Arimaa Forum - Print Page


    
      
        Arimaa Forum
        (http://arimaa.com/arimaa/forum/cgi/YaBB.cgi)
      

        Arimaa >> Events >> 2005 World Championship
        
(Message started by: MrBrain on Oct 1^st, 2004, 7:44am)

Title: 2005 World Championship
Post by MrBrain on Oct 1^st, 2004, 7:44am

I'm extremely disappointed that we went with an even faster time control than the one we used last year, which was already too fast. I really thought that if we could have a system that would better match up times when people could play, that we could have gone to something reasonable, like 2 minutes a move. I guess I'll sign up for the tournament, but at 1:15 per move, I don't have a realistic chance of playing a good game. Even the best players won't be able to put together quality games at that speed. Do we forget already that the last game of the championship was decided by one player running out of time? :(

Title: Re: 2005 World Championship
Post by 99of9 on Oct 1^st, 2004, 10:40am

on 10/01/04 at 07:44:51, MrBrain wrote:

I'm also slightly surprised that the human time controls for the finals are not 2 minutes, because then they would be the same as the bots will play at in the bot tournament.

on 10/01/04 at 07:44:51, MrBrain wrote:

Even the best players won't be able to put together quality games at that speed. Do we forget already that the last game of the championship was decided by one player running out of time? :(

Well, I would dispute this. Currently most games between top players are played at even faster time controls than 1:15. This is because games between top players often go for more moves than games between more erratic players, but they have similarly limited time in their lives. In the game you refer to, if time limits had been 3 min per move, and players had used their entire time, the game would probably have gone for almost 8 hours (it went to 66 moves, and may well have taken another 14). Three games of that is too much to commit to.

Perhaps a compromise that would help everyone would be to put some time in the reserve at the start of the game, and make a larger reserve. Then normal moves would still be played at a reasonable pace, but players who need to think once in a while would have that option.

I would support 1m15s/10m/100/10/15m

Title: Re: 2005 World Championship
Post by Fritzlein on Oct 1^st, 2004, 11:38am

I agree that there is an issue here, but there is quite a bit of gray area. What is a "quality game"? Clearly, the longer players have to think, the better they will play objectively. Someone who likes to play at four minutes per move could well claim that speeding up to two minutes per move degrades the quality of the game, but then a postal player might say that games played at four minutes per move display an acute lack of planning and are full of outright errors.

A more important issue of "quality game" is the enjoyment of a game as a spectator sport. Is it fun to watch? I know as a fan of chess that games played at two and a half minutes per move are very taxing on the patience of the spectators. In sports in general, it is fun to watch both sides straining to their utmost to win, even if there are mistakes. And the best time control for spectators is probably between 30 seconds and 60 seconds per move.

I think there must be a balance. On the one hand the games shouldn't merely be about who can avoid a blunder. For interest's sake there should be identifiable plans and strategies. On the other hand, the goal isn't merely mistake-free Arimaa, it is interesting Arimaa to watch, and that requires not going too slowly.

Title: Re: 2005 World Championshipt
Post by omar on Oct 1^st, 2004, 12:19pm

Personally as a player I perfer longer time controls also. It gives me a chance to be more relaxed and play a better game. But there are other considerations besides what's best for the players.

People have time constraints and our players are distributed all over the world, so trying to keep the games within a resonable time where more people can fit it into their schedual to play and watch is very important.

In an absolute sense the quality of a game is better with longer time controls, but in a relative sense the quality of a game at any time control is still good for that given time control. If the tournament used postal time controls the absolute quality of the games might be even better and there might be a different person who wins the championship at that time control than at the interactive but slow time control. And at a faster interactive time contorl a third person might win. So the actual time control does not matter too much; someone or other will be good at what ever time contorl we chose.

Now one can also get very concerned about the probability of the same person winning the tournament again and that probability going down as the time control gets faster. That is, are the tournament results repeatable. But I don't like to get too concerened about that because even with very long interactive time contols I dont think the probability would be that much better (the format of the tournament effects this probability and we would have to have tournaments with many more games; like a round robin; to make make major improvements).

I would much rather have more people being able to play and watch the World Championship games than sacrificing that for things like quality of the games and repeatability. The place where these things matter more (well at least to me) is in the Arimaa challenge match; thus the longer time controls there :-)

If you practice for your matches by playing games with a faster time control than what is used in the tournament, you will not have any problems.

Title: Re: 2005 World Championship
Post by MrBrain on Oct 1^st, 2004, 2:05pm

"If you practice for your matches by playing games with a faster time control than what is used in the tournament, you will not have any problem"

No, this is not the case for me. I know from experience that I am a deep thinker, and not a fast thinker. I could practice all day at a faster time control, and I'll still make blunders at the faster time control. Arimaa is a game of strategy. And as such, it should have a longer time control. At shorter time controls, tactics takes over, and strategy goes out the window. This is a huge shame.

