Accuracy of recent kifus

macelee · Post by **macelee** » Sun Jul 21, 2013 11:20 pm

weiqi.tom.com is a major Chinese source of recent professional game records. Unfortunately, a senior editor has left his post recently and the temporary staff replacing him is much less experienced. For this reason, a lot of game records published there contain very significant errors - incorrect move sequence, incorrect date, komi, etc. What's more worrying is that some other 'collections' simply blindly copy/paste things from weiqi.tom.com without checking the details. Here is just one absurd example with a game being 625 moves long :evil:

http://gokifu.com/s/1f6m-gokifu-2013071 ... 5p%29.html

I tried my best to help correct the source. Make sure to view the recent games at Go4Go.net where games are properly checked before going in the database.

Mef · Post by **Mef** » Sun Jul 21, 2013 11:28 pm

I'm sure TMark might have a few words to add about the accuracy of game record reporting...

xed_over · Post by **xed_over** » Sun Jul 21, 2013 11:33 pm

Game recording is hard. I only do top level amateur games, but they are still above my level, and I have to pay very close attention.

Last year at the AGA-Tygem pro tourney, after boasting that I don't usually make mistakes, the very next game I made a big mistake while Myoung Wan Kim was trying to do live commentary. He had to come upstairs and correct my recording while the game was still in progress.

ez4u · Post by **ez4u** » Sun Jul 21, 2013 11:45 pm

macelee wrote:weiqi.tom.com is a major Chinese source of recent professional game records. Unfortunately, a senior editor has left his post recently and the temporary staff replacing him is much less experienced. For this reason, a lot of game records published there contain very significant errors - incorrect move sequence, incorrect date, komi, etc. What's more worrying is that some other 'collections' simply blindly copy/paste things from weiqi.tom.com without checking the details. Here is just one absurd example with a game being 625 moves long

http://gokifu.com/s/1f6m-gokifu-2013071 ... 5p%29.html

I tried my best to help correct the source. Make sure to view the recent games at Go4Go.net where games are properly checked before going in the database.

Any idea on how things are done in China currently? In Japan essentially everyone switched over to recording on PC's a couple of years ago. There was nothing formal about it apparently. Some pros started to do it after their games and everyone realized it was much easier and more useful that way, so more or less everyone switched en masse. AFAIK, the Nihon Kiin still has no policy about it. Presumably it improves accuracy later since no outsider has to transcribe the game from a hand-written record.

macelee · Post by **macelee** » Sun Jul 21, 2013 11:54 pm

Compiling a collection is equally hard and it is very time-consuming. For example, http://weiqi.tom.com/qipu/201307/39jpmry-wgzsxyzs.sgf show a recent game incorrectly marked as from Japanese Meijin tournament. This incorrect information was then shared by others (e.g. http://gokifu.com/s/1f31-gokifu-2013071 ... 9p%29.html). When I checked at the Nihon Kiin website I could not find a reference to this game. Further investigation revealed that the game was actually played in the Kisei tournament. It takes quite some effort to double-checked these details. I think TMark must have a lot of similar stories to tell.

macelee · Post by **macelee** » Mon Jul 22, 2013 12:31 am

ez4u wrote:Any idea on how things are done in China currently? In Japan essentially everyone switched over to recording on PC's a couple of years ago. There was nothing formal about it apparently. Some pros started to do it after their games and everyone realized it was much easier and more useful that way, so more or less everyone switched en masse. AFAIK, the Nihon Kiin still has no policy about it. Presumably it improves accuracy later since no outsider has to transcribe the game from a hand-written record.

I don't know exactly what is the normal practice in China. But many tournament photos on the internet do show that people still keep record on kifu papers. The fact that many SGF files contain incorrect move sequences also suggests that they are transcribed later.

Even some game records from Japan, I believe, are transcribed from newspapers. A while ago I use a computer program to process all the SGFs found in Mr Kin's Go News page and compared them against Chinese sources. Over the 4000 games checked there are about 100 are different. Occasionally move sequences are slightly different, but both making sense.

TMark · Post by **TMark** » Mon Jul 22, 2013 1:11 am

Whenever the China Cities League published their game records in book form, I would go through the book lookiing for "new" games. For every game that we did have, I would check the final position, move numbers, date, names and results. There was always a fairly high percentage of games in the book which differed from the online record, most noticeably when there were extra moves on the online record not in the book and the result was reversed. We generally took the view that the published book trumped the online copy.

Best wishes.

gogameguru · Post by **gogameguru** » Mon Jul 22, 2013 1:13 am

Yeah. This is a major headache as everyone doing this kind of thing is already pressed for time...

bazilxp · Post by **bazilxp** » Mon Jul 22, 2013 1:41 am

Mef wrote:I'm sure TMark might have a few words to add about the accuracy of game record reporting...

Heh, did notice this terrible part, 625 moves.....

One question remain who provide games without errors?

bazilxp · Post by **bazilxp** » Mon Jul 22, 2013 1:52 am

macelee wrote:
ez4u wrote:Any idea on how things are done in China currently? In Japan essentially everyone switched over to recording on PC's a couple of years ago. There was nothing formal about it apparently. Some pros started to do it after their games and everyone realized it was much easier and more useful that way, so more or less everyone switched en masse. AFAIK, the Nihon Kiin still has no policy about it. Presumably it improves accuracy later since no outsider has to transcribe the game from a hand-written record.

I don't know exactly what is the normal practice in China. But many tournament photos on the internet do show that people still keep record on kifu papers. The fact that many SGF files contain incorrect move sequences also suggests that they are transcribed later.

