Re: [Gtkboard-devel] help writing a game

[nelson <gn...@ce...>]

El lun, 27-09-2004 a las 01:48, Arvind Narayanan escribi=F3:
> Hi!
>=20
> Well first of all its about a year since I worked on gtkboard, so I
> might not remember all the details right, but anyway...
>=20
> On 26 Sep 2004 23:37:23 +0100, nelson <gn...@ce...> wrote:
> > Hello :), Im writing a game for gtkboard, is an entertainment pencil an=
d
> > paper game,Im firstly trying to write the Machine to Human type, but Im
> > a bit confused how to connect my machine-turn C code to the game_eval
> > and  game_movegen hooks, you know, I have my own c algorithm for the
> > machine to choose its move from the existent board, but there is no an
>=20
> The game_eval and game_movegen functions assume that you want to do
> game tree evaluation (alpha-beta and generalizations thereof; see
> http://en.wikipedia.org/wiki/Alpha-beta_pruning in case you are not
> familiar with those algorithms). game_eval and game_movegen are hooks
> for the eval and movegen algorithms in minimax search.

Yeah, wikipedia is great !, I could understand a bit more the game_eval
and game_movegen trough the well explained concepts at wikipedia, people
will find useful to begin looking at
http://en.wikipedia.org/wiki/Game_tree

> However, I think you're not using minimax, so you don't want to use
> those. There's a game_search function, which lets you search the
> position yourself and return the best move. If game_search is
> non-null, it will be called and minimax/alpha-beta will not be
> executed.

Yes, that game_search is exactly what I wanted, I hadn't noticed it at
the docs :).

> > easy hook where only put the position chosen by the Machine (that
> > position computed by a function of mine)...instead it seems gtkboard
> > forces me to use its functions for searching and calculating new
> > movements (game_eval and game_movegen), that sure are more powerful and
> > faster that my own but they are not well documented and I cant find out
> > how to use them. Questions about that:
> >=20
> > - Is there an easy hook where to put the machine move computed by me
> > (avoiding game_eval and game_movegen)?
> >=20
> > - What value Am I supposed to put on the float *eval argument?
> >=20
> > - How Can I control the turn flow ?, because in my game a human can mak=
e
> > three moves in 3 clicks (3 events) and then pass the turn to the
> > machine,but it seems gtkboard passes the turn between human and machine
> > automatically after a one succesful move...
>=20
> Ah! As far as the program is concerned, when you complete one "move",
> the other player gets the turn. So you are supposed to pretend that a
> move has not been completed (by returning 0 in X_getmove() rather than
> 1) and still update the board on the screen. There's some ugly stuff
> like RENDER_* to do that, but not difficult. Lots of games do such
> things, such as mastermind.
>=20
Ok, I finally have this behaviour working with the RENDER thing, but I
looked at the code of HIQ game, because I use game_event_handler and not
game_getmove, and hiq has this behaviour pretty clear with the
game_event_handler hook.

> Feel free to ask me again if something was not clear.
>=20
> > My game also will have levels of difficulty (machine thinking a bit
> > smarter).
> >=20
> > Explaining of the game:
> >=20
> > Initial Board:
> >=20
> > X X X X X X X
> >   X X X X X
> >     X X X
> >=20
> > Simple Rules:
> > Loose the one who removes the last piece.
> > In a turn, you can only remove pieces from one row, but you can remove
> > as many pieces as you want, for example the entire row.
>=20
> In other words: Nim. Yes, you definitely don't want to use alpha-beta
> for that one :)
>=20
> Fascinating stuff though. I spent about two months reading up the
> Conway Berlekamp theory and I loved every bit of it. It would be nice
> to see some Nimlike games in gtkboard.
>=20
Have a link to that theory, for curiosity?, what do you mean by nimlike
games? you mean simple games like mine?..

> Good luck
>=20
> Arvind
Thank You, I think I could send you the game in next days because your
answers solved my bigger problems :)...

> > PD: Thanks and sorry for the long text...