this substraction of 184 points, there are 3 other Go players that have achieved a rating over 2800 (Kato Masao 2805, Cho Hun Hyeon 2803 and Cho Chikun 2803). For chess, there are also three such players (Topalov 2813, Kramnik 2809 and Anand 2803). Chess has 14 players with an 2750+ rating. In Go, there are 47 such players. Chess has 43 players at 2700+, Go has 111 (including the first westerner in the list, Michael Redmond, at 2701). It is not much of a surprise that the numbers in Go are larger, because the go list spans a larger time period. Of the 43 chess players with rating 2700+, only 5 have achieved that peak before 1994 and most have attained it after 2000. In the Go list on the other hand, you will find many top ratings from the 70's, 80' and 90's.
Reasons for this might be:
- Chess is a much younger sport, professionalization of the sport is much more recent.
- Chess levels have gained recently due to the support of computer programs in analyzing.
Oh, and one more point: You say that comparison becomes meaningless when go ratings extend to 3000, which is not true. Elo rating do not as such define any 'absolute' skill level. The only reason top player have ratings 2700-2800 is because the ratings were aimed at putting the average player at around 1500. If they had chosen to put the average player at 2000, Kasparov would have achieved 3351. Had they aimed at 1000, he would have scored 2351. These numbers are only meaningfull in relation to eachother. By definition they only tell you what the probabilities of winning/drawing/losing between two players are.
Koblentz? FYI, FIDE ratings are published for all players with ratings > 1400 (not 2200 as speculated in some comments below, evem my own). See http://www.fide.com/official/handbook.asp?level=B0209.
Proposal for comparing Go and Chess ratings and titles
Reading Koblentz's and Herman's comments it seems to me that they overestimate the go ranks. I suspect they forgot that the level of Go in the western world is much lower than it is in the Far East. I think that even though an EGF 6d may be something in the western world, it is much less impressive when the Far East is taken into account (I once read a story about a couple of EGF 6d players participating in the WAGC in Japan who were all challenged and beaten by the door man of the hotel where they stayed during the tournament. If I remember well, one of them was Mark Boon who wrote Goliath).
Velobici: In Japan, it is possible that a door man is an ex-insei...especially an ex-insei that sacrificed high school success in the pursuit of a career as a professional. In Asia, you never know what to expect given that there are so many former aspiring professionals compared to the number of professionals.
There are some problems when trying to match EGF ratings and Elo ratings:
- Elo winning percentages may not match with EGF percentages, because the EGF rating is not based on winning percentages alone, it is also based on handicap differences. Only between 10k and 3d a winning percentage of 64% between consecutive ranks seems to justify a 100 points difference between ranks. So the EGF scale is not a pure Elo scale.
For instance: The 30 EGF points difference between professional ranks is based on the rule that one handicap stone equals 100 EGF points. The difference between professional ranks is about 1/3 of a handicap stone, so the EGF difference is set to 30 points.
- I think the virtual absence of draws in go is the reason why there are so many go ranks, compared to chess. A Chess winning percentage of 64% (100 Elo) would include many draws. So in go a similar disparity in skill would lead to a considerably larger winning percentage (especially for strong go players, who can most probably secure a win from a small positional advantage, while strong chess players would most probably only get a draw from a small positional advantage). Differences in skill are not as convincing in chess as in go.
- I think the previous point causes the EGF range to be about 1.5 times as wide as the chess Elo range.
We can let the last two items (rating range versus winning percentage associated with 100 Elo points difference) cancel each other by compressing the EGF range of about 4500 to the chess Elo range of about 3000.
I tried to make a comprehensive list by
- matching the extremes of the EGF and Elo ratings
- comparing qualifications associated with Go ranks and Chess ranks
- considering the subranges that exist in the Go Ranks (kyu, dan, pin): the difference between a 3d and 4d is probably similar to the difference between 4p and 6p, and it is also bigger than the difference between a 10k and 11k, which itself is bigger than the difference between 27k and 28k. The difference in winning percentage between an 6d and an 7d is probably the largest difference existing between two consecutive Go ranks. Also, a 1p may be a bit weaker than a 7d. I think EGF 7d is closer to 2p.
I used the USCF classification mostly. FIDE ratings seems to have lower ratings in the top region ( http://math.bu.edu/people/mg/ratings/report94.ps). Another complication is that FIDE ratings and titles seem to suffer from inflation (between 1972 and 2005 the number of GM worldwide rose from 88 to 900) ( http://en.wikipedia.org/wiki/Grandmaster_%28chess%29), which seems like 100 Elo inflation in the last 15 years mostly. (the pin ranks also suffer from inflation, but to a lesser degree: about 2p-3p in the last 80 years?)
