It is currently Sun Nov 19, 2017 6:47 pm

All times are UTC - 8 hours [ DST ]




Post new topic Reply to topic  [ 68 posts ]  Go to page 1, 2, 3, 4  Next
Author Message
Offline
 Post subject: Iyama's world ranking
Post #1 Posted: Fri Nov 04, 2016 2:44 am 
Tengen

Posts: 4235
Location: Cambridge, UK
Liked others: 159
Was liked: 2083
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
This thread is for discussion/whinging/ranting on Iyama's world ranking on goratings.org and comparing his strength to Chinese/Korean pros. by78 please feel welcome to tell us how weak he is after ever loss here.


This post by Uberdude was liked by 3 people: dfan, ez4u, kimidori
Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #2 Posted: Fri Nov 04, 2016 3:15 am 
Tengen

Posts: 4235
Location: Cambridge, UK
Liked others: 159
Was liked: 2083
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
by78 wrote:
Uberdude wrote:
by78 wrote:
Uberdude said: "Just like there's no way Lee Sedol is in the top 10 because he lost to Ichiriki Ryo?"

Did Lee lose to Ichiriki in a seven game match?

No. But top Korean and Chinese pros have never beaten Takao in a seven game match with 8 hours each either. So whilst they are considerable favourites in the shorter international matches, I could well believe Takao might do somewhat better against them in such a format. It's worth pointing out that Gu Li was reluctant to the 4 hour time limits of his jubango against Lee Sedol because he saw it as a disadvantage.



Ahhh, the time limit excuse again...

I really do miss the 1980s and early 90s, before the time limit excuse was invented and when the Japanese super-9-Dans (Cho, Kobayashi, Takemiya, Kato, et al.) were sweeping all before them internationally.

Or are my memories faulty? Did Japan have shorter time controls 25-30 years ago?


I of course agree that in the 80s / early 90s Japan was a far stronger force in international Go (they weren't sweeping Nie Weiping though). And as I have said before I actually agree with you that Iyama is probably not in the world top 10 now. What I find tiresome is the repetitiveness, glee and certainty with which you say he is over-ranked.

As for the time controls, I don't think back in the 80s the top Japanese pros beat Chinese/Korean ones in 8 hour games. What were the time limits of the SuperGo matches? I presume the early international tournaments won by Japanese players (Fujitsu, Ing, Asian TV etc) had similar time limits to they do today. What I am saying is that 8 hour games are different to 2 or 3 hour ones, and I am open minded enough to believe that Takao could well play comparatively better in the former. Chinese/Korean pros have no experience of such games, I think it would be great if they did but it seems it's not a format that attracts domestic sponsors. As evidence that foreign pros might not like such long games I gave the example of Gu Li not wanting long time limits against Lee Sedol. Lee Sedol was also apparently reluctant to have long time limits against AlphaGo because he was worried he would fatgiue whereas the computer would not. So it seems quite plausible to me that Takao's experience of many 2-day games means he is relatively good at them. (Less relevant, but another example of time limits making a difference: Dinerstein 3p could probably beat me on 2 or 3 stones in a 1 hour game, but I beat him even in a 24-hour game). So whilst Takao might only have something like a 10% win chance against some top Chinese/Korean pro in a 2 hour game, in a 2-day game it might be higher. I'm not saying >50%, but maybe 30%? We really have very little evidence to say anything with confidence.


Last edited by Uberdude on Fri Nov 04, 2016 3:28 am, edited 1 time in total.
Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #3 Posted: Fri Nov 04, 2016 3:16 am 
Tengen

Posts: 4235
Location: Cambridge, UK
Liked others: 159
Was liked: 2083
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
Back in the other Iyama thread we had:

Uberdude wrote:
by78 wrote:
handa711 wrote:
Takao Shinji is leading 3-0 in Meijin.


If Takao wins the Meijin, expect Goratings.org to rank him among top 10 in the world. LOL!


No. You have a poor understanding of the Whole History Rating algorithm if you expect that. For one thing winning 4 games won't give him the 150 points or so he needs to get to top 10, that would need far more games. Secondly he already has 3 of the 4 wins he needs in the goratings system and is still down at #52. Cue whinging that even 52 is too high for Takao.


Takao duly won the Meijin title, but rather than being in the top 10 on goratings.org as by78 predicted he is at #57 today (go4go has added the final game).

Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #4 Posted: Fri Nov 04, 2016 3:26 am 
Gosei

Posts: 1901
Liked others: 12
Was liked: 2814
I'm not saying national pride doesn't matter, but all this sniping at Japanese players always brings up in my mind the old joke about Nietzsche writing "God is dead" and then God's finger appears and writes "Nietzsche is dead." In this case, guess which one is Iyama (or Takao) sitting looking at his bank account.

