[OLPC-Games] Play Go issues for laptop

Paul Barchilon shimari at comcast.net
Thu Nov 20 22:48:10 EST 2008


Dear Walter and Andres,

I have been so excited about the OLPC project that I finally put my  
money where my mouth is and have just purchased one through the give  
one get one campaign.  I expect it in the next few days, and I can't  
wait to play with it!  I will also take it to my children's club and  
have all the kids play your program.  This will give us a good  
indication of the programs playing strength, which would be critical  
for any kind of handicap mode.

Glad you have found my feedback helpful.  I would still very much like  
to discuss the possibility of creating an interactive tutorial  
directly within the game.  It would be a much more natural way to  
learn.  As I mentioned before, I would be happy to write all of the  
content, you would just need a way to have the program play out the  
moves and offer options.  I would also like to edit the Wiki page on  
Go, or replace it completely.  Would that be okay with you guys?  Is  
there a review process or something that I should go through?

Also, I wanted to share some interesting comments from AGF President  
Terry Benson.  He is referring to area counting instead of territory  
counting.  Area counting counts both the territory and the stones on  
the board.  This is the traditional Chinese method.  Territory  
counting is the traditional Japanese method and counts only the empty  
space.  Terry's point is that if the program used area counting, then  
it wouldn't matter that the computer wasted moves to capture every  
dead stone or group in its territory, it also wouldn't affect the  
person's score.  It wouldn't solve the problem of teaching beginners  
bad moves, but it would definitely be easier to program.  Terry's  
comments are below:

I am curious about the underlying rule set and rule concept being used  
and, while I've read everything, I can't quite be sure. We are (I  
hope) using an area concept. The computer appears to count "territory"  
and so beginners will lose if they make unnecessary moves. This is one  
of the reasons I think territory is bad (besides the inability to see  
when the game is over). Capturing all the dead stones is "unnecessary"  
except when you can't see that the stones are dead. What you and I  
would consider necessary playing out is, to the pros, just as  
unnecessary as the dead stones on the left of the screen shot. It's  
all relative. The point is to be able to let the players play out any  
situations unclear to them with no penalty.

If the computer counts everything - living stones and territory - it  
will solve a bunch of problems. The new WMSG rules use Ing Fill-in  
which is equivalent to area. So anything we do should be area based.  
Beginners don't need pass stones if they count stones and territory  
and I would make that explicit. ("Players in some countries make the  
counting easier by only counting the empty territory points. But in  
that case, players must hand over a pass stone as a prisoner each time  
they pass.") The new WMGG rules do not require white to play last and  
- in fact - deduct one point if white passes first (meaning black  
played last). This adjustment in the komi (either 6.5 or 7.5 depending  
on who passes first) makes a 6.5 komi meaningful.

That's background info. For the case here, the point is that when you  
pass you hand over a stone. If black plays last, white doesn't (as  
under AGA rules) hand over an extra pass stone. This is close to WMSG  
rules.

Thanks much,

Paul

Paul Barchilon,
American Go Foundation

http://tigersmouth.org
http://agfgo.org




On Oct 24, 2008, at 6:58 AM, Walter Bender wrote:

