Re: [computer-go] Pattern matching - example play

[Darren Cook <darren@xxxxxxxxx>]

> > your prediction rate sounds great. does it vary much between the
> beginning,
> > middle and end stages of the game?
> Yes.
> The prediction graph looks like a bathtub.
That is very interesting; do you have the graph online anywhere?

I mentioned a few weeks back how I thought 46% was impossible - there are 
just so many equally good moves on the board at any time. So it seemed like 
a bug or over-training. Frank didn't follow-up; perhaps I'm in his kill list?

However then I realized there is probably an urgent move most of the time in 
the middlegame, so I realize a good performance in the middle game would 
bring up the overall average.

So the prediction graph (defined as number of times they guessed the next 
move correct on their first try) I would expect for a pro on a pro game 
would be 20% or so in the first few moves, rising quickly to 50-80% from 
move 20 to move 200 or so, then trailing off down to 30% in the endgame as 
they chose the small gote moves in arbitrary order. A ko fight might show as 
a dip in the middle game.

This opinion is based most on watching pro commentate on games. Anyone care 
to propose why they think the graph would look different?

Have you noticed any correlations between game style and the prediction rate 
in each phase of the game? E.g. by fuseki, by country of the players, by 
time limit, etc.

Darren


> At the start, the system's Fuseki patterns are used.
> Then the Joskie patterns are used and a whole lot of patterns that are not
> striclt Joseki but large patterns that are common in the first 100 moves.
>
> The more "tactical" a game is with (for me) an "erratic" game development (I
> mean games that not are about a slow building up of equilibrium but are more
> like one big exploding fight), the less well the pattern system predicts. I
> have seen prediction rates as low as 19% for such games.
>
> At the end, the pattern system also works well because end patterns are
> about filling dame etc. Well-established ways to finish up things. So I
> often see a 60% prediction rate in the first 100 moves or so.
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/