Re: [computer-go] how to have a computer go tournament? (summary)

[David G Doshay <ddoshay@xxxxxxx>]

I have a few more thoughts, embedded below:

On 25, Jan 2005, at 3:55 AM, Ray Tayek wrote:

> i have put together an outline of what i think many of you (from the 
> computer go mailing list) have said. (please see below). please attack 
> viciously, and point out anything you disagree with or anything i have 
> missed. please answer anything with a question mark if you can. once 
> the dust settles we can go offline and get out of your hair.
>
> thanks again for all of the good ideas and information.
>
> 0. 2005 cotsen open will be held on the weekend of April 30th.
Oh, much earlier this year ... I hope I'm ready!

> 1. a computer go tournament
>         1. types and costs
>                 1. world class $10k? (get most? of the top programs 
> 15?-25?)
This is lots more "motivation" than I need. But I already live in 
California.

>                 2. upper middle ground $6k (get some 8?-15?)
>                 3. lower middle ground $3k (get some 5?-10?)
>                 4. unofficial $1k (mostly for some travel & lodging on 
> an as needed basis for the authors and a small prize for 1st place, 
> get 3?-6? )
>                         1. authors of many faces, smart go, proably 
> gnugo, maybe others? will allow their programs to be used without them 
> assuming some sane moderation by a rational human.
>         2. setup
>                 1. moderate? hassles with equipment setup, ups?, 
> security
> 2. entering a program in the tournament
I would like to propose another option that I see as a middle ground: 
Have the computer competition before the human competition and let the 
top 2 programs enter the human competition. Hopefully human competitors 
will not object as loudly if the programs have already been culled down 
to just the best.

>         0. players receive a full refund of their entry fee if they 
> play all 5 games so the only cost to the player is travel/lodging.
>         1. lots of rules to hassle, as the aga has flavours of 
> computer involvement.
>         2. involves aga, ing, the td, sensei, etc.
>         3. aga appears to not be very interested in computer go
>         4. previous experience indicates that the players will revolt.
Perhaps, perhaps not. Another thing that can be done it to have a 
check-box on the entry form to find out those who would refuse to play 
against a computer and try to adjust accordingly. I do not know if you 
have an elimination tourney planned, but if a program makes the cut 
then it should not get a forced loss because a person refused to play 
it.

>         5. *might* work if you provided some $ incentive?
>         6. probably not worth the trouble now.
I would hope that the people would be more open minded.

> 3. players whose opponents don't show up play a program.
>         1. give then the win.
>         2. offer a small prize $10-$20 to play a program.
>         3. person who beats up the program the worst gets $100 (most 
> stones or earliest resign)
>         4. how often will this occur and in what band? (what are mf, 
> smart go's, and gnugo's approximate igs/nngs ratings? 5k? 10k?)
ratings near 8 seem common.

>         5. we have some real beginners, any reason not to let the 
> programs play absolute beginners (20kyu-30kyu)? (maybe they could play 
> on 13x13 boards? aga rules on this?).
I have no objection other than wanting to be able to enter the human 
competition if in the top 2.

>         6. game will be played on a real board, so requires an 
> operator and pc each time.
I agree that for any game that is part of a competition this is the 
easiest. We would have to determine what the rule is about 
mis-transfered moves (should an honest mistake be a forced resign or 
should take-backs be allowed ... one  ... or a small number?).

Even if we do get our pc referee ready by this early date, I think play 
through it should only be on a demonstration basis, just a way to keep 
some programs playing against each other without so much human 
intervention.

> 4. demo of computer go playing software
>         1. two pc's, cables, and a projector
>         2. minor setup and maintenance
>         3. security or humping required for equipment
I probably would have the worst problem on this item, and I am willing 
to drag my stuff around as needed.

> 5. offers of assistance (onsite)
>         1. probable (due to timing) from "Bob Myers" <rtm@xxxxxxxxxxxxxxxxx> of 
> http://intelligentgo.org/
>         2. probable (due to timing) "David Fotland" 
> <fotland@xxxxxxxxxxxxxxxxx> http://smart-games.com/ (many faces)
>         3. probable (due to timing) David G Doshay <ddoshay@xxxxxxxxxxxxxxxxx> 
> (game referee program
Again, I also have to say maybe to the pc referee part too, 
particularly at this early date.

> and *maybe* some way to play against his cluster?) 
> http://currents.ucsc.edu/04-05/07-26/cluster.html
 I will not be able to bring the cluster mentioned in that article, but 
if there is an internet connection we could play it remotely for those 
who trust me to be telling the truth about what is on the other end of 
the wire. I am working on getting a traveling cluster together to bring 
to LA, and it would be less powerful than the one in the machine room 
at UCSC.

But the main point in my offer was the possibility of assisting as a 
sponsor more than as an organizer, as long as nobody fears a conflict 
of interest if I am both a sponsor and a competitor.

>         4. possible? (due to timing) "Anders Kierulf" or a proxy 
> <anders@xxxxxxxxxxxxxxxxx> http://smartgo.com/
> 6. other
>         1. permission from the authors of many faces, smart go, and 
> (probably) gnugo to use their programs in an unofficial tournament 
> provided we have a sane human running it.

Cheers,
David


_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/