It is currently Tue Mar 19, 2024 7:04 am

All times are UTC - 8 hours [ DST ]




Post new topic Reply to topic  [ 11 posts ] 
Author Message
Offline
 Post subject: Fan Hui
Post #1 Posted: Sun May 15, 2016 8:49 am 
Dies with sente

Posts: 96
Liked others: 0
Was liked: 14
Not sure I've seen it mentioned in any of the posts about AlphaGo, post-series, but Fan Hui seems to have improved a lot and has put it down to AlphaGo..?

"Since his loss, Fan has been training against AlphaGo in DeepMind’s offices – and his world ranking has jumped from 600 to 300 in the last three months alone. ‘He just won the European pro Go championship in February, and he won with a full score; he beat absolutely everybody,’ Hassabis says, with a trace of pride. Fan is on the best form he’s ever been, and he attributes it to AlphaGo."

Cited from http://s.telegraph.co.uk/graphics/projects/go-google-computer-game/

Top
 Profile  
 
Offline
 Post subject: Re: Fan Hui
Post #2 Posted: Sun May 15, 2016 9:07 am 
Gosei

Posts: 1589
Liked others: 886
Was liked: 527
Rank: AGA 3k Fox 3d
GD Posts: 61
KGS: dfan
According to http://www.goratings.org/players/1480.html, he then lost his next 6 games, although it was against pretty strong opposition.

Top
 Profile  
 
Offline
 Post subject: Re: Fan Hui
Post #3 Posted: Sun May 15, 2016 10:32 am 
Judan

Posts: 6725
Location: Cambridge, UK
Liked others: 436
Was liked: 3719
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
I don't see any significant evidence in his results that Fan Hui has got stronger: he beat Shikshin, Sumra, Jabarin, and Dinerstein to win that European Pro Championship: all decent wins but he was the significant favourite in all those games and has been beating that crowd of top Europeans to win the last 3 European Go Congresses before AlphaGo. Fan typically only loses to visiting Asian pros/top amateurs, plus the occasional (maybe 1 in 5 to 1 in 10) loss to top Europeans; for example I recall him losing to Ilya in the Polish EGC weekend tournament (but he won their more important game in the main tournament). He then lost against top active Asian pros at the IMSA Mind Games and Ing cup as expected. He even lost to Mingju Jiang (the Chinese 7p living in America for ages) which is a game he could win on a good day. As for world 600 to 300, that is nonsense and a meaningless blip in goratings which has since been corrected (he's 521 now, Mingjiu is 702 FWIW). Now maybe he has got a bit stronger and he didn't get hammered as much as he would previously by top pros like Mi Yuting, but the problem with the sort of international tournaments he plays is he is up against top pros of Korea/China/Japan/Taiwan and he'd need to have gotten a lot stronger to beat them, whereas maybe now he could beat some anonymous low to mid pro who he would have lost to before, but those pros don't go these events. Ali Jabarin and Andy Liu did exactly that at those Japanese Sankei prelims and exchange matches.

Top
 Profile  
 
Offline
 Post subject: Re: Fan Hui
Post #4 Posted: Sun May 15, 2016 11:13 am 
Tengen

Posts: 4380
Location: North Carolina
Liked others: 499
Was liked: 733
Rank: AGA 3k
GD Posts: 65
OGS: Hyperpape 4k
I agree with everything Uberdude said.

I wouldn't expect go ratings to be even close to reliable until a player has:
1) at least a dozen games against Asian professionals (probably more)
2a) a few wins against Asian pros, or
2b) many losses against lower ranked pros.

The rationale behind 2b is that getting repeatedly clobbered by really strong players doesn't tell you enough. Against that level of competition, you might get end up with zero wins whether you're an average pro or a kyu player like me.

A good example is a Japanese youngster named Onishi Ryuhei. I've been watching him since he's either 15 or 16 and already playing pretty well. He has 26 games in the database, a number of wins, but he still lost 70 rating points overnight around the time of the Globis cup.

_________________
Occupy Babel!

Top
 Profile  
 
Offline
 Post subject: Re: Fan Hui
Post #5 Posted: Sun May 15, 2016 11:52 am 
Judan

Posts: 6725
Location: Cambridge, UK
Liked others: 436
Was liked: 3719
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
As an example of the unreliability of goratings with few games (and disjoint populations), I was amused to see it ranked Thomas Debarre, the 6d amateur from France above Rin Kaiho 9p from Japan and winner of numerous titles and I think at one point the most pro career wins. For sure Rin is getting old and weaker, but really 6d ama weaker?!

Top
 Profile  
 
Offline
 Post subject: Re: Fan Hui
Post #6 Posted: Sun May 22, 2016 2:56 am 
Dies with sente

Posts: 80
Liked others: 0
Was liked: 19
Uberdude wrote:
As an example of the unreliability of goratings with few games (and disjoint populations), I was amused to see it ranked Thomas Debarre, the 6d amateur from France above Rin Kaiho 9p from Japan and winner of numerous titles and I think at one point the most pro career wins. For sure Rin is getting old and weaker, but really 6d ama weaker?!


Goratings once had Iyama at world number two! That alone should dissuade anyone who has any common sense to take Goratings seriously. The man behind the ranking is a great programmer, but he's no statistician.

