|
||||||||||||||
|
Title: Ideal time control proportions Post by Fritzlein on Dec 1