I have been following this sort of statistics for ages and I have just made a new discovery. According to an old GoGoD article, Nie Weiping "appeared in a 363-move marathon against Komatsu Hideki 8-dan in the 8th Japan-China Supergo series on 6 December 1993. This game apparently claims the record for the number of ko captures (59)."

http://www.go4go.net/go/games/sgfview/10228

Several times I got close to this records.

I was going through old Korean Kuksu games when I made the following discovery. A game played between Seo Bongsoo 9p and Kim Seungjun 6p on 1998-08-12 is 364 moves long and has 64 ko captures! Making it more special is that the first ko started quite late at move 152! I also think that the ko fights in this game are more interesting because they are mostly about life and death rather than yose. Enjoy the game.

http://www.go4go.net/go/games/sgfview/45012

Neat find! To save time for lazies on iPad (where using SmartGo Kifu is more convenient, it's game 1998-08-12f

_________________
Geek of all trades, master of none: the motto for my blog mostlymaths.net

Go Blackie!

What would be the definition of a ko capture? Your URLs yield
"Only registered users have the previlige [sic] to view ..." but for
http://homepages.cwi.nl/~aeb/go/games/games/CJSuperGo/CJSG-08-09.sgf
I count 60, not 59, ko captures, at moves 64, 67, 86, ..., 310, 341, 359.

I found the other game at
[url]http://jinix.sf.net/go/sgf/06.韩国-经典/徐奉洙对局集/9/00889__徐奉洙_Vs_KimSeungJoon.sgf[/url]
(hope that the non-ASCII survives) and agree with 64 ko captures.

Some searching yields examples with up to 91 ko captures. Some of these may be amateur games. An old game, probably you have it, is Ito Matsujiro vs Yasui Sanchi (1839-05-15, 405 moves, 70 ko captures).
A modern example is Kim Hyeoimin 6p vs 金顕燦 1p (2012-04-14, 372 moves, 70 ko captures), see
http://www.kihuu.net/sgf/k00000111334.sgf.

Although GoGoD compiled long lists of record-breaking games, as indicated by my sniffy "apparently" this particular record, along with record number of captures, is rendered awkward by the fact that Japanese and Korean records often omit the final ko fight and simply add something like "Black wins the ko". My guess is that it is the longer such fights that would be omitted most, but either way this practice affects not just the number of captures but also indicators such as the average length of a game.

Sometimes the pros will just agree without play about who wins the final ko, but can play it out for quite a long time (relying on memory, for even up to an hour). Do bear in mind that not every game had a game recorder and the winning pro was responsible for supplying a game record from memory (and memory was often unreliable even for non-ko-fight games, which is why there are alternative records for so many China-Japan games, recorded in this case by memory on both sides; in keeping with being your most reliable database, GoGoD usually gives or alludes to both versions).

The proper definition is derived from the definitions here http://home.snafu.de/jasiek/ko.pdf as a play on a ko-intersection so that it remains a ko-intersection and that captures at least one ko-string. However, I'd guess that a more naive, possibly false definition would be used when counting record numbers.

To an administrator: Did I do something wrong in the above url? Or can't one use non-ASCII urls?

To macelee: One more example: in the 2005-09-11 quadruple ko game Lee Changho vs Chang Hao with 369 moves, I count 67 ko captures.

To JF: Yes, of course; game length and number of captures are functions of a given game record; the historical truth may be unknown or different. They also depend on the precise definitions used. (In your example I count 60 ko captures and don't know how to set up definitions that would give 59. In my modern example with 70 a slightly different definition of ko capture would make it 69.)

To RJ: Ugh.
(How can a definition be false? A definition says: this word or phrase will have that meaning in the present context. A definition is true by definition )

To JF: "lists of record-breaking games" - nice! Can you name some records? We can look and see whether there are even more extreme games.

Thanks for the contribution. It would be interesting to write a small program to check these things automatically.

Yess. Program. Written. (Myself, I couldnt reliably count to sixty.)

There is a famous long game, Yamabe Toshiro vs. Hoshino Toshi
(411 moves, of which 104 captures, 66 ko captures).
I Googled for it, and found
http://senseis.xmp.net/?FamousGamesInvolvingKo
where several of the games mentioned here also occur.
Unfortunately most of the sources given require a password or registration, so are rather useless.
But then, Google also knows about
https://badukmovies.com/pro_games/4762-hoshino-toshi-yamabe-toshiro

I mentioned an example with 91 ko captures, so should provide a pointer. Here is one.
http://web2go.board19.com/gopro/go_view.php?id=13709
People will say: "that doesn't count". But then one needs a tighter definition of precisely what the conditions are.