2019-02-11 Go | Chess --------------------------------------------------|------------------------------------------------------------------ Rank | EGF rating | USCF rating | FIDE rating | Rank -------------------------------------|------------|-------------|-------------|-------------------------------------- AlphaGo Zero (AI) | 3270 | | 3450 | (engine) Stockfish 8 AlphaGo Master (AI) | 3210 | | 3250 | (engine) Rybka 4 | 3180 | | 3150 | (engine) Rybka 3 | 3090 | | 3000 | (engine) Rybka 2 AlphaGo Lee (AI) | 3030 | | 2900 | classic meijin | 3000 | | 2850 | kasparov top title holder | 2970 | | 2800 | 9p | 2940 | | 2750 | super GM 7p | 2880 | | 2700 | 5p | 2820 | 2700 | 2650 | 3p | 2760 | 2600 | 2550 | GM 7d/1p | 2700 | 2500 | 2450 | IM 6d | 2600 | 2400 | 2350 | FM 5d | 2500 | 2300 | 2250 | CM 4d | 2400 | 2200 | | master 3d | 2300 | 2100 | | expert Mogo (program) 2d | 2200 | 2000 | | 1d | 2100 | 1900 | | class A 1k | 2000 | 1800 | | 2k | 1900 | 1700 | | class B 4k | 1700 | 1600 | | 6k | 1500 | 1500 | | class C 8k | 1300 | 1400 | | 10k | 1100 | 1300 | | class D 12k | 900 | 1200 | | 15k | 600 | 1100 | | class E 18k | 300 | 1000 | | 21k | 0 | 900 | | class F 24k | -300 | 800 | | 27k | -600 | 700 | | class G 30k | -900 | 600 | | 34k | -1300 | 500 | | class H 38k | -1700 | 400 | | 42k | -2100 | 300 | | class I, first game by adult? 46k | -2500 | 200 | | 50k | -2900 | 100 | | class J, first game by child?
Note: I tried to match each rank to the center of the corresponding rating range, not the lower bound.
So my current EGF rating may compare to a USCF master.
See RankWorldwideComparison for comparison between EGF ranks ans AGA ranks
This fit seems to support the following rough conversion factors between differences in Go ranks and Elo points:
- 2p ranks (60 EGF) = 100 Elo
- 1d rank (100 EGF) = 100 Elo
- SDK ranks (100 EGF) = about 50 Elo
- DDK ranks (100 EGF) = about 30 Elo
The fit could be smoothened, and molded a bit for higher accuracy, but i think the round numbers resulting from these simple conversion factors give a clearer picture.
xela: I find this interesting. In chess, the difference between "professonal" and "amateur" isn't as clearly defined as in go. I would have thought "International Master" would be equivalent to a very strong amateur, maybe European 6 dan. Are there any statistics available on what proportion of chess players are IMs versus what proportion of European go players are 6 dan? (Is that a meaningful question?)
petri Proportion on National Masters of all players who play seriously in tournaments is about 2%, and very very few of those get IM title. IM is certainly not good amateur but middle rank professional. As an example Finnish chess association has about 8000 players with rating and only 11 them have Elo high enough for IM title. And GO ranks are qute far away form each other. Elo difference of 4dan and 5dan is about 250 Elo points. Meaning that on amateur scale just about all chess titles go inside one dan rank. GOR is not Elo as it artificially makes one rank 100 points.
OscarBear?: Hello petri, can you expand a little on why there is a 250 Elo difference between 4-dan and 5-dan? (Maybe on a separate page) petri Thats rather simple (estimate is not accurate) Elo differencce of 200 points means that stronger player should get 75% of points in the long run. Now if you look the EGFRatingSystem page you will find that that probability of EGF 4d winning an opponent of 100 pts stronger is mere 23.6%. Which is less than 25 hence the Elo difference would be bigger than 200 pts. And on 6d it is already below 20% which translates to about 240 Elo pts
Also, at the low end of the table: I've never seen Elo ratings below 1000--does the scale really go all the way down to zero? I thought "raw beginner" level was around 600, although I can't remember where I got that number from. petri In theory there is no lower limit. In practice associations do not go below 100. Which is already "almost knows the rules" level
(It's a long time since I've taken chess seriously, so forgive me if those sound like silly questions. You're probably more up-to-date on this than I am.)
Davos: To me also, it seems as if the distinction between amateur and professional is clearer in go. To my understanding, class A and below are amateurs, while IM and above are professionals. Chess ranks in between are not clearly amateur of professional. Is this correct? The cause of this difference in distinction may be that the organisation of go professionals is national and for chess it is international. The Japanese pro organisation provides all its members with a salary, which the FIDE doesn't (I suppose). On the other hand, I think it won't be easy to support a family with the salary of a 2p professional in Japan. But in Japan 2p pros are usually females or young males (who can rely some other provider in the household). Note: there are go amateurs who can compete with professionals, sometimes as strong as a modern 9p: SekiyamaSendaiu, YasunagaHajime, Cho Seokbin, FernandoAguilar, JieLi. So some "very strong amateurs" are indeed as strong as IM when compared to chess, but they are rare. I don't think much can be concluded from the rating distribution of European go players. Analyzing it would be similar to analyzing the FIDE rating distribution of South Africa.
At the lower end the classification is unclear in chess as well as in go. An adult who completed a beginner's course in a real life Go club is usually considered 25k. The levels below 25k are almost exclusively used for kids (like the lower classes in chess). In the meantime I read that a Chess rating of 100 is considered to be the very lowest, even for children. Thats sounds as if 100 Elo is about 35k in Go.
I don't know if i am really up-to-date on this. I'm not a chess player, I just tried to use as much information as I could find about the chess classification of the USCF en FIDE. In the meantime I investigate further and I incorporate new information in the table as I go.
Notes:
- According to http://www.rvf.ru/engl/rvf.php?id=8&id1=expos&eid=420 there are at least 118 million chess players in the world (about 4 times the estimated numbers of go players) of which 88000 are professionals.
- According to http://www.fide.com/ratings/topfed.phtml there are 10000 players holding a FIDE title, of which 2700 are IM and 1000 are GM. This figure may be comparable to 1500 professional go players, which seems realistic to me. The bigger figure of 88000 professional chess players mentioned in the previous note probably includes playes who make a living from chess without holding a FIDE title. I guess there are a lot of strong non-pro go players too making a living from go.
- According to http://www.fide.com/ratings/top.phtml?list=men about 20 players would currently qualify for the (unofficial) super GM title (FIDE 2700+ rating). I arbitrarily selected the top title holders to be comparable, because this is also a very small group of the very strongest concurrent go players.
Davos: Recently I found some clues that seem to confirm your (xela) thoughts. First I thought that CM would be 1p, which would mean that an IM would be a fairly strong professional. but later i found other sources suggesting that IM is usually where professional chess begins. Also i've seen quite a few chess players stating ratings of 2250 casually, suggesting strength but clearly not considering themselves (nearly) professionals. So i suspect that a 2250 USCF rating has a prestige similar to a EGF 4d-5d rank. I updated the table to match this. I also dropped a few ranks at the top (like the kisei rank, because i think the difference between a classic mejin and a classic kisei is probably quite small after all).
What still confuses me is the naming of the USCF ranks and your (xela) thought about IM being a top amateur. Class A is usually mentioned as the top amateur class. This would suggest at least EGF 5d to me. But there seems to be a huge gap between Class A and IM, so how can they both be top amateurs? I now suspect that the top amateur qualification of Class A is a misnomer and that your thought is closer to the truth.
xela: I don't know much about how the US "class" system works. My thoughts were based on observations in Australia and the UK, that "being able to play in a serious tournament without losing every game" and "being able to earn a living as a full time chess player" are two very different things, and my impression was that IM status implies the first but probably not the second.
<HermanHiddema>: Hi Dave! The first link you give is probably unreliable (also given that it is a webpage seeking investments). The 88000 figure is phrased like this:
"'There are 88 415 professional chess-players, whose rating is calculated & registered by FIDE (01 01 2006 data) '"
Now unless I am very much mistaken, I think what they are claiming here is actually:
"'There are 88 415 players whose rating is calculated & registered by FIDE (01 01 2006 data). They are therefore professional players.'"
Also, the 118 million players total is doubtful. I have looked long and hard, but I have not been able to find a reliable number of total chess players worldwide. Figures like this usually count anyone who knows the rules of chess as a player.