Title: Re: 2005 World Championship
Post by 99of9 on Oct 1^st, 2004, 2:42pm

on 10/01/04 at 14:05:09, MrBrain wrote:

No, this is not the case for me. I know from experience that I am a deep thinker, and not a fast thinker. I could practice all day at a faster time control, and I'll still make blunders at the faster time control.

I think it would be worth trying in arimaa. You have a limited record against anyone but bot_shallowblue, so I can't say anything conclusive from your record - but you have beaten speedy 1-0, and lost to bomb 0-2. Since these are the same algorithm, this possibly implies that you are _better_ at short time controls!?

You would be surprised how much strategy can fit in even a 30-45 second time control. When I have played these time controls, about one quarter of the games have been decided by tactical errors by one player or the other, an half have been decided by out-strategising, and another quarter have been decided by flag fall.

Title: Re: 2005 World Championship
Post by MrBrain on Oct 1^st, 2004, 2:58pm

No, I have played many games besides Arimaa, mainly chess. I do very poorly in quick time controls, and very well in long time controls. If the championship were in a long time control, I would have a realistic chance of winning against anyone on here, despite my limited record. With a short time control, I have none.

I feel very strongly that 2 minutes is an absolute lower-end limit on a reasonable time control for Arimaa. If the challenge match and the computer match are at 2 minutes, the human match should also be at 2 minutes.

Omar always states that the reason for the fast control is due to availability of players. Well, if that's a huge concern, then why isn't there a reasonable mechanism to choose a time that's convenient for both players? Last year's scheme was inadequate to find a time that's mutually convenient. I have no problem playing an 8-hour game, as long as it won't bump into anything else that I've committed to. Having to pick 30 slots out of 59 makes that impossible. It's a "World Championship", so it should be played at a rate that makes a very serious game that includes strategy possible.

Title: Re: 2005 World Championship
Post by MrBrain on Oct 1^st, 2004, 3:01pm

The other option is that if the human match is 1:15 per move, then the computer match and the challenge match should be at 1:15 per move. Pick a time and stick to it. Don't have one time control for one type of match, and a different one for computers. That makes no sense.

Title: Re: 2005 World Championship
Post by clauchau on Oct 1^st, 2004, 4:31pm

I very much like it at 1:15 (and 2:00 later on).

Title: Re: 2005 World Championship
Post by Fritzlein on Oct 2^nd, 2004, 3:00pm

I agree with MrBrain that playing well at a certain speed isn't merely a matter of practice. Some people naturally excel more at one time control than at another. But that doesn't really enter into the argument of what time control is best for the World Championship. Surely MrBrain isn't saying, "A longer time control is more appropriate because a longer time control gives me a better chance of winning."

The most serious issue to me is to avoid a time control that makes strategic planning impossible and/or unimportant relative to tactics. The nature of the game does change depending on the time contol. However, I do not think 75 seconds per move is so fast that tatics will dominate.

How have I been able to beat bot_speedy five games out of six at 30 seconds per move? Did I outduel it at tactics? Not at all. I don't think any of us can top bot_speedy in tactics at that lightning pace. Yet I won anyway. Clearly my victories were due to good strategy (or rather very poor strategy by speedy). At the much slower pace of 75 seconds per move, strategy becomes even more possible and more important relative to tactics.

After the fact, we will be able to judge from the games themselves whether the strategy took a back seat to tactics across the board. If I had to bet in advance, however, I would bet that the games will be more strategic than in last year's championship, despite the faster time control, simply because the level of play has generally risen. And indeed, it is partially the rising level of play that is adding more moves to the games, and increasing the pressure to have less time per move.

Title: Re: 2005 World Championship
Post by MrBrain on Oct 3^rd, 2004, 11:57am

I must admit that a significant part of my disappointment is the realization that I will not have a realistic chance of winning at these controls.

But my stronger protest is that there are different time controls for different games and different types of games (computers, humans). I just feel very strongly that there should be one official time control. This should be included in the Arimaa challenge itself. Otherwise, the challenge loses some meaningfulness. When the computers get better, can we then increase the time control to give the humans more time to develop strategy?

But anyway, I should apologize for being so negative in these posts. I wouldn't be so much so if I didn't feel so strongly and very positive about the game of Arimaa and what quality in games and strategy should be acceptable. I very much appreciate all the work that Omar and his family and friends have put into the creation of the game. Also the hard work and money put into offering events such as these championships and the challenges. Thank you.

Title: Re: 2005 World Championship
Post by Fritzlein on Oct 3^rd, 2004, 12:10pm

I agree that time control for the games of the Arimaa Challenge should be written into the rules of the challenge. It would indeed be silly to lengthen the time controls, once computers start getting close, in order to give humans a better chance. The only thing I might support is increasing the numbers of games in the match, so a fluke decision is less likely. But the time control ought to be set in stone to make the challenge as meaningful as possible.

I also am sympathetic to the argument that the time controls for the human World Championship should be the same as those for the Arimaa Challenge. If one function of the human championship is to pick a representative of the human race to beat up the computers, then why not use a similar format?

On the other hand, if I remember correctly, the human representative in the Arimaa Challenge match is not determined by the human World Championship. Those are two different things. Omar and Aamir explicitly have the right to pick whoever they want, with no reference to tournaments or ratings or anything. So in that sense the two events aren't really linked, even though they are close in time.

Title: Re: 2005 World Championship
Post by omar on Oct 5^th, 2004, 5:56am

Last year we did not have as much actual experience with the different time controls as we do now. So I had chosen a time control of 3/3/100/15/8 for the Arimaa challenge match just to make sure there would not be any need to rush. Having played the challenge match myself, I now think it was a bit too much. I was getting bored waiting for Bomb to move and had plenty of time to take snack breaks (I don't drink coffee) between moves. Anyone who has played Bomb or GnoBot version 2004CC will probably feel the same. This year I've reduced it down to 2/2/100/10/8.
Games at this time control will run for about 2 to 4 hours and a really long game could go for about 5 or 6 hours. I think this is a really good time control for offical Arimaa games where the players are able to commit this kind of time for the games. I plan on using this for all Arimaa challenge games and the computer championship games in the future. Im sure the computers have no complaints about it and there is only one human involved. I would also like to use this time control for the human championship games, but I just don't think that people can commit that much time. In the human championship we are dealing with lots of people and people in different time zones; so as a practical consideration we have to use a faster time control for this event so that most games can fit into about two hours. I don't mind having different time controls for different events. I think we have to flexible and adjust based on the needs of the event.

With faster time controls the thing I worry about most is that blunders will overtake both strategy and tactics. The chance of a player losing due to missing a goal threat or because they did not notice the clock increases dramatically. So there is probably some lower limit for the time controls before it becomes unacceptable.

The games we are playing in the EU vs US match will give us some idea of how fast we can make the time controls before blunders overtake strategy and tactics. The one minute per move time control for these games is a bit faster than what we will be using for the knock out games in the world championship. Of the 3 games played so far only my game with Claude ended on time, the other two seemed to go OK.

Playing games at a fast time control does help one learn to avoid blunders and watch the clock more carefully. But it can never subsitute for the higher quality stratigic play that is possible with slower time controls.

Someday if the Arimaa world championship games mean much more and the players don't mind commiting the time, we can go to the slower time control. But then if ESPN start to cover it we might have to use faster time contorls again :-)

Title: Re: 2005 World Championship
Post by MrBrain on Oct 5^th, 2004, 10:30am

For myself, I would really prefer a 2/2/100/10/8 time control for all future official matches, regardless of opponents. From your personal correspondence, I understand you've improved the scheduling mechanism. This should be sufficient to ensure that everyone will be able to finish games at this speed. Like you said, 2 to 4 hours is typical. Almost everyone could start a game at 8 pm at least one night of the week, for example. Then, even if a game goes very long (6 hours), it just becomes a late night. I think in today's world, most people are used to having a late night every now and then.

Title: Re: 2005 World Championship
Post by 99of9 on Oct 7^th, 2004, 7:36pm

on 10/05/04 at 10:30:52, MrBrain wrote:

Almost everyone could start a game at 8 pm at least one night of the week, for example. Then, even if a game goes very long (6 hours), it just becomes a late night. I think in today's world, most people are used to having a late night every now and then.

1) Two players would always be involved. Unless their one night a week coincidentally lined up, one night a week is useless.

2) Who's 8pm are you talking about? In case you haven't noticed, some of us do not live in the same timezone as you. Your 8pm is past my midnight, and that's when you want the game to START???

I think you'd be better off arguing about weekends - they are easily the best chance to line up schedules.

However, I agree strongly with Omar that lining up the schedule of two busy people in different timezones is _very_ difficult. You may remember that I forfeited my quarter final against you last year because I was unable to line up time for even one game of arimaa on one particular week, and that was without even trying to match it to your schedule. I'd suggest that too many forfeits is just as bad or worse than too many timeouts.

