Re: computer-go: Spirit of the game

[Don Dailey <drd@xxxxxxxxxxxxxxxxxxxxx>]

   From: "Grajdeanu, Adrian" <adrian.grajdeanu@xxxxxxxxxxxxxxxxx>

   > From: Don Dailey [mailto:drd@xxxxxxxxxxxxxxxxx]
   > TT scoring does   indeed capture the  spirit  of  the game.
   >>    From: "Grajdeanu, Adrian" <adrian.grajdeanu@xxxxxxxxxxxxxxxxx>
   >>    TT scoring does not capture the sprit of the game. 


   We're arguing over what is the 'spirit of the game'.
   I see the 'spirit of the game' as: control most territory using the least
   number of stones.
   This statement reflects jap scoring. Under chinese scoring the spirit of the
   game is: control the most territory. Under TT scoring the spirit of the game
   is: eradicate your opponent from the board. Jap scoring 'spirit of the game'
   is stronger then the others and thus is the most difficult to implement. A
   player that plays best under jap scoring will have no problem playing best
   under chinese scoring. Yet the other way around it doesn't hold true because
   an extra deffensive move comes for free in chinese scoring, but at a cost
   for jap scoring. A player that plays best under jap scoring can play its
   best up until the point when it would pass and then switch into 'kill at all
   costs' mode and achieve best play under TT scoring. A player that play best
   under TT scoring will fall short under jap scoring as it doesn't know the
   best time to pass.

This is  the silliest  thing I ever  heard!!  You  are  defining "best
play" by  the rules themselves!  Playing  inside your own territory is
NOT best play in chinese scoring either unless the game is essentially
over and there is no progress to made, in which  case it is neutral in
Chinese and  BAD in Japanase only  because it is  defined to be bad by
the rules!!!

I could make a rule change in Chess too  and say that  if you have a 3
move checkmate and take 5 moves to do it instead,  you lose!!!  Then I
could claim that  this version  is more  in  "the spirit" of the  game
because it forces you to win quicker.  Why is this  more in the spirit
of the game to win 2 moves quicker?  Because I am defining it to be so
by the rules!!   The proof is that  you can win  a game using the  old
rules that the new "stricter" rules just knows is bad!!


   In the light of jap 'spirit of the game', here are the problems I see with
   TT scoring:
   1. require removal (capture) of dead stones.
   2. don't penalize for playing inside own territory.

No,   these are   not   "problems."  Removing  dead   stones  is not a
"problem", it's a requirement.  If  it's not done explicitly it's done
in the minds of the  players, why is  this always overlooked?  You can
make a program still do this if it's so important  to you, the scoring
system is not relevant.   Chinese  rules are essentially Tromp  Taylor
rules with this mental shortcut allowed.

You ARE penalized for playing inside your own territory no matter what
ruleset you use.  You make  it sound likes  it's perfectly ok to waste
time moving into your  own territory  in  Chinese.  In  either scoring
system, a given move has the same status, it's either good or bad.

The only caveat,  which you seem  to think is  a big deal, is that  in
Chinese scoring you  are not "punished by  decree" for filling in your
own territory after it doesn't really matter anyway.



   And here are my comments to them:
   1. I certainly understand your concerns regarding cheating the nets from
   learning the end game. Yet that happends only if a referee stops the game
   when it knows the end. In my case I plan on having the nets play as long as
   they consider playing to be meaningful. In this case a stubborn net may
   challenge the opponent and try to save dead stones. The opponent needs to
   learn to kill, or otherwise it looses and eventually doesn't survive. So at
   least in theory I am not concerned of not learning the end game. But a net
   that is good and 'sees' that a group is dead no matter what, shouldn't try
   to kill it as it would mean to play inside territory and thus fall pray to
   what I see the second shortcoming of TT scoring.

   2. It is generally bad practice to play inside own territory when there is
   disputed area to be conquered. And a net need to learn this. TT scoring can
   help. But a net also needs to learn that if there is no more disputed area,
   it is better to pass then to play inside own territory. TT scoring won't
   help here.

   Now, I will clarify what I understand jap scoring, chinese scoring and TT
   scoring.
   TT scoring counts as points the stones on the board, the territory fully
   enclosed by own color and prisoners.
   Chinese scoring counts as points the stones on the board, the dead opponents
   strings, the territory fully enclosed by own color (ignoring dead strings)
   and prisoners.
   Jap scoring counts as points the dead opponents strings, the territory fully
   enclosed by own color (ignoring dead strings) and prioners.

   Throughout this message I purposly ignored rules differences as suicidal,
   super-ko, or whatever else not clearly specified above.


All I'm saying is that I think that you don't have to be worried about
which   ruleset  you use.   I   believe  you are attaching    a lot of
importance to this.  I feel that you are feeling  guilty for not using
Jap scoring and seem to think that it is degrading your results.  What
you  are really looking  for is a  way to  get an  evaluation and good
Japanese  end  of game scoring  really   is  just  a fancy  evaluation
function (which is why it's hard to do a good one by computer.)

Don