Needed for next ACCA event

Discussion of chess software programming and technical issues.

Moderator: Ras

bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: Needed for next ACCA event

Post by bob »

hgm wrote:Well, creating N new accounts is a proven solution that will certainly work. We know ICS are designed to handle thousands of accounts and allow many hundreds of them to be logged in at any time.

Of course if you are willing to develop something to do this with a single bot it would be more than welcome. Beware that it is not only important how many games you can observe (and 20 might not be enough), but also how many handles you can follow simultaneously.
I am pretty certain you can't follow every tournament participant. Several years back you could only follow 1. This was changed to 2 at some point in time. I will log in later today to see if it has been increased beyond that.

That said, I think I would still prefer a single application. It is not difficult to create N sockets and connect them to ICC using a different account for each. There there is just one program to set up, and since we have a fairly frequent ACCA/CCT event, it would probably be preferred to not have someone starting N copies of something to monitor N games...
User avatar
hgm
Posts: 28387
Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller

Re: Needed for next ACCA event

Post by hgm »

OK, so in any case we would need N accounts.

N XBoards could of course be started using a script. But I agree that a single application is conceptualy simpler. But that would be dependent on you making that application. All I wanted to point out is that there already exists a solution today, albeit a more cumbersome one.

I hope we can agree on a common kibitz format that will be recognized by the PGN collection bot and wil be sent by XBoard. Currently XBoard kibitzes the final Thinking Output line starting with

!!! <score>/<depth> <other stuff>

where the format of the "other stuff" is not really critical, but the !!! <score>/<depth> is used for recognition of info to include in the PGN.
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: Needed for next ACCA event

Post by bob »

hgm wrote:OK, so in any case we would need N accounts.

N XBoards could of course be started using a script. But I agree that a single application is conceptualy simpler. But that would be dependent on you making that application. All I wanted to point out is that there already exists a solution today, albeit a more cumbersome one.

I hope we can agree on a common kibitz format that will be recognized by the PGN collection bot and wil be sent by XBoard. Currently XBoard kibitzes the final Thinking Output line starting with

!!! <score>/<depth> <other stuff>

where the format of the "other stuff" is not really critical, but the !!! <score>/<depth> is used for recognition of info to include in the PGN.
I'm more than agreeable to having a common kibitz format. I've never liked adding "score=" or such, but since everyone kibitzes in a different way, I did that to avoid the usual questions on ICC. If everyone were to go to a uniform format, the questions should go away just as cleanly, except for the "score" meaning, since there is still the WPOV, PPOV issue...
CRoberson
Posts: 2094
Joined: Mon Mar 13, 2006 2:31 am
Location: North Carolina, USA

Re: Needed for next ACCA event

Post by CRoberson »

hgm wrote:OK, so in any case we would need N accounts.

N XBoards could of course be started using a script. But I agree that a single application is conceptualy simpler. But that would be dependent on you making that application. All I wanted to point out is that there already exists a solution today, albeit a more cumbersome one.

I hope we can agree on a common kibitz format that will be recognized by the PGN collection bot and wil be sent by XBoard. Currently XBoard kibitzes the final Thinking Output line starting with

!!! <score>/<depth> <other stuff>

where the format of the "other stuff" is not really critical, but the !!! <score>/<depth> is used for recognition of info to include in the PGN.
ICC will allow you to observe multiple games. It will not allow you to follow multiple users. That is fine with me because you only need to
follow half the participants of the tournament and that list changes with every round. So, observe seems the way to go which means I don't need
multiple accounts.

Also, I don't want to start up 20 xboard sessions to do this for a tournament with 40 participants. It is easy to see that solution doesn't scale. Too much work.
Software is suppose to reduce work load not increase it. I think starting up one session and observing 20 via the one session is less work and a cleaner solution.
User avatar
michiguel
Posts: 6401
Joined: Thu Mar 09, 2006 8:30 pm
Location: Chicago, Illinois, USA

Re: Needed for next ACCA event

Post by michiguel »

CRoberson wrote:
hgm wrote:OK, so in any case we would need N accounts.

N XBoards could of course be started using a script. But I agree that a single application is conceptualy simpler. But that would be dependent on you making that application. All I wanted to point out is that there already exists a solution today, albeit a more cumbersome one.

