Re: computer-go: Go Devil

[Vincent Diepeveen <diep@xxxxxxxxx>]

At 02:38 PM 9/23/00 -0400, you wrote:
>
>Part 2 of Go-Devil.
>
>In my last  post, I talked about  rating curves in  Chess and tried to
>make a case that the very best human  players actually don't play very
>well compared to what would be possible by omniscient entities.

That thought is completely nuts, and totally wrong.

Give a grandmaster a pawn up directly and he'll destroy you completely.
Now you can argue that he mates you in 50 moves instead of 34,
but you'll lose chanceless.

Fact is that top gm's make on average 1 move a game which is obviously
not the best. Apart from that, some make under big time contraints
now and then an error (a big one) somewhere near time trouble.

Now chess is like this that you play with very little pieces compared
to go, so a big error is directly turning the game in favor of your opponent.

The world top gm's of 2700 and above however are playing a type of game
which is far better. Usually only a single 'dubious' move at most a game,
which they find dubious, but i don't mind playing plenty of those moves
in a single game.

Now of course the question is directly: "how can some lose to programs?"
I wonder about that too.

There are different things. First of all we must not forget that
not all games of them are that well. Only games where they play for
something important are real good usually. Playing each day a game
is obviously not very good for the level of the games.

Not seldom we see in the so many-th round many errors occurring, usually
from both sides.

Still the real question is: "why do they suck (with a few exceptions)
against computers?". Question is: they didn't show the last of their
tongue.

See the games Kasparov-Karpov. In those games you will not see many
errors. In fact the matches Kasparov-Karpov are real high quality games,
the best ever shown. At most you can point down that nowadays
different opening lines get played, but the games itself real high
quality. On average less as 1 real bad move a game. I remember only 1 game
where a blunder (deciding the game directly) has been made by Karpov
in a position where he was already under immense pressure the whole game.

Yet the games kasparov-deep blue had about 6 bad moves a game at least,
from kasparov side even complete blunders in game 2 (resigning in 
a drawn position) and game 6 several blunders. Also kasparov didn't play
his usual style. He kind of gave Deep Blue several games 9 stones up
so to speak, and very obviously he didn't try to win the games
in the opening, which would have been a real peanut for Kasparov.

Yet the years before that, despite all those bad moves, this was enough
to beat the programs. I think Kasparov's only mistake was to play the
18 stone up game against deep blue in game 6 in the last match we know,
otherwise we would consider nowadays that supercomputer programs suck
incredible, but that pc programs are way better on a slow machine.

Also very surprising were the games human-computer at Paderborn
after the world champs 1999. We clearly see there that grandmasters
just do *something* except play what they normally play. 

"we don't get paid to win there". 

On the other hand even very bad GMs which played over the internet against
Rebel managed sometimes from already bad positions for the GMs to get
easily a draw, or even win the game. So they were playing on the computer,
which is already worse as playing in a tournament hall at your own real
3d board (less distraction etcetera and especially an extra dimension).

Add to this that last years we see a new generation of grandmasters who
play tactical far better as grandmasters from some tens of years ago.

Also their insight in the game is in my opinion way better. 

We even see players like Kramnik (who is gonna play Kasparov for 
the world title in octobre) find new ideas in very old systems, 
which are at least 100 years old.

The whole assumption that these players make a lot of mistakes is
completely wrong. If you take into account what level they play at,
in important games for them, then you'll also clearly see that the
level is much higher as most think.

>I have some further thoughts on this taking it from a different angle.
>In the previous post I made reference to Kasparov's play and held this
>up  as an example of  the best human play.   However  there is kind of
>chess that is generally  acknowledged to be of  a much  higher quality
>than this,  and it's called   correspondence chess.  I have  talked to

The average correspondence player loses against Kasparov blindfolded.

The average correspondence player plays a good opening, but this opening
is litterary out of the books he has, which are made by grandmasters
way under Kasparov's level.

Kasparov completely finishes an opponent if Kasparov has openingsadvantage,
he has shown in strong tournaments that he usually doesn't even need
an openingsadvantage. Basic point is that they will win a won position,
in the style they can chose, where correspondence players are simply
to dependant upon their resources to play.

I played myself correspondence chess, and the level is real pathetic
from technical viewpoint,
and usually dominated by good openings and bad technique to finish the
game. Actually several games in so called 'mastergroup' there i
could as 17 to 19 year old boy from a lost position convert 
into won positions! 

It appeared i was playing basically chessprograms, which still dominate
the lower levels of correspondence chess. In fact many chessprograms
including my own have many national correspondence titles already 
(of course not the opening played by the programs). Rebel especially
has countless titles all over the world. 

In general people who hardly can play chess over the board join
such competitions, and very few of them, as there is hardly money to win.

No money no good players!

The only high level games in correspondence chess are very few world
league games and some country versus country games on the highest boards.

Those high level games you can however count on a few hands, and are usually
dominated by a few GMs and IMs or masterclass players who play for fun
correspondence chess, so basically dominated by players who over the board
are playing quite well too.