> Great feedback!! Especially in light of consideration of our learning
> goals. Thanks.
>
> -walter
>
> On Fri, Oct 24, 2008 at 4:02 AM, Paul Barchilon  
> <shimari at comcast.net> wrote:
>> Hi Andres,
>>
>> Here is a list of most of the problems.  Some of these may be  
>> caused by
>> running the program on the emulator.  I was able to fix the view  
>> menu,
>> thanks for the suggestion.  With komi the scoring is correct,  
>> except the
>> half point issue.  The number of prisoners I had in the games I  
>> played did
>> not make it evident that there was komi, I thought they just  
>> weren't being
>> counted, or in one case added to the wrong color.
>>
>> 1) A player should be able to choose color as well as board size.
>>
>> 2) Your program is strong enough to beat any beginner, so you need  
>> to create
>> a handicap module.  With handicaps, the AI could play people at 18  
>> different
>> ranks, guaranteeing a lot more interest in the game.  If a beginner  
>> simply
>> loses every game, they will rapidly lose interest.
>>
>> 3) The scoring mechanism is abstract and confusing.  A visual  
>> display of
>> prisoners would be helpful.  The game lists prisoners as "score"  
>> which is
>> not correct.  Komi is not indicated anywhere, but that should also be
>> something you can choose along with handicap.  It should be  
>> displayed on the
>> side, along with "Prisoners" for each side.
>>
>> 4) The computer waits to display the move you make until after it has
>> decided on its move.  This makes a person think their move hasn't  
>> been
>> registered.  When the move does appear, it comes at the same time  
>> as your
>> prior move.  When there are many stones on the board, it is extremely
>> difficult to find where the computer moved - especially if it is  
>> not in the
>> same area.  The last played stone for either color should have a  
>> circle or
>> some other mark in it.
>>
>> 5) The program does not understand the concept of dead stones.  It  
>> must
>> surround and capture stones in order to score them.  In actual play  
>> when
>> both players realize a group or a stone has no possibility of  
>> making two
>> eyes,
>> the group is simply abandoned.  It is not necessary to spend extra  
>> stones to
>> capture them, they should be removed automatically at the end of  
>> the game.
>> Obviously this is a difficult assessment for the computer to make.  
>> KGS
>> solves this problem by asking the player to mark any dead stones.   
>> Some kind
>> of pop up menu that explains the concept and tells a person to mark  
>> the
>> stones prior to scoring would work.  It is a little tricky because  
>> new
>> players won't know which stones are dead either.  Solving this  
>> problem will
>> take some creativity, but I am sure we can come up with something.   
>> I have
>> screen captures to demonstrate this, but I am not attaching them to  
>> this
>> message as doing so seems to cause the message to be blocked from  
>> the list.
>> I will try sending the screen caps to your personal e-mail and see  
>> if that
>> works.
>>
>> 6) The program makes unnecessary moves at the end of the game.  It  
>> appears
>> to think that all stones must be connected into a single group.  It  
>> also
>> fills points and makes eyes where they are not needed.  This could  
>> actually
>> cause the program to lose a game it has won, and it could possibly  
>> cause it
>> to fill a liberty that would put its own group in danger, or create  
>> a seki.
>> Again, I have a screen cap for this.
>>
>> 7) I get hovering ghost stones that get stuck on the board.  They are
>> semi-transparent, and occur at random.  This may be something the  
>> emulator
>> is causing.
>>
>> 8) I have had difficulty starting a second game with the AI every  
>> time.  It
>> fails to clear the board if you click "Play against Play Go." If  
>> you hit the
>> reload arrow, it clears the board, but then nothing happens after  
>> you play
>> your stone.  It seems to be in two player mode.  If you click play  
>> against
>> play go again, still nothing happens.  Then if you place a second  
>> stone, the
>> AI kicks in and you are white and the computer plays black, but the  
>> person
>> had to play the first two stones.  I had to keep clicking on reload  
>> and new