I hope we can agree on a common kibitz format that will be recognized by the PGN collection bot and wil be sent by XBoard. Currently XBoard kibitzes the final Thinking Output line starting with

!!! <score>/<depth> <other stuff>

where the format of the "other stuff" is not really critical, but the !!! <score>/<depth> is used for recognition of info to include in the PGN.
ICC will allow you to observe multiple games. It will not allow you to follow multiple users. That is fine with me because you only need to
follow half the participants of the tournament and that list changes with every round. So, observe seems the way to go which means I don't need
multiple accounts.

Also, I don't want to start up 20 xboard sessions to do this for a tournament with 40 participants. It is easy to see that solution doesn't scale. Too much work.
Software is suppose to reduce work load not increase it. I think starting up one session and observing 20 via the one session is less work and a cleaner solution.
But you can only observe games that already started. You will have to wait for the game to start and quickly type observe <participant>. You will also miss the beginning. I really wonder whether this is a real solution.

Miguel
CRoberson
Posts: 2094
Joined: Mon Mar 13, 2006 2:31 am
Location: North Carolina, USA

Re: Needed for next ACCA event

Post by CRoberson »

bob wrote:
hgm wrote:OK, so in any case we would need N accounts.

N XBoards could of course be started using a script. But I agree that a single application is conceptualy simpler. But that would be dependent on you making that application. All I wanted to point out is that there already exists a solution today, albeit a more cumbersome one.

I hope we can agree on a common kibitz format that will be recognized by the PGN collection bot and wil be sent by XBoard. Currently XBoard kibitzes the final Thinking Output line starting with

!!! <score>/<depth> <other stuff>

where the format of the "other stuff" is not really critical, but the !!! <score>/<depth> is used for recognition of info to include in the PGN.
I'm more than agreeable to having a common kibitz format. I've never liked adding "score=" or such, but since everyone kibitzes in a different way, I did that to avoid the usual questions on ICC. If everyone were to go to a uniform format, the questions should go away just as cleanly, except for the "score" meaning, since there is still the WPOV, PPOV issue...
I like the idea of a common kibitz format for the same reasons. However, it has problems. Some of the people are using interfaces that
force a kibitz format on them. If we use a common format, it would have to be adopted by those interfaces or we find common ground with those interfaces.
The latter may be the best solution.

I don't see why we need a special prefix such as !!!. We get a prompt that says "Telepath kibitzs: ...." already. I am speaking from
the collection point of view.
User avatar
Zach Wegner
Posts: 1922
Joined: Thu Mar 09, 2006 12:51 am
Location: Earth

Re: Needed for next ACCA event

Post by Zach Wegner »

It should be possible if the solution is a dedicated telnet client. Just watch for the tomato announcements and observe instantly once a game starts. That's not quite perfect though, since it's still possible to lose book move kibitzes at the very beginning. Following is really the best way there.

The problem with the xboard solution as I see it:
--requires a standard format that most people have not implemented. I think history shows that a requirement like this is going to be broken by at least a few people. It would be very easy for this requirement to be overlooked by those that don't frequent this forum.
--requires a standard format that I don't like. This isn't really a big deal, but I don't like the way the standard format looks. I wrote my kibitz format primarily for myself; it looks nice to me and provides lots of information. With non-standardized formats it's also much easier to tell which engine is kibitzing.
--people that use other interfaces like Arena and ChessPartner, especially under UCI, are pretty much screwed, since they can't kibitz in the standard way.
--the xboard solution is lossy. It only saves the score and the depth, but many programs, like mine, output much more information. A PV is essential information at the least, and ideally everything else that an engine outputs should be saved.

So perhaps the best solution would be to set up a bunch of accounts to follow each participant (I'm not sure if guests can follow... if not, ICC could set something up I guess), and then have xboard save all of the kibitz information, regardless of format, and leave it as comments in the PGN. That should be easy enough to do with already existing code. This duplicates each game, but a bit of redundancy won't hurt.
User avatar
hgm
Posts: 28387
Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller

Re: Needed for next ACCA event

Post by hgm »

CRoberson wrote: I don't see why we need a special prefix such as !!!. We get a prompt that says "Telepath kibitzs: ...." already. I am speaking from the collection point of view.
The !!! was added to the format to identify the line that belongs to the actual move played. Some engines kibitz multiple iterations per move. Some kibitz before the move, some kibitz after it. Some kibitz while pondering. It also reduces the confusion with non-compliant formats that happen to look similar, like those starting with depth/score in stead of score/depth.

