Re: [computer-go] New CGOS - need your thoughts.
I'm voting for 2 time settings: One normal and one fast (so maybe 5 min and 1 min on 9x9). -- GRATIS für alle GMX-Mitglieder: Die maxdome Movie-FLAT! Jetzt freischalten unter http://portal.gmx.net/de/go/maxdome01 ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] New CGOS - need your thoughts.
From what I can see, there is resistance to this idea - so what I'm going to do is to provide venues which are standalone but makes it possible later to add a time control.In other words for now there will be only 1 time control per board size but the server will be flexible enough that other venues can be added if the server ever gets popular enough that we have 40 or 50 players always on line. But they will be separate venues scheduled independently. - Don On Tue, Jun 16, 2009 at 8:08 AM, Isaac Deutsch i...@gmx.ch wrote: I'm voting for 2 time settings: One normal and one fast (so maybe 5 min and 1 min on 9x9). -- GRATIS für alle GMX-Mitglieder: Die maxdome Movie-FLAT! Jetzt freischalten unter http://portal.gmx.net/de/go/maxdome01 ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] New CGOS - need your thoughts.
Maybe we could agree that 1 day out of 7 in a week would be played on 6 times faster time controls. The same bots, connections, logins, the same number of games per week. Different rating of course. This would be a problem only for hardcoded bots with no time control. The advantage would be that we would see how different algorithms (bots) scale. If the ratings would be very similar for most bots, it would mean that we can get faster testing of new ideas. We would know which ideas can be tested of fast time control. Lukasz 2009/6/16 Don Dailey dailey@gmail.com: From what I can see, there is resistance to this idea - so what I'm going to do is to provide venues which are standalone but makes it possible later to add a time control. In other words for now there will be only 1 time control per board size but the server will be flexible enough that other venues can be added if the server ever gets popular enough that we have 40 or 50 players always on line. But they will be separate venues scheduled independently. - Don On Tue, Jun 16, 2009 at 8:08 AM, Isaac Deutsch i...@gmx.ch wrote: I'm voting for 2 time settings: One normal and one fast (so maybe 5 min and 1 min on 9x9). -- GRATIS für alle GMX-Mitglieder: Die maxdome Movie-FLAT! Jetzt freischalten unter http://portal.gmx.net/de/go/maxdome01 ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] New CGOS - need your thoughts.
I vote for 2 venues, each optional. Separate rating pools is a must. Łukasz Lew wrote: Maybe we could agree that 1 day out of 7 in a week would be played on 6 times faster time controls. The same bots, connections, logins, the same number of games per week. Different rating of course. This would be a problem only for hardcoded bots with no time control. The advantage would be that we would see how different algorithms (bots) scale. If the ratings would be very similar for most bots, it would mean that we can get faster testing of new ideas. We would know which ideas can be tested of fast time control. Lukasz 2009/6/16 Don Dailey dailey@gmail.com: From what I can see, there is resistance to this idea - so what I'm going to do is to provide venues which are standalone but makes it possible later to add a time control.In other words for now there will be only 1 time control per board size but the server will be flexible enough that other venues can be added if the server ever gets popular enough that we have 40 or 50 players always on line. But they will be separate venues scheduled independently. - Don On Tue, Jun 16, 2009 at 8:08 AM, Isaac Deutsch i...@gmx.ch wrote: I'm voting for 2 time settings: One normal and one fast (so maybe 5 min and 1 min on 9x9). -- GRATIS für alle GMX-Mitglieder: Die maxdome Movie-FLAT! Jetzt freischalten unter http://portal.gmx.net/de/go/maxdome01 ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] New CGOS - need your thoughts.
Whatever the eventual decision is - personally I would love a fast-play venue as an alternative, with separate rating - please don't worry too much about engines with fixed playouts, or engines that cannot handle certain time limits. The GTP client sitting between the engine and server will be able to protect the engine, by either keeping it out of games it cannot support or issuing it with reconfiguration commands. Christian Michael Williams wrote: I vote for 2 venues, each optional. Separate rating pools is a must. Łukasz Lew wrote: Maybe we could agree that 1 day out of 7 in a week would be played on 6 times faster time controls. The same bots, connections, logins, the same number of games per week. Different rating of course. This would be a problem only for hardcoded bots with no time control. The advantage would be that we would see how different algorithms (bots) scale. If the ratings would be very similar for most bots, it would mean that we can get faster testing of new ideas. We would know which ideas can be tested of fast time control. Lukasz 2009/6/16 Don Dailey dailey@gmail.com: From what I can see, there is resistance to this idea - so what I'm going to do is to provide venues which are standalone but makes it possible later to add a time control.In other words for now there will be only 1 time control per board size but the server will be flexible enough that other venues can be added if the server ever gets popular enough that we have 40 or 50 players always on line. But they will be separate venues scheduled independently. - Don On Tue, Jun 16, 2009 at 8:08 AM, Isaac Deutsch i...@gmx.ch wrote: I'm voting for 2 time settings: One normal and one fast (so maybe 5 min and 1 min on 9x9). -- GRATIS für alle GMX-Mitglieder: Die maxdome Movie-FLAT! Jetzt freischalten unter http://portal.gmx.net/de/go/maxdome01 ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ -- Christian Nentwich Director, Model Two Zero Ltd. +44-(0)7747-061302 http://www.modeltwozero.com ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] New CGOS - need your thoughts.
On Mon, 15 Jun 2009, Brian Sheppard wrote: Please don't do anything that decreases the frequency of games in order to accommodate programs that want to play on multiple venues. Keep venues strictly separate. Programs that want to play on multiple venues can just log in multiple times. I second that opinion. If there is a second venue, I'd prefer longer time controls. Christoph ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] New CGOS - need your thoughts.
I agree with David. Have one time control per board size. I like the 5-minute controls for 9x9. You can take your program down for extensive offline testing and still get 100 games per day. That is far more data than you can analyze. Still, the speed is fast enough for ratings to stabilize in a day or two. The argument for longer time controls is that it encourages the development of new algorithms. New algorithms are usually slower. It might take 10 man hours to quickly code up a new idea. Sure we can optimize it to run 10 times quicker, but that takes another 90 man hours. We want to see how well an idea works before spending all that effort. If the same bot plays both 5 minute games and 20 minute games, and the server can show different ratings for each venue, it would show how well that program's algorithms scale. Darren P.S. I'm not competing any time soon, so this cannot really count as a vote for longer time controls. -- Darren Cook, Software Researcher/Developer http://dcook.org/gobet/ (Shodan Go Bet - who will win?) http://dcook.org/mlsn/ (Multilingual open source semantic network) http://dcook.org/work/ (About me and my work) http://dcook.org/blogs.html (My blogs and articles) ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] New CGOS - need your thoughts.
On Mon, Jun 15, 2009 at 9:43 PM, Jason House jason.james.ho...@gmail.comwrote: Given all the negative reaction to nested time control, I have to say I like it. The pool won't be diluted as long as there's an obvious main venue. A good compromise might be to have only 2 venues, one such as David suggested and another one that is quite a bit faster. Another possibility is to make BOTH venues mandatory - but my fear is that some programs may not be able to play fast enough and would always time out.Or they may not implement a proper time control algorithm and thus would not be able to adapt to 2 different time controls without being reinitialized with different parameters. Sent from my iPhone On Jun 15, 2009, at 7:20 PM, Don Dailey dailey@gmail.com wrote: I've been working on the new server and I'm almost at the point where I can think about time controls - and since this is primarily for developers, I would like to get your thoughts. First, a brief explanation of how the time control works. When the client starts up it will inform the server of which venues it is willing to play in. It must choose an available boardsize and then any of N different time controls. Initially, N will probably be 2 or 3. For each board size, a time control is called a venue. Let's assume there are 3 venues for boardsize 9x9. The time control for each venue will be significantly different from the others. One will be very fast, one will be very slow and there will be one in between. Each time control will be in sync with the others and the process will be recursive. So the basic scheduling algorithm is to NOT start a new round for a given venue until any players who have registered to play in this venue and are currently playing in FASTER venues are available for scheduling. In addition to this, new rounds are not scheduled for any particular venue as long as the next slower venue is stalled waiting for these faster venues to complete. I hope this idea allows more choice and keeps players busy a greater percentage of the time by allowing them to fill dead space with fast games. Each bot can choose which venues to play in. If you only want to play fast games, then you can. Now the questions I pose to you are these: How many venues for each boardsize? (two, three, more?) What time controls should they be? It's almost certainly the case that certain combinations of time control venues will work together better than others. There will always be the issue of waiting for games to complete and in fact this may make the problem a bit worse for those programs that only want to play in the longest venue. I suggest that each venue is spaced at least a factor of 2 apart in time. For instance 1 minute, 2 minutes, 4 minutes, etc. My own suggestion for 9x9 is to have 3 venues of 1 minute, 5 minutes and 15 minutes per game per player. It's also not too late to change our minds and not have venues if we think the disadvantages outweigh the advantages. - Don ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] New CGOS - need your thoughts.
I'll express my opinion here, but keep in mind that my engine (cogito) has only played 44 games as of now on CGOS. I have a few problems with separate time controls. --It dilutes the rating pool. If there is only one time control, everyone can play everyone. If there are separate time controls, then there will probably be some players that only play in certain time controls. Thus the rating pools must be kept separate to not introduce bias. Separate rating pools reduce the amount of useful data available. --There are better ways to accomplish the same goals. As you suggested, you could simply wait until half of all players are idle, and then start another round. You could take this even further. I suppose in the current CGOS you have some measurement on whether two players would make an acceptable match in the next round? Whenever a player becomes idle, and there is at least one other player idle, try and match them if they are close enough. There wouldn't be any more rounds, but I think this would be a better solution. --It introduces complexity. Some players (like mine) don't have any time control code. Mine has to be recompiled to play at different numbers of playouts. This is because it's an ultra minimal engine, only 1937 characters of C at the moment (by IOCCC counting rules). It plays on CGOS using an adapter shell script. I'd rather not have to rewrite that! There are plenty of players that play at fixed playout counts. Zach ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] New CGOS - need your thoughts.