Even some game records from Japan, I believe, are transcribed from newspapers. A while ago I use a computer program to process all the SGFs found in Mr Kin's Go News page and compared them against Chinese sources. Over the 4000 games checked there are about 100 are different. Occasionally move sequences are slightly different, but both making sense.

what about if to look to Korean source? compare three sources ?

TMark · Post by **TMark** » Mon Jul 22, 2013 4:44 am

Here is a game that appeared in the Korean Yearbook and in Go World.

I have since found another copy, supposedly, in a Chinese book. Here are the first 20 moves as in that book.

Slight difference in the sources, neh?

Best wishes.

vier · Post by **vier** » Wed Oct 30, 2013 7:12 am

macelee wrote:Here is just one absurd example with a game being 625 moves long

Ha, that is nothing. http://web2go.board19.com/gopro/go_view.php?id=57179 is a game with 3759 moves.

EdLee · Post by **EdLee** » Wed Oct 30, 2013 1:24 pm

In addition to the medium (paper kifu vs. computer), another factor, perhaps more importantly,
is the level of the game relayers/recorders. Anyone who has watched live pro games on tom.com
or sina.com, etc. has seen the Undo's, sometimes very frequently. Massive Undo of long sequences.
This is most likely because the relayers/recorders are not Go pros. In Japan, my understanding is
they have Go pro relayers and recorders (maybe the new 1p's or other lower dan pros).
Not so in China, especially for these "news" companies -- I don't know who they hire as relayers
and recorders, and I don't know their Go level.

(Back in 2006, I was lucky enough to be at two separate pro tourneys
at the Chinese Go Association headquarters in Beijing. In the main tournament hall,
I think over 50 pros were playing; over 25 live pro games.
There were reporters and relayers at certain tables -- for example, at Chang Hao's table.
Everyone was using laptops. There was no paper kifu. This was over 7 years ago.
I think the problem is less likely about paper vs. computer;
it's more likely the Go level of the relayers-recorders.)

Does anyone know if they hire strictly Go pros as relayers-recorders in Korean pro tourneys ?
(For example, Samsung, Nongshim, LG, other live TV pro blitz games, etc.)

xed_over · Post by **xed_over** » Wed Oct 30, 2013 3:16 pm

EdLee wrote: another factor, perhaps more importantly,
is the level of the game relayers/recorders. Anyone who has watched live pro games on tom.com
or sina.com, etc. has seen the Undo's, sometimes very frequently. Massive Undo of long sequences.
This is most likely because the relayers/recorders are not Go pros. In Japan, my understanding is
they have Go pro relayers and recorders (maybe the new 1p's or other lower dan pros).
Not so in China, especially for these "news" companies -- I don't know who they hire as relayers
and recorders, and I don't know their Go level.

I disagree.

I've got a few amateur kyu players that could out record many dan level or even new pro level recorder any day of the week.

While I agree that playing skill level does bring some advantages to the table, it can also prove to be a slight deterrent. Higher skilled players often get distracted watching the game, or imagining out their own sequences while waiting with long thinking times. Lower skilled players are forced to pay closer attention since they can't always read out the possible variations.

Game recording is more about being detailed oriented and paying attention, than it is about playing skill.

And game records with thousands of moves are more likely due to computer automation errors than human transcription errors.

EdLee · Post by **EdLee** » Wed Oct 30, 2013 5:07 pm

xed_over wrote:I've got a few amateur kyu players that could out record many dan level or even new pro level recorder any day of the week.

Perhaps apples and oranges.

Of course, there exist kyu players who can relay-record pro games at near-100% accuracy. ^(*)
The question is what is the average/typical accuracy of all amateur kyu relayers-recorders;
in particular, what is the average/typical accuracy of the relayers-recorders
hired by news companies such as tom.com and sina.com.

Now let's look at the average/typical accuracy and ability of pro relayers-recorders.
In particular, the younger pros in China, Korea, and Japan.
(For example, new 1p's or new low dan pros, especially under 20 years of age.)
One of these younger pros can play 5 or even 10 simul games,
then review from memory alone all 5 to 10 games. Most of them can do this.
AFAIK, extremely few amateur kyu players have this ability.

- What is the average/typical Go level of the relayers-recorders hired by companies such as tom.com, sina.com ?
- What is the average/typical accuracy of these relayers-recorders ?

- What is the average/typical Go level of the relayers-recorders for pro tourneys in Japan ? (I'm guessing 100% pros.)
- What is the average/typical accuracy of these relayers-recorders ? (I'm guessing 100%.)

The answers to these questions can help explain the frequent Undos
in live, relayed pro games on servers such as tom.com and sina.com.

Saying that there exist (a few?) good kyu recorders does not answer the questions.

^(*) At the last Toyota-Denso cup at UCLA (2008?),
I was the 5k-relayer at Table 1, for all 5 or 6 games.
I think my relaying was 100% accurate for all but one game
which was near-100% accurate (until they started blitzing ko moves
at 0.5-seconds during byoyomi.) I don't remember now.
I believe you yourself have relayed-recorded many high dan games.
Unfortunately, this doesn't tell us much about what's happening at tom.com and sina.com.

Life In 19x19

Accuracy of recent kifus

Accuracy of recent kifus

Re: Accuracy of recent kifus

Re: Accuracy of recent kifus

Re: Accuracy of recent kifus

Re: Accuracy of recent kifus

Re: Accuracy of recent kifus

Re: Accuracy of recent kifus

Re: Accuracy of recent kifus

Re: Accuracy of recent kifus

Re: Accuracy of recent kifus

Re: Accuracy of recent kifus

Re: Accuracy of recent kifus

Re: