[Computer-go] Are 4 'easy to avoid errors' common to all MC programs?

[Petr Baudis]

On Fri, Jan 21, 2011 at 10:47:49AM +0000, Jonathan Chetwynd wrote:
> Are 4 'easy to avoid errors' common to all MC programs?

Note that easy to spot (for human) does _not_ mean easy to avoid. :-)

> 1	when behind (even slightly), Fuego appears to collapse and play
> rubbish, in my experience this is most easily perceived in the late/
> middle game.

It might be interesting to compare Fuego and Pachi in this regard; Pachi
is of similar strength (I believe latest git version rather stronger
than Fuego on the same hardware now, but we did not perform rigorous
testing yet), but uses dynamic komi to avoid this behavior. At the point
it is behind too much, it breaks down rather abruptly as well, but the
point should come much later (especially if the loss of advantage is not
sudden).

> 2	Fuego plays out ladders, which it cannot capture.

There are some techniques to avoid playing out ladders, but this
description covers many possible scenarios - playing out ladders against
opponent's ladder-breaker, playing out ladders where the laddered group
has too many liberties, when it can counter-capture some of our stones,
etc. Some are harder to address than others.


On Fri, Jan 21, 2011 at 08:31:04PM +0900, Darren Cook wrote:
> Dynamic komi. (E.g. if it thinks it has only 30% chance of winning at
> 7.5pt komi, but if you reduce the komi to 5.5pts it thinks it has a 55%
> chance of winning, then reduce the komi to 5.5pts: it will play an
> intelligent looking endgame and lose by 1.5pts.)
> 
> This will make it weaker overall (because it won't try so hard to cause
> trouble)

What do you base this statement on? It's rather controversial for me. :-)

> I.e. program endgame is generally stronger than the humans of the same
> rank; chances are a 1-dan human will make a few 1pt or 2pt errors during
> the endgame.

I also think this is not obviously true at all. My observations have
been that MCTS does not perform too well at all in very close endgames.
(Though it is not a big disadvantage in practice since it is in the
nature of MCTS to strive for deciding the game ASAP, i.e. in the middle
game.)

-- 
				Petr "Pasky" Baudis
Computer science education cannot make an expert programmer any more
than studying brushes and pigment can make an expert painter. --esr