Some more reliable figures:
- The USCF has 90,000 active tournament players ( http://www.libraryjournal.com/article/CA421026.html )
- All online chess servers together have at most 1,000,000 players ( http://en.wikipedia.org/wiki/Talk:Chess/Archive3#.22As_of_2005.2C_it_was_estimated_that_285_million_people_play_online..22 )
Another intersting figure: The IGF claims that there are 10,000,000 go players in China, 9,000,000 in Korea and 5,000,000 in Japan. And these are considered conservative estimates because they are used as background information for the attempt to get Go recognized as an olympic sport, so they do not want to be caught overstating their case. </HermanHiddema>
Davos: Hi Herman! You are right: the source for the 118 million (and 88000) estimate does not look very reliable, but it is the only estimate I found about the global size of chess. I indeed expect this number to represent the number of people worldwide who know the rules of chess (in the Netherlands this would be a few percent of the total population). But I guess the estimate of 30 million go players represents the same thing. I don't think there would be more than a few percent of the Japanese population who know the rules. I've lived in Japan for a year, and although everybody knew the game (and seem to respect you for being a fairly good player), most people did not know how to play.
I think both the 30 million and the 118 million figures are inherently unreliable. I think the margin of error could well be 30 or more% in both figures (the boundary of the both groups is vague too).
I too would not compare the 88000 figure to go professionals. I think it is as you say: the number of players with a FIDE rating. This figure probably includes large numbers of master level and possibly even expert level players (comparable to 6d and 5d go players according to my proposal).
<HermanHiddema:> I've found several sources, but sadly none of them are reliable.
As for the numbers of go players, those are conservative estimates. In Europe, there are, it seems, at least 20.000 active go players. This is the number of players on the EGF rating list, so it is the number of players appearing on tournaments in the last 12 years. Some of those may not be active anymore, but there are plenty of players active who do not go to tournaments.
10 million for China then does not seem so far fetched. It is only 50 times the number of players in Europe, which has only half the population, so Go need only be 25 times as popular.
For comparison, the number of Dutch go players in the ratinglist is 919. The number of Dutch chess players in their ratinglist is about 14000. Making Chess 15 times as popular as go, apparently. … </HermanHiddema> (more later …)
Anonymous: I would just note that the most talented chessplayers of all time (e.g., Fischer, Kasparov, etc.) took years of study and play to achieve master level (say, around 2200 FIDE). More typical of a reasonably talented child with exposure to strong players might be 4 to 6 years of study to reach a 2200 FIDE level. This might be one basis for comparing go and chess ranks.
However, my very rough impression as a beginner go player (around 10 kyu) and a more experienced chessplayer (around 2350 FIDE) is that chess strength is much more knowledge-based that go strength, which would explain why it takes even very talented chessplayers years of effort to reach a fairly modest level. It may be that go development is more even than chess development because there is not such a dramatic "knowledge threshold" that one needs to overcome.
My other comment is that, at least in the United States (I can't speak from personal knowledge about other countries) very few players of even IM strength (say, 2400+ FIDE and 2500+ USCF) can make a living from chess (in fact, a number of promising young GMs have essentially given up chess because of the bleak financial prospects). There are exceptions (e.g., relatively weak players that teach chess in schools, etc.) but probably not that many. I would estimate that a chessplayer would need to be at least 2500 to 2550 FIDE to have a reasonable chance of making a living playing chess, and even then most of their income would probably come from giving lessons and it would be a very modest income. I'm fairly confident that the vast majority of players below 2400 FIDE are purely amateurs. So if you defined professional as someone who might make a very modest living from the game, maybe 2500 to 2550 FIDE would represent a kind of baseline for a professional player.
Herman: Professional in the go sense is slightly different from professional in the general sense. A Go professional is someone who has passed a pro-exam or has been promoted by special recommendation, but that does not necessarily mean that they can make a living from it. Although many professionals do receive a baseline salary from their pro association, there is some movement away from this, and in Korea they have recently made some big changes to the payment structure, which means that many weaker professionals will probably no longer be able to make a living in go. This probably means that many of the 1p-3p players will have difficulty with this (equivalent to FIDE 2450–2550 according to the above table), which would support your argument that players need to be at least (equivalent to) 2550 or so to make a modest living from it.
Jono64a: Elo ratings are named after the Hungarian-American physicist and chess master and Wisconsin state champion Arpad Emmerich Elo (né Élő Árpád Imre 25 Aug 1903 – 5 Nov 1992). Thus Elo is not an acronym or initialism, so should not be written in all-capitals.