A propos not much, the world chess championship is about to kick off between the No. 1 and No. 10 players on the Elo list.

Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #5 Posted: Fri Nov 04, 2016 6:46 am 
Dies with sente

Posts: 112
Liked others: 1
Was liked: 11
Uberdude wrote:
What I am saying is that 8 hour games are different to 2 or 3 hour ones, and I am open minded enough to believe that Takao could well play comparatively better in the former. Chinese/Korean pros have no experience of such games, I think it would be great if they did but it seems it's not a format that attracts domestic sponsors. As evidence that foreign pros might not like such long games I gave the example of Gu Li not wanting long time limits against Lee Sedol. Lee Sedol was also apparently reluctant to have long time limits against AlphaGo because he was worried he would fatgiue whereas the computer would not. So it seems quite plausible to me that Takao's experience of many 2-day games means he is relatively good at them. (Less relevant, but another example of time limits making a difference: Dinerstein 3p could probably beat me on 2 or 3 stones in a 1 hour game, but I beat him even in a 24-hour game). So whilst Takao might only have something like a 10% win chance against some top Chinese/Korean pro in a 2 hour game, in a 2-day game it might be higher. I'm not saying >50%, but maybe 30%? We really have very little evidence to say anything with confidence.


While they might not have experience with 8 hour tournaments, it's not like they don't have experience with 8 hour games. I remember watching weiqitv, and Peng Quan commenting on how some of the players tried to play super long games for fun (I can't remember who it was, but it's the guy the nickname "cow demon king" from the journey to the west because he always makes sure to use up all his time in games).

And the general consensus from chinese pros, I think, is that after a certain point, having more time doesn't make your play better. In fact, it might make it worse, since you start second guessing yourself and stuff. I think they would like to see more 4 hour games, but not more than that.

Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #6 Posted: Fri Nov 04, 2016 6:55 am 
Tengen

Posts: 4235
Location: Cambridge, UK
Liked others: 159
Was liked: 2083
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
idontgetit wrote:
And the general consensus from chinese pros, I think, is that after a certain point, having more time doesn't make your play better. In fact, it might make it worse, since you start second guessing yourself and stuff. I think they would like to see more 4 hour games, but not more than that.

Maybe for modern ones with little experience of it, but do you really think Go Seigen (who was a relatively fast player), Kitani, Fujisawa Hosai (slow) and the greats of old played worse when they had 12 hours for big matches 60+ years ago than they would with a few hours? If they thought they could play better faster they could have used less time, but they didn't.

Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #7 Posted: Fri Nov 04, 2016 7:19 am 
Tengen

Posts: 4044
Location: North Carolina
Liked others: 411
Was liked: 662
Rank: AGA 3k
GD Posts: 65
OGS: Hyperpape 4k
I think the question of whether Iyama's rating on goratings boils down to three subquestions:

1) Are the other Japanese players overrated? If so, Iyama will be overrated, because he'll receive too much credit for beating them.
2) Could it be that Iyama just performs better against Japanese opponents than international competition? If so, then it's a bit harder to say what his rating "should be", but you can at least say it's higher than a hypothetical scenario where he plays more games against non-Japanese opponents (i.e. he's overrated).
3) Is WHR fundamentally flawed? The algorithm only works if you can estimate a player's performance against 3500 ELO players by looking at their performance against 3300 or 3200 players. Perhaps that estimate is systematically skewed.

Each of these three seems possible. You could do statistical tests for (1), but as far as I know, no one has tried to test it. For (2), there's not enough data to be really confident. Iyama's results against other top 20 players are somewhat worse than you'd expect, but I'm not sure they are statically significant (there's a thread where Remi does the calculation, but I've lost it).

(3) is much more difficult to test, but it's also the most speculative.

Quote:
And as I have said before I actually agree with you that Iyama is probably not in the world top 10 now. What I find tiresome is the repetitiveness, glee and certainty with which you say he is over-ranked.
Could not agree with the second sentence more. I take it to be an open question whether goratings is accurate, and I think it's reasonable for people to say they feel like it's over-estimating Japanese players. What I find annoying is the variety of dumb "proofs" people give that it must be inaccurate. Related: elsewhere, I said that Dr. Bae Taeil's system was misjudging Japanese players because of the wild swings in their ratings. I still am suspicious of them, but I shouldn't have acted like I knew it was wrong.