>> game repeatedly to get it to launch another game.  The first one  
>> always
>> worked, the others are problematic.  Could be an emulation issue.
>>
>> 9) I see no way to choose opponents, and no way to negotiate color  
>> with them
>> either.  If two kids are playing over their local network, they  
>> need to be
>> able to do this.  Perhaps this functionality has not been added to  
>> the
>> program yet.  A chat interface within the program would be helpful  
>> for this
>> kind of thing, although perhaps kids could use the OLPC chat  
>> program while
>> they are playing.  Whatever the case, some communication is  
>> necessary to
>> start the game.
>>
>> 10) The introduction page, from Wikipedia, is not really a good  
>> guide for
>> kids.  Our Tigersmouth explanation is better, but probably needs  
>> alteration
>> for this project too.  The Interactive Way to Go might be the best  
>> choice,
>> as it is already in a great many languages.  It needs Java to run  
>> though.
>> Does Java come pre-installed on the laptops?  I tried to download and
>> install it, but it wouldn't work.  We don't want kids to have to  
>> download
>> something separate to learn how to play.  A tutorial within the  
>> program
>> might be a better way to explain the game to kids.  We could write  
>> the words
>> and create the moves if that is something you might be able to  
>> implement.  A
>> learn the rules button could launch this mode, kids could click on  
>> a button
>> to advance after they read each description and looked at the  
>> example.
>>
>> We should play on KGS and talk about these things.  I think you are  
>> three
>> hours earlier than I am, depending on where in Uruguay you are.  I  
>> am GMT
>> -7.  Evenings are good for me, but I could play during the day if  
>> necessary.
>> Next Wednesday, the 29th, would work for me.  Other days are okay  
>> too, let
>> me know what works for you.  the 13 year old who has a laptop said  
>> he might
>> be able to join us also.  He thinks he is about 19 kyu, he cannot  
>> beat your
>> program.
>>
>> I am very glad your English is so good, my Spanish is terrible ;)
>>
>> Thanks,
>>
>>
>> Paul Barchilon,
>> American Go Foundation
>>
>> http://tigersmouth.org
>> http://agfgo.org
>>
>>
>>
>>
>>
>> On Oct 23, 2008, at 1:40 AM, Andrés Ambrois wrote:
>>
>>> On Thursday 23 October 2008 05:15:36 you wrote:
>>>>
>>>> I sent this message last week, but it came back with a message  
>>>> that it
>>>> needed moderation (I guess because I included a screen shot).  It
>>>> appears to have never been posted to the list, so here it is again
>>>> without the image:
>>>>
>>>> Hi Andres,
>>>
>>> Hi!
>>>
>>>> I spent a good deal of time playing with the program and overall  
>>>> was
>>>> quite impressed!  It looks great, and plays fairly well.  There are
>>>> many little problems, and a couple of major ones.  The scoring  
>>>> engine
>>>> got the wrong result on every game I played, there are clearly some
>>>> mistaken instructions in the module itself.  Should I post the
>>>> problems to this list or would it be better to talk about them on  
>>>> the
>>>> phone, or during a game on KGS?  This week is looking pretty busy  
>>>> for
>>>> me, so it would have to be next week for the latter.  I am on  
>>>> mountain
>>>> time in the US, where are you?
>>>
>>> Thanks! I wonder how the scoring engine is failing. What it does  
>>> is take
>>> your territories, subtract the prisoners your enemy has taken, and  
>>> if
>>> you're
>>> White add 5.5 komi. The scores are currently displaying as  
>>> integers, this
>>> is
>>> stupid and I need to fix it.
>>>
>>> Please post any problems you find :). I'm from Uruguay, so your  
>>> phone
>>> bill is
>>> going to be an issue =P.
>>>
>>>> Also, I am unable to right click in the software.  I think the
>>>> emulation doesn't recognize my mouse.  Option click or command  
>>>> click
>>>> don't trigger it either.  This means I can't access the jabber  
>>>> online
>>>> play option you were mentioning.
>>>
>>> Right clicking is not necessary, if you just leave the mouse over  
>>> what
>>> you're trying to "right-click" on for a few seconds, it will have  
>>> the same
>>> effect.
>>>
>>>> When I browse the web, sites are enlarged to the wrong screen
>>>> resolution.  As a result, fonts are huge, and you can only see  