I'm for keeping the number of pools small, to keep their sizes large. Peter Drake http://www.lclark.edu/~drake/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] New CGOS - need your thoughts.
On Mon, Jun 15, 2009 at 11:18 PM, Zach Wegner zweg...@gmail.com wrote: I'll express my opinion here, but keep in mind that my engine (cogito) has only played 44 games as of now on CGOS. I have a few problems with separate time controls. --It dilutes the rating pool. If there is only one time control, everyone can play everyone. If there are separate time controls, then there will probably be some players that only play in certain time controls. Thus the rating pools must be kept separate to not introduce bias. Separate rating pools reduce the amount of useful data available. This would not dilute anything if everyone played in all venues, it would only increase the total amount of data. Of course if players opt out of certain levels it would dilute it somewhat. Separate ratings is something I definitely planned on doing. --There are better ways to accomplish the same goals. As you suggested, you could simply wait until half of all players are idle, and then start another round. You could take this even further. I suppose in the current CGOS you have some measurement on whether two players would make an acceptable match in the next round? Whenever a player becomes idle, and there is at least one other player idle, try and match them if they are close enough. There wouldn't be any more rounds, but I think this would be a better solution. Actually an early version of CGOS did this and it does not work so well. What happens is that when the first match is over, BOTH of those players are now available and they play each other again and this never ends. So I had this hack to prevent consecutive matches against the same players. It helped a little but the real problem is that you need a lot of players available to prevent lots of mismatches while still maintaining diversity. CGOS wants to give you many different players to play, while not giving you too many ridiculous mismatches. --It introduces complexity. Some players (like mine) don't have any time control code. Mine has to be recompiled to play at different numbers of playouts. This is because it's an ultra minimal engine, only 1937 characters of C at the moment (by IOCCC counting rules). It plays on CGOS using an adapter shell script. I'd rather not have to rewrite that! There are plenty of players that play at fixed playout counts. Yes, I agree complexity is an issue. I can handle this in the server but I want it to be easy to make a simple bot that works. - Don Zach ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
RE: [computer-go] New CGOS - need your thoughts.
If more than one venue is mandatory I probably won't be able to join, since I want to spend my limited programming time making the engine stronger, not programming multiple time controls. Please allow me to play with just a singe time limit without changing my cgos interface code. David From: computer-go-boun...@computer-go.org [mailto:computer-go-boun...@computer-go.org] On Behalf Of Don Dailey Sent: Monday, June 15, 2009 7:02 PM To: computer-go Subject: Re: [computer-go] New CGOS - need your thoughts. On Mon, Jun 15, 2009 at 9:43 PM, Jason House jason.james.ho...@gmail.com wrote: Given all the negative reaction to nested time control, I have to say I like it. The pool won't be diluted as long as there's an obvious main venue. A good compromise might be to have only 2 venues, one such as David suggested and another one that is quite a bit faster. Another possibility is to make BOTH venues mandatory - but my fear is that some programs may not be able to play fast enough and would always time out. Or they may not implement a proper time control algorithm and thus would not be able to adapt to 2 different time controls without being reinitialized with different parameters. Sent from my iPhone On Jun 15, 2009, at 7:20 PM, Don Dailey dailey@gmail.com wrote: I've been working on the new server and I'm almost at the point where I can think about time controls - and since this is primarily for developers, I would like to get your thoughts. First, a brief explanation of how the time control works. When the client starts up it will inform the server of which venues it is willing to play in. It must choose an available boardsize and then any of N different time controls. Initially, N will probably be 2 or 3. For each board size, a time control is called a venue. Let's assume there are 3 venues for boardsize 9x9. The time control for each venue will be significantly different from the others. One will be very fast, one will be very slow and there will be one in between. Each time control will be in sync with the others and the process will be recursive. So the basic scheduling algorithm is to NOT start a new round for a given venue until any players who have registered to play in this venue and are currently playing in FASTER venues are available for scheduling. In addition to this, new rounds are not scheduled for any particular venue as long as the next slower venue is stalled waiting for these faster venues to complete. I hope this idea allows more choice and keeps players busy a greater percentage of the time by allowing them to fill dead space with fast games. Each bot can choose which venues to play in. If you only want to play fast games, then you can. Now the questions I pose to you are these: How many venues for each boardsize? (two, three, more?) What time controls should they be? It's almost certainly the case that certain combinations of time control venues will work together better than others. There will always be the issue of waiting for games to complete and in fact this may make the problem a bit worse for those programs that only want to play in the longest venue. I suggest that each venue is spaced at least a factor of 2 apart in time. For instance 1 minute, 2 minutes, 4 minutes, etc. My own suggestion for 9x9 is to have 3 venues of 1 minute, 5 minutes and 15 minutes per game per player. It's also not too late to change our minds and not have venues if we think the disadvantages outweigh the advantages. - Don ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