_________________
Occupy Babel!


This post by hyperpape was liked by: dfan
Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #8 Posted: Fri Nov 04, 2016 7:47 am 
Judan

Posts: 7584
Liked others: 1377
Was liked: 1161
KGS: Kirby
Tygem: 커비라고해
These rating systems are just models. It's expected that they are not 100% accurate, and would be surprising if they were.

Various presidential polls predict different margins for the US presidential elections these days. You can criticize the data used or model of a particular poll, but if you care that much, make your own poll.

If someone else's prediction of truth is not accurate, it's not the end of the world.

_________________
What have I learned from this game that I'll be able to use in my own game if the opportunity presents itself?


This post by Kirby was liked by: Waylon
Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #9 Posted: Fri Nov 04, 2016 10:01 am 
Dies with sente

Posts: 112
Liked others: 1
Was liked: 11
Uberdude wrote:
idontgetit wrote:
And the general consensus from chinese pros, I think, is that after a certain point, having more time doesn't make your play better. In fact, it might make it worse, since you start second guessing yourself and stuff. I think they would like to see more 4 hour games, but not more than that.

Maybe for modern ones with little experience of it, but do you really think Go Seigen (who was a relatively fast player), Kitani, Fujisawa Hosai (slow) and the greats of old played worse when they had 12 hours for big matches 60+ years ago than they would with a few hours? If they thought they could play better faster they could have used less time, but they didn't.

I don't know, but I wouldn't be surprised if the quality of their games were not better than 4 hour games*.
Just because people (even the very best people) do something a certain way doesn't mean it's the best way to do it. There have been lots of misconceptions in every field by the top people of their time.

Logically they'd try to think as long as possible if they thought that it worked better. But if they actually did experiment, maybe they'd realize differently.


*If they were not allowed to intensely study the game with all of their students/friends when they stop the match. Like, if you're allowed to play sequences on the board during breaks and study with a bunch of people, that's not really a show of your own skill. IMO that's cheating.

Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #10 Posted: Fri Nov 04, 2016 10:19 am 
Tengen

Posts: 4235
Location: Cambridge, UK
Liked others: 159
Was liked: 2083
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
idontgetit wrote:
I don't know, but I wouldn't be surprised if the quality of their games were not better than 4 hour games*.

I actually started a thread trying to address that question by looking at mistakes in Gu vs Lee jubango games vs Go vs Kitani Kamakura matches vs a modern faster one, but as tends to happen got busy, didn't finish and the thread went off on tangents. If you or someone else would like to try to quantify the relative amount of mistakes or some other measure of quality in games of different time limits then I'd be very interested. Wouldn't it be great if we could use AlphaGo to analyse the games and give some kind of average delta from its best move (like CrazyStone can but it is far weaker than pros so useless for this) as the approach I used there has the obvious defect that what I, via my interpretation of the professional commentators, classify as a mistake, could itself by mistaken (as could AlphaGo, but less so as it's very strong and it's more impartial, though by being trained on modern Go might have a preference for that).
viewtopic.php?f=13&t=13207


This post by Uberdude was liked by: Bantari
Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #11 Posted: Fri Nov 04, 2016 10:48 am 
Gosei

Posts: 1901
Liked others: 12
Was liked: 2814
Quote:
I don't know, but I wouldn't be surprised if the quality of their games were not better than 4 hour games*.


As is well known, I am not a fan of Mickey Mouse time limits. But my stance has always been that taking away time harms the quality of go, yet I have never automatically assumed that giving extra time is always beneficial. We have some anecdotal evidence both ways on that.

On the side that extra time is a good thing, we can point to the fact that players in two-day games almost invariably use the evenings or even the overnight period to think about the games more. They must think it's worth the extra tiredness it incurs. Added to that, pros always review their games and usually find better moves somewhere - that, too, may suggest if they had more time during the game they might have played better.

But against that, tiredness must be a factor. Not necessarily in a current game, but cumulatively over a period, leading more to sickness rather than bad moves, although there are known instances of tired bad moves. It was concern over health that led to the Oteai being cut from 13 hours each. But in those days players might only play one serious game a week, so a cut to 10 hours might have been sufficient to restore the health balance. Since then, the number of tournaments and games has grown spectacularly, and so it has been sensible to cut time limits for health reasons alone.

I don't think there has ever been a serious view that extra time is wasted. It's just a matter of compromise with other factors such as health, scheduling and sponsors' interests.