Title: Re: 2005 World Championship
Post by MrBrain on Oct 8^th, 2004, 8:59am

Well, all I'm saying is that hopefully the new scheduling mechanism will be much better than last year. I could have been unlucky with the one last year, and not been able to finish a game. By being able to rank times 1, 2, and 3, both players will undoubtably find a time that suits the both of them. The 8 pm thing was just an example. Not saying that it's going to be 8 pm for everyone, obviously. And last year, your situation was unique. You were out the entire week. In that case, no time control or scheduling mechanism would help.

Title: Re: 2005 World Championship
Post by 99of9 on Oct 8^th, 2004, 11:23am

on 10/08/04 at 08:59:23, MrBrain wrote:

By being able to rank times 1, 2, and 3, both players will undoubtably find a time that suits the both of them. The 8 pm thing was just an example. Not saying that it's going to be 8 pm for everyone, obviously. And last year, your situation was unique. You were out the entire week. In that case, no time control or scheduling mechanism would help.

I agree that the new scheduling system is better than the previous one. But I don't think it is SO much better that it suddenly opens up 4-6 hour slots in everyone's 2-person-combined timetables.

To me, your 8pm example implied that it was going to be easy to find a weeknight time at which both players would have 4-6 hours free, and that anyone unable to accept a 2am finish every now and then was not living in todays world. Am I correct about your implications?

[As it happens, yes, I personally am willing to live with a 2am finish.] My point in reply was not that timezones mean that 8pm is not universal. My point was that even if you can accept a 6 hour game that finishes at 2am, your European opponent would have to accept a 6 hour game that may finish at later than 7am. Not many people I know would be willing to accept that unless it was incredibly important, or they had nothing important on the next day. Disparate timezones mean that finding mutually acceptable times to play is much harder than your simplistic argument implied.

Just to be sure this is clear, let's just take the example of a Frenchman willing to stay up until 2am that has to play a New Yorker who is able to knock off his job at the not unreasonable time of 6pm, and immediately start arimaa. Let's try to line up a weeknight time for these people. They luckily turn out to have absolutely nothing on on weeknights... Note that 6pm New York = approx 11pm France. This kind of game clearly cannot go on for the hours you suggest for the European. If both are workers, there is no other alternative during the week (unless the US player wants to wake up at 1am, in which case they might sneak in a quick game before the European goes to work).

So for your time control, intercontinental games involving two weekday workers would all need to be played on the weekend.

You are right that my situation last year would not have been helped by a time control or scheduler. But it is not that unique, and it is simply the end of a busy-ness spectrum that most of us find ourselves on. I'm not saying that anything should have been done differently last year. Human constraints mean that I forfeit, and that was fine by me. My point was that even solving a one-body problem for 6 weeks in a row is not always trivial, a two-body problem is automatically much more difficult, so we have to understand Omar's compromises on game timing.

If we insist on 4ish hour games, then the argument I outlined earlier suggests that no arimaa player should be able to go for a weekend away during a 6 week period without forfeiting! I'd much rather get well-scheduled games set up and accept the fact that one or two players won't be able to handle the pace and may blunder.

Title: Re: 2005 World Championship
Post by MrBrain on Oct 8^th, 2004, 12:48pm

Well, you've clearly thought this through a lot better than I have. So I'll agree with you that scheduling is hard. But I still feel strongly that official games shouldn't be less than 2 minutes per move (I think even that's fast), regardless of the scheduling difficulties. If the games are worth playing at all, they should be played under reasonable time controls. It's a "World Championship". People should be willing to go to some lengths to play in a championship. Last year, most of the time slots I picked were in the middle of the night. It's not that I'd like to play in the middle of the night, but rather that the scheduling mechanism forced me to pick 30 slots.

Title: Re: 2005 World Championship
Post by 99of9 on Nov 16^th, 2004, 9:50am

Hi Omar,

The prediction contest for the WC is not allowing predictions for the first round. I know that's what your rules officially stated, but that was before we figured out such a neat prediction scoring method that even works for inequal games. I think it would be better to include these first two games.

Toby

Title: Re: 2005 World Championship
Post by omar on Nov 17^th, 2004, 6:28pm

Thanks for that reminder. I've changed it so that we can predict on all the games.

Title: Re: 2005 World Championship
Post by Fritzlein on Nov 17^th, 2004, 6:38pm

Thanks for changing it. With only 10 participants this year, there will only be 11 or 12 games total, so including the first two in the prediction contest makes a significant difference. I think the scores do balance out reasonably well for the players involved. On average the underdog will score fewer points, but if the underdog pulls off an upset, s/he gains a lot of points on the entire field, so it's basically a wash.