Top
 Profile  
 
Offline
 Post subject: Re: Fan Hui
Post #7 Posted: Sun May 22, 2016 6:31 am 
Lives in sente

Posts: 827
Location: UK
Liked others: 568
Was liked: 84
Rank: OGS 9kyu
Universal go server handle: WindnWater, Elom
by78 wrote:
Uberdude wrote:
As an example of the unreliability of goratings with few games (and disjoint populations), I was amused to see it ranked Thomas Debarre, the 6d amateur from France above Rin Kaiho 9p from Japan and winner of numerous titles and I think at one point the most pro career wins. For sure Rin is getting old and weaker, but really 6d ama weaker?!


Goratings once had Iyama at world number two! That alone should dissuade anyone who has any common sense to take Goratings seriously. The man behind the ranking is a great programmer, but he's no statistician.


Let's mop the path of our bowls, first :).

One problem at least appears to be that the only game results goratings.org has are those in which macelee, on go4go, can find game records of. This means that most of the games included into the database are major title matches, domestic main tournament matches and main draw international matches.

This leaves out many smaller preliminary matches which are (un)surprisingly important in determining relative strength between player's lower down in the rating list.

I doubt this is just something specific to goratings. Dr Bea Teal might have access to these preliminary results at least in Korea, but despite this, one of the main reasons why only the top 100 players are listed, is probably to do with them having the most reliable ratings. Compare them to the top 100 on goratings.org's list, like for like, as we don't know who the 300th rated person is on Dr Teal's list: it might be more surprising.

Despite this, goratings (and mamumamu0413, possibly) ranked Ke Jie as number one before Dr Teal's list. It's an impressive database and rating algorithm, not to be classed as ineffective, I'm sure... ...


e.g. It's hard to find game record's of all but the last (or possibly second-to last) round of the preliminaries of even international-scale tournaments, so players who passed through the recent LG Cup preliminaries, for example, would not have as large a rating increase as they should in goratings, but maybe in other rating lists, causing a meaningful disconnect between the different algorithms concerning average-strength/median pro's and up.

However, this most likely has little effect upon the top 100, as the competition a top 100 player would have to route would probably be statistically much weaker on average, and many of the top 20~30-ranked competitors playing in the competition may be seeded into the main draw, in any case.

_________________
On Go proverbs:
"A fine Gotation is a diamond in the hand of a dan of wit and a pebble in the hand of a kyu" —Joseph Raux misquoted.

Top
 Profile  
 
Offline
 Post subject: Re: Fan Hui
Post #8 Posted: Sun May 22, 2016 6:40 am 
Gosei

Posts: 1589
Liked others: 886
Was liked: 527
Rank: AGA 3k Fox 3d
GD Posts: 61
KGS: dfan
by78 wrote:
Goratings once had Iyama at world number two! That alone should dissuade anyone who has any common sense to take Goratings seriously. The man behind the ranking is a great programmer, but he's no statistician.

Coulom's paper on the Whole History Rating system used by goratings.org is here (PDF link), if any statisticians want to comment.

Top
 Profile  
 
Offline
 Post subject: Re: Fan Hui
Post #9 Posted: Sun May 22, 2016 7:12 am 
Judan

Posts: 6725
Location: Cambridge, UK
Liked others: 436
Was liked: 3719
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
I don't think WHR is so bad, it's more a problem of garbage in and garbage out. Highly disjoint player populations are always going to be hard to compare, particularly when there is a player in one head-and-shoulders above all his compatriots. And quite possibly Bae Taeil's system has more tweaking and weights applied to matches that reflects our human knowledge RE Japan. I think Remi puts all new pros into the system with the same initial rating: if he used a lower one for Japanese pros than Korean/Chinese (to reflect that they are generally weaker) then the results could be better. But even so I don't agree with all the goratings hating that seems to happen, people seem to think it has some official/authoritative standing and claims to be more than it is.

Top
 Profile  
 
Offline
 Post subject: Re: Fan Hui
Post #10 Posted: Sun May 22, 2016 10:59 am 
Dies in gote

Posts: 24
Liked others: 15
Was liked: 7
Rank: EGF 2d
KGS: fightback
I agree with Uberdude about WHR and goratings.

First, because of the database used, this ranking focuses on the top 200, perhaps 300 players, the rest is just here to give an idea.

Secondly, WHR use all the games available to calculate a rank at a given date, including the more recent ones. So the ranking that's displayed today is not the final one, because futur games will change it. The Iyama's case is a good exemple. Iyama reach his top the 5th of november 2015. This day, he was only number four, behind Ke, Park and Lee, and less than a point in front of Shi (I don't include Alphago). So WRH never really had Iyama at world number 2, it was just waiting for more games.

In my opinion, 4th for Iyama wasn't wrong. Or more exactly, based only on his results, assuming I don't have any other knowledge about go (like an anlgorithm), I don't see anything wrong with that. He won 32 games in a row, and even if none was agaisnt a top pro, there's 17 of them against a 3300+ opponent (3300 give you a rank of 107 today). Now assuming having a 90% chance of winning a game is being much better than an opponent, the probability of winning 17of such games in a row is 16,7%. So for 6 months, Iyama was so much above an average top 100 player than being number 4 is not surprising. I'm not saying that he totally deserved his rank, but given the results available, I think the algorithm do a really good job overall.


This post by findol was liked by: hyperpape
Top
 Profile  
 
Offline
 Post subject: Re: Fan Hui
Post #11 Posted: Fri Jun 24, 2016 8:33 pm 
Beginner

Posts: 11
Liked others: 0
Was liked: 0
Thanks for the info
royal1688 download

Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 11 posts ] 

All times are UTC - 8 hours [ DST ]


Who is online

Users browsing this forum: No registered users and 1 guest


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