[gnugo-devel] 10 games

[david doshay]

Here is the result of a 10 game sequence of Black = GNU Go 3.4 v.s 
White = our modifications of GNU Go 3.4 to add 4 parallel lookahead 
paths.

3 of the lookahead paths are the top 3 moves selected by GNU Go, the 
4th is the top move selected by GNU Go if we were to pass and Black 
were to choose (unless this is an illegal move for white, in which case 
the 4th choice is the 4th highest move for white). We often choose to 
take "their" best move over our best move. In other words, we are 
testing the path suggested by the Go proverb "Your opponent's move is 
often your best move."

We then do an 8 move lookahead using GNU Go's first choice for each 
player. At the end of the 8 move lookahead we use get_influence to 
evaluate the board. We then do a normalized linear combination of the 
influence and the original move values to select between the 4 moves. 
We are still working on the way we combine the values, and there are 
some other issues involving using "their" best move that we are working 
out. For example, if their best move is a monkey jump, and we select 
the path the starts with our defense against the monkey jump, at 
present our code plays the same point black would have played. We are 
changing that right now.

Game 1: W+48.5
Game 2: W+14.5
Game 3: B+36.5
Game 4: W+11.5
Game 5: W+24.5
Game 6: B+16.5
Game 7: W+5.5
Game 8: W+21.5
Game 9: W+5.5
Game 10: B+0.5

I poked around for a while and did not find the default komi used in 
the above scoring.

We have the .sgf files of these games and a logfile containing the 
evaluation criteria, should anyone have more interest. We are still 
changing the details, so this set of 10 games will not really compare 
with future sets.

Cheers,
David