PageType: HomePage    

As i had been playing go for while at KGS I decided to try and extract the data from the game records and see if there was anything to learn.

I cannot really share the code now as its not exactly user friendly but I could make it available and tidy it up if people are interested.


  • I win more often than lose as white: 504 wins to 448
  • I lose more often as black than win: 214 wins to 269
  • Although I appear to win and lose in equals amounts I have been promoted 4 to 5 stones: 718 wins to 717
  • I suppose you can very weakly conclude its better to be white
  • I play white far more often than black
  • I play too often ... and don't really seem to have got any better ...

Welcome any questions or ideas for analysis.

Hikaru79: I, for one, am _very_ interested in your idea... it's always been an idea of mine to try something similar, but I never have. ^__^ Perhaps you could send me your program for beta testing? ;)

I'll try to retrieve it ... its lost in a project somewhere. Its a java application, hope this is not a problem. Might be quicker to write again as its ugly code. Here is my algorithm:

- you connect to your user stats with the URL and the parameters:

gameArchives.jsp?pageMax=MAX&user=MyUserName&pageNum=" + String.valueOf(i + 1)

Where MAX is the number of pages of results you have and i is the count.

- grab the stream and add to a file; - repeat for i to M

You then have the file ... now you have to parse it which is really awful. All the data is kept in <TD> elements so its not too bad but there are inconsistencies.

Malweth: If you win more often as White on KGS, this is not a problem if you have no rank. If you do have a rank and are playing ranked games without handicap, it makes sense that you more often win as White. You will always be white (unless you change it manually, which very few people do) when playing a weaker player and black when playing a stronger player. Therefore your statistical results will tend to make sense in the long run. Check out the bottom of [ext] my games archive. I just added a statistics section for my own curiosity.

Tas: Getting stronger while having equal numbers of wins and losses is not strange. People should in general be getting stronger, but everyone can not win more than they lose. People who improves fast is likely to be slightly underrated and thus win more than 50%, while people who improves slowly is likely to be slightly overrated (as their previous oponents has gotten better) and therefore lose more than 50%. That you have an equal amount of wins an losses probably just mean that you have gotten stronge at an average rate. Could be interesting to see if the rated bots lose a lot more than average on kgs.

pedro last edited by on February 6, 2008 - 10:50
RecentChanges · StartingPoints · About
Edit page ·Search · Related · Page info · Latest diff
[Welcome to Sensei's Library!]
Search position
Page history
Latest page diff
Partner sites:
Go Teaching Ladder
Login / Prefs
Sensei's Library