FWIW I detect increasing unrest in the pro world over ridiculously short time limits (and know there was unrest before). They will never go away but, I suspect, will increasingly be seen as either PR or exhibition-type games or team efforts rather than individual efforts (i.e. where the objective is to score points for a team rather than combine a win with a masterpiece purely for one's own reputation). China now seems to be hovering between 2, 3 and 4 hours each. Japan seems to favour 5 hours for top games, otherwise 3 mainly. Korea still has a bigger burden of short games and will take longer to rein them in, but the will to change seems there.

I can't imagine we'd ever go back above 5 hours in ordinary tournament play. Personally I'd respect 3 or 4-hour play, but am not yet sure about 2 hours. 1 hour is still too fast for me as a fan.

Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #12 Posted: Fri Nov 04, 2016 12:41 pm 
Tengen

Posts: 4044
Location: North Carolina
Liked others: 411
Was liked: 662
Rank: AGA 3k
GD Posts: 65
OGS: Hyperpape 4k
Kirby wrote:
These rating systems are just models. It's expected that they are not 100% accurate, and would be surprising if they were.

Do you really think this is a constructive addition to the discussion? Forgive me for being a little snippy, but I think it's borderline condescending.
Kirby wrote:
If someone else's prediction of truth is not accurate, it's not the end of the world.
This is true. It's not worth getting upset about, even if it seems like by78 is out to aggravate everyone more than anything else.

That said, some of us find it an interesting topic, and we're gonna talk about it.

_________________
Occupy Babel!

Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #13 Posted: Fri Nov 04, 2016 1:03 pm 
Judan

Posts: 7584
Liked others: 1377
Was liked: 1161
KGS: Kirby
Tygem: 커비라고해
I'm not trying to stop you from talking about it. I just don't understand the apparent agitation.

For example, if by78 doesn't think goratings is sufficiently accurate, that's OK. I'd recommend for him to propose a better ranking. Having data to show its performance would be even more convincing.

Not trying to be condescending, but this seems like common sense to me.

_________________
What have I learned from this game that I'll be able to use in my own game if the opportunity presents itself?

Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #14 Posted: Sat Nov 05, 2016 2:48 am 
Dies with sente

Posts: 112
Liked others: 1
Was liked: 11
Uberdude wrote:
idontgetit wrote:
I don't know, but I wouldn't be surprised if the quality of their games were not better than 4 hour games*.

I actually started a thread trying to address that question by looking at mistakes in Gu vs Lee jubango games vs Go vs Kitani Kamakura matches vs a modern faster one, but as tends to happen got busy, didn't finish and the thread went off on tangents. If you or someone else would like to try to quantify the relative amount of mistakes or some other measure of quality in games of different time limits then I'd be very interested. Wouldn't it be great if we could use AlphaGo to analyse the games and give some kind of average delta from its best move (like CrazyStone can but it is far weaker than pros so useless for this) as the approach I used there has the obvious defect that what I, via my interpretation of the professional commentators, classify as a mistake, could itself by mistaken (as could AlphaGo, but less so as it's very strong and it's more impartial, though by being trained on modern Go might have a preference for that).
http://www.lifein19x19.com/forum/viewto ... 13&t=13207


I think that's extremely hard to do, because you'd also have to take into consideration the strength of the players. For example, if Wu Qingyuan games were commentated by a 1930s Japanese pro, that pro is almost certainly much weaker than current top pros, so the amount of things he could see is much less than what Ke Jie or Park Junghwan would be able to see. But then, Ke Jie and Park Junghwan don't give public reviews of old games.

If we had a program like AlphaGo who can more objectively evaluate moves, that would work. Do you think that the 20 hour (or however long) games of the past were as high a quality as the 2-3 hour international tournaments today, uberdude?

Personally, I think it won't be. But then I'm even quite a bit weaker than you, so I wouldn't be able to tell at all. (Note that this question is just an aside, I'm not using it to prove that longer games don't help)

Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #15 Posted: Sat Nov 05, 2016 3:04 am 
Lives in gote

Posts: 306
Liked others: 56
Was liked: 96
Rank: igs 3d+
For what it's worth (from Go World 102, p. 42):
Quote:
Readers who keenly follow the Japanese go scene will be disappointed to see that we have only one game from the Kisei title match in this issue. Go World has always maintained a policy of presenting all the games from the best-of-seven title matches in Japan, that is, the Kisei, Meijin and Honinbo. One reason is that the two-day format seems to produce games richer in content and drama than the one-day games of the other titles. A second reason is that these matches are treated as the highlights of the tournament year in the go press, so good commentaries are usually available.