Even from the collection POV some of these point are relevant. Unless you want to bloat the PGN excessively by including the kibitz for all iterations, and don't care if the kibitz messages end up with different moves as they were for.

Putting marker in front of the captured kibitz lines also offers more freedom to users that feel the sstandard kibitz requirements are too restricting, and want to kibitz more or more often. They can do that now without any chance of confusing the collection process, by simply leaving off the marker.

I don't think that white or own POV matters much in this case. The score info is not used for anything (like adjudication) and people looking at the PGN later will know how the participating engines score their moves. The de-facto standard for reporting score in WinBoard engines during thinking (as opposed to pondering) is own POV.
CRoberson
Posts: 2094
Joined: Mon Mar 13, 2006 2:31 am
Location: North Carolina, USA

Re: Needed for next ACCA event

Post by CRoberson »

Zach Wegner wrote:It should be possible if the solution is a dedicated telnet client. Just watch for the tomato announcements and observe instantly once a game starts. That's not quite perfect though, since it's still possible to lose book move kibitzes at the very beginning. Following is really the best way there.

The problem with the xboard solution as I see it:
--requires a standard format that most people have not implemented. I think history shows that a requirement like this is going to be broken by at least a few people. It would be very easy for this requirement to be overlooked by those that don't frequent this forum.
--requires a standard format that I don't like. This isn't really a big deal, but I don't like the way the standard format looks. I wrote my kibitz format primarily for myself; it looks nice to me and provides lots of information. With non-standardized formats it's also much easier to tell which engine is kibitzing.
--people that use other interfaces like Arena and ChessPartner, especially under UCI, are pretty much screwed, since they can't kibitz in the standard way.
--the xboard solution is lossy. It only saves the score and the depth, but many programs, like mine, output much more information. A PV is essential information at the least, and ideally everything else that an engine outputs should be saved.

So perhaps the best solution would be to set up a bunch of accounts to follow each participant (I'm not sure if guests can follow... if not, ICC could set something up I guess), and then have xboard save all of the kibitz information, regardless of format, and leave it as comments in the PGN. That should be easy enough to do with already existing code. This duplicates each game, but a bit of redundancy won't hurt.

Yes, moding the format to a strict standard screws all the UCI folks using an interface that doesn't comply.

Guests can watch tournament games as long as the games are unrated which is why our games are unrated.
I tested that myself during the last tournament.
CRoberson
Posts: 2094
Joined: Mon Mar 13, 2006 2:31 am
Location: North Carolina, USA

Re: Needed for next ACCA event

Post by CRoberson »

hgm wrote:
CRoberson wrote: I don't see why we need a special prefix such as !!!. We get a prompt that says "Telepath kibitzs: ...." already. I am speaking from the collection point of view.
The !!! was added to the format to identify the line that belongs to the actual move played. Some engines kibitz multiple iterations per move. Some kibitz before the move, some kibitz after it. Some kibitz while pondering. It also reduces the confusion with non-compliant formats that happen to look similar, like those starting with depth/score in stead of score/depth.

Even from the collection POV some of these point are relevant. Unless you want to bloat the PGN excessively by including the kibitz for all iterations, and don't care if the kibitz messages end up with different moves as they were for.

Putting marker in front of the captured kibitz lines also offers more freedom to users that feel the sstandard kibitz requirements are too restricting, and want to kibitz more or more often. They can do that now without any chance of confusing the collection process, by simply leaving off the marker.

I don't think that white or own POV matters much in this case. The score info is not used for anything (like adjudication) and people looking at the PGN later will know how the participating engines score their moves. The de-facto standard for reporting score in WinBoard engines during thinking (as opposed to pondering) is own POV.
I want to collect everything from the engines not just the last kibitz for a given move.