>several master on this subject as well as one former world champion in
>correspondence chess  and the consensus is  that the level of  play of
>top  players   FAR  exceeds    that  of over    the  board   play.  In

The average top GM game is of a much higher level as the average
correspondence game, yet there are of course a few advantages to the
correspondence players: they hardly blunder for example. They use the
latest openings books, but despite that they are outgunned in opening
against any 2600+ GM who invents his own new theory instead of
litterary selecting lines from just published books, therefore with
induction the above hypothesis is wrong.

>correspondence play,  each  player has  many hours  or  even  days  to
>respond to a single move, and uses this time to do serious analysis.

>As a thought experiment, if you could give a top correspondence player
>the   ability to  make  the same  high   quality moves  but  in normal
>tournament time controls, you  would  have a  player far  superior  to
>Kasparov in over  the board play.  I  have  never heard a  good player
>deny this conjecture and I have talked to several about it.

I doubt this. Kasparov prepares a lot at home too. It's different
of course if your correspondence player is called Kramnik.

Basically this theoretical possibility is supporting your opinion,
but if you take the average correspondence player who has a NATIONAL
TITLE : "national champion correspondence chess", then Kasparov
will destroy him completely IF HE WANTS TO.

Question is: does he need to destroy him by showing his latest opening
lines prepared, or will Kasparov use such lines in the upcoming match
Kramnik-Kasparov?

>When games of the top players are analyzed later, it is common to find
>flaws in the  play.  It is so common  in fact that  it is clear we are
>playing far from perfectly.  There is a still a very large gap between
>the current level of play of top humans and perfect play.

What is your definition of perfect play?

  - a) not losing drawn positions, 
       winning a won position, and getting the best out of lost positions,
       so sometimes converting them to a win?  
  - b) mating an opponent using optimal play so if it's a mate in
       56 theoretical then also finding this mate in 56 behind the board?

If your definition of perfect play is b, then you're right.
If it's a or something similar, then you're dead wrong.

I advice you to analyze some games from Kramnik with some strong
players, using published analyzes and everything you have.

A good example is the game he played against Polgar in a demonstration
game at television recently. Try to analyze his moves in this 1 hour
a game level instead of having discussions like this!

They are always started by people who can't play chess at all!

>Because of our limited perception, it is  difficult for us to see much
>farther than from our own vantage point.  To us, brilliant and perfect
>play is simply whatever the best of us  can achieve.  We are impressed
>with   our own   accomplishments.  But to   a  God,  what  we consider
>brilliant and perfect might very well be simple minded and naive.
>
>I would like to also point  out that player perfect  moves most of the
>time does  not  mean you  play good  Chess  (or Go)   in general.  The
>difference between a  great player and  an  excellent player might  be
>that the excellent  player plays  perfetly  90% of  the moves and  the
>Great player gets it  right 93% of  the time.   Over a long  series of
>games the great player will  amass a huge lead, especially considering
>that the bad moves of the weaker player will get punished more often.
>
>In a real sense, there is no such thing as a good or excellent move in
>2 player games  of perfect information like  Chess and Go.  There  are
>only bad moves.  A good move is a human label  applied to a move which
>is difficult to find or is  "pretty", a Go Devil  kind of move!  A bad
>move is  a  move that doesn't   maintain  the game  theoretic outcome.
>There is a maxim in chess that says the winner is the next to the last
>person to make a mistake!   
>
>Finally, I don't know how to apply this to Go.   I feel that we are at
>least a few hundred elo rating points from perfect chess, even as many
>as 1000 (which means a hypothtical player exists who would rarely ever
>lose to Kasparov, and yet there still exists a player who would rarely
>lose to this player!)
>
>Christoph gave the rough  rule of thumb  that  1 stone equals 100  ELO
>rating points.  If this is  approximately right, then  the idea of the
>best players  being  close to perfect  (within 1   or 2  stones) seems
>fairly ludicrous.  200 rating points (or two stones?) means that among
>equal players, the handicapped player would still win about 15% of the
>time although losing about 85% of the games.
>
>I think Go  is MUCH more profound and  difficult to master than Chess,
>and this by a really wide margin.  I have heard similar numbers quoted
>for  chess, where it's been  hypothesized that we  are only 200 or 300
>points away from perfection.  I even admit that it's hard to imagine a
>player  who could achieve  a score of  over 80% against  the great and
>wonderful Kasparov, but I think that kind of  thinking is what ties us
>in knots, we really shouldn't make our guesses by comparing to players
>we worship because they are too close to God in our minds!
>
>
>Just for reference here is a simple table of expectancies that the ELO
>rating system predicts.  For instance the last entry indicates that if
>you are 400 points higher rated, you will win 97% of the games against
>your weaker opponent.  Most of  the experiments I mentioned earlier in
>computer chess, for those not familiar, were based on playing hundreds
>of  games against programs   of  various depths,  and calculating  ELO
>ratings   based on winning   percentages.  In the  experiments, it was
>shown  that  by searching 1  ply deeper  you  could  obtain a stronger
>program, one that could  expect to win about 80% or more of the time.
>
>
>    ELO diff    wins %
>    --------    ------
>        0        0.50
>       20        0.53
>       40        0.58
>       60        0.62
>       80        0.66
>      100        0.69
>      120        0.73
>      140        0.76
>      160        0.79
>      180        0.82
>      200        0.84
>      300        0.93
>      400        0.97
>
>