>>>> the top
>>>> left corner of a web site, even one that is at 800x600.  Is there a
>>>> way to control this in the program, or from outside of it?  Is it
>>>> doing this because I am on a Mac and using an emulator?  I attach a
>>>> screen shot of our Tiger's Mouth site so you can see the problem.
>>>
>>> I believe this is a problem with the emulator. Maybe this can help  
>>> you
>>> http://wiki.laptop.org/go/Emulating_the_XO/Quick_Start/Mac and more
>>> specifically this:
>>>
>>> http://wiki.laptop.org/go/Emulating_the_XO/Help_and_tips#Font_Size_correction
>>>
>>> If that doesn't work, you can at least reduce the font size in  
>>> Browse in
>>> the
>>> "View" Menu, or by pressing Ctrl+-.
>>>
>>> Cheers!
>>>
>>>> Thanks,
>>>>
>>>> Paul
>>>>
>>>> http://tigersmouth.org
>>>>
>>>> Paul Barchilon
>>>> 1005 36th St.
>>>> Boulder, CO 80303-2139
>>>> 303-440-7124
>>>>
>>>> On Oct 12, 2008, at 1:50 PM, Andrés Ambrois wrote:
>>>>>
>>>>> On Sunday 12 October 2008 17:28:15 Paul Barchilon wrote:
>>>>>>
>>>>>> Hi Walter,
>>>>>>
>>>>>> That was exactly the right question to ask!  I was trying the  
>>>>>> LiveCD
>>>>>> method with XO-LiveCD_080919.iso, which didn't work.  When I
>>>>>> created a
>>>>>> "new PC" in Qemu and told it to use olpc-redhat-stream-ship.2-
>>>>>> devel_ext3.img then it worked.
>>>>>>
>>>>>> Still trying to figure out how to get the screen size to the
>>>>>> appropriate dimensions, and now I have to figure out how to use  
>>>>>> your
>>>>>> OS.  From there I will load PlayGo and get back to all of you.
>>>>>> Thanks
>>>>>> for the help, I am glad I got in finally!
>>>>>
>>>>> Great! I'm glad you got it. To install PlayGo just head to
>>>>> http://wiki.laptop.org/go/PlayGo with the Browse activity and  
>>>>> click
>>>>> on the
>>>>> PlayGo.xo link in the box on the upper right. Then you can  
>>>>> launch it
>>>>> from the
>>>>> Journal or in the list view in the Home.
>>>>>
>>>>>> Here is another comment from the kid who is giving me feedback  
>>>>>> on his
>>>>>> laptop:
>>>>>>
>>>>>> From    leinad
>>>>>> Date    10/11/2008 23:39
>>>>>> Subject RE: one thing about playgo i can
>>>>>> i dont think it is online i think you have to be physically near
>>>>>> another XO...
>>>>>> and i do not have access to wi-fi that frequently...
>>>>>> also i noticed that the program does not take away dead stones...
>>>>>>
>>>>>>
>>>>>> I am not sure if he means it doesn't take dead stones off during
>>>>>> play,
>>>>>> or at the end, both are important though.  I will definitely  
>>>>>> want to
>>>>>> play a game with Andres on KGS, I think we could figure out all  
>>>>>> the
>>>>>> problems quite quickly that way.
>>>>>
>>>>> I'm not sure about what he means by that either. The game does  
>>>>> take
>>>>> away
>>>>> dead stones, and supports collaboration. To play it "online" you
>>>>> will need to
>>>>> connect to an online jabber server. If you have a recent version  
>>>>> of
>>>>> the
>>>>> system, this is done by right clicking on the "XO guy" in the home
>>>>> view ->
>>>>> Control Panel -> Network and typing: jabber.laptop.org in the
>>>>> "Server" field.
>>>>> Then, you will need to click on the "tick" mark and restart Sugar.
>>>>>
>>>>> Sure, I'm on KGS now as a guest, my handle is aambrois. Please go
>>>>> easy on me
>>>>> =P.
>>>>>
>>>>>> Thanks,
>>>>>>
>>>>>> Paul
>>>>>
>>>>> Thank you!
>>>>>
>>>>>> http://barchilonceramics.com
>>>>>>
>>>>>> Paul Barchilon
>>>>>> 1005 36th St.
>>>>>> Boulder, CO 80303-2139
>>>>>> 303-440-7124
>>>>>
>>>>> --
>>>>> -Andrés
>>>
>>> --
>>> -Andrés
>>
>>
>> _______________________________________________
>> Games mailing list
>> Games at lists.laptop.org
>> http://lists.laptop.org/listinfo/games
>>
>>
>
>
>
> -- 
> Walter Bender
> Sugar Labs
> http://www.sugarlabs.org
> _______________________________________________
> Games mailing list
> Games at lists.laptop.org
> http://lists.laptop.org/listinfo/games

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.laptop.org/pipermail/games/attachments/20081120/44721df5/attachment-0001.htm 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: AGFsig.jpg
Type: image/jpeg
Size: 4032 bytes
Desc: not available
Url : http://lists.laptop.org/pipermail/games/attachments/20081120/44721df5/attachment-0001.jpg 


More information about the Games mailing list