To my mind this raises two comments:
- "content and drama" is not exactly the same thing as "quality" (if by "quality" we mean "absence of errors"
- as the editors mentioned, the fact that these games are treated as the highlights of the year means better (longer? more thorough?) commentaries. It cannot be ruled out that, if as much time were spent on commenting the games from one-day tournaments, maybe they would be found to be as rich in content and drama.

Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #16 Posted: Sat Nov 05, 2016 3:54 am 
Lives in gote

Posts: 436
Liked others: 1
Was liked: 38
Rank: KGS 5 kyu
I think the problem with WHR is that Japanese pros don't play that much against Chinese or Korean pros. So you actually have 3 groups of players, C, K, and J but C-K also has bridges between them. The number of bridges from J to C/K is small so it is hard to estimate how strong J players are.

Imagine if only one japanese player ever played international titles. So out of hundreds of chinese/korean and japanese only one player linked them. Then whole rating system would depend on that player alone. Japanese players should really play more international games.

Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #17 Posted: Sat Nov 05, 2016 8:28 am 
Lives in sente

Posts: 1208
Liked others: 258
Was liked: 322
Rank: 5d
GD Posts: 1000
From my position as a watcher of pro games I think a big factor in the shorter time limits is the watchability, in real time, of those games. This must definitely influence sponsors since their commercials would get greater exposure. Personally I've enjoyed watching real time long time limit games on the internet. I try to find the next move and what the players might be thinking about. As for errors, even with unlimited time games people will make
errors because no one, not even alphago, plays perfect go. My impression is that the short time limits push players to make gambles, playing risky moves, and hoping the opponent will make a mistake. Of course, people make mistakes in slow games too. But I want players to win by outplaying their opponents, not relying on chance. I admire the attitude of Otake Hideo 9p who resigned a game when he was dissatisfied with his play in a game in which he was ahead. Another admirable thing is how a two stone game between Dosaku (w) and Yasui Shunchi in 1683 considered Dosaku's masterpiece even though Dosaku lost by one point. There is so much more in go than winning.


This post by gowan was liked by: Waylon
Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #18 Posted: Sat Nov 05, 2016 9:02 am 
Dies with sente

Posts: 112
Liked others: 1
Was liked: 11
gowan wrote:
From my position as a watcher of pro games I think a big factor in the shorter time limits is the watchability, in real time, of those games. This must definitely influence sponsors since their commercials would get greater exposure. Personally I've enjoyed watching real time long time limit games on the internet. I try to find the next move and what the players might be thinking about. As for errors, even with unlimited time games people will make
errors because no one, not even alphago, plays perfect go. My impression is that the short time limits push players to make gambles, playing risky moves, and hoping the opponent will make a mistake. Of course, people make mistakes in slow games too. But I want players to win by outplaying their opponents, not relying on chance. I admire the attitude of Otake Hideo 9p who resigned a game when he was dissatisfied with his play in a game in which he was ahead. Another admirable thing is how a two stone game between Dosaku (w) and Yasui Shunchi in 1683 considered Dosaku's masterpiece even though Dosaku lost by one point. There is so much more in go than winning.


Actually, pros tend to make less "gambles" in faster games. The thing is, pros usually don't like to completely lose after a single, huge battle. They would rather pick a way that would be less risky, and with less time, many more moves become risky (because with more time, they can read it out and realize that's it's actually not risky).

Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #19 Posted: Sat Nov 05, 2016 2:32 pm 
Lives with ko

Posts: 162
Liked others: 12
Was liked: 180
Rank: UK 2d Dec15
KGS: mathmo 4d
IGS: mathmo 4d
goratings seems to be just based on win and loss of games from go4go data?

If so, that seems pretty objective.

I think I understand right that other systems often place extra emphasis on winning/doing well in world title competitions (other sports often do this I think). And that would hit Iyama's rank for not turning up.

Top
 Profile  
 
Offline
 Post subject: Re: Iyama's world ranking
Post #20 Posted: Sat Nov 05, 2016 2:51 pm 
Tengen

Posts: 4044
Location: North Carolina
Liked others: 411
Was liked: 662
Rank: AGA 3k
GD Posts: 65
OGS: Hyperpape 4k
dhu163 wrote:
If so, that seems pretty objective.
It is objective in the sense that it's not driven by bias, but it may not be accurate. I believe it's using something like a prior for each player, and so if there are too few games between Japanese and non-Japanese players, the resulting rating for Japanese players will be too influenced by the prior.

_________________
Occupy Babel!

Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 68 posts ]  Go to page 1, 2, 3, 4  Next

All times are UTC - 8 hours [ DST ]


Who is online

Users browsing this forum: No registered users and 4 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group