Hi Peter, Talking about (program) strength in general is tricky. There may be big differences depending on board size, time control, the opponent (familiarity with the program), specifics of the task (opening, endgame, semeai, tsumego, or full games), regional differences (1d AGA != 1d EGF), so I think you should try to make the question more precise.
I have some ideas on program strength before 2006, but it's rather anecdotal. The big problem, especially with classical programs, is that once their weaknesses are know they can typically be exploited rather easily. My impression is that, against naive opponents, for 9x9 'classic' programs got to something in the order of 3-6 kyu, while for 19x19 it's more in the order of 6-10 kyu. Perhaps organize a big tournament with some top engines from the last 20 years, possibly mixing in some human anchors, and infer the relative strengths from that? Maybe Stefan Mertin's data could be of some use (see http://igosoft.com/) BR, Erik On Sat, Sep 27, 2014 at 1:48 AM, Peter Drake <[email protected]> wrote: > I'm giving a talk on computer Go (and how to play Go) to a regional > computer science conference in a couple of weeks. > > Has anyone compiled a table or graph of the best programs' strength over > time? Ideally it should make the jump after Monte Carlo clear. > > I am aware of these: > > http://senseis.xmp.net/?KGSBotRatings (only goes back to 2007) > > http://www.computer-go.info/h-c/ (a bit difficult to condense into a > single line graph) > > > -- > Peter Drake > https://sites.google.com/a/lclark.edu/drake/ > > _______________________________________________ > Computer-go mailing list > [email protected] > http://dvandva.org/cgi-bin/mailman/listinfo/computer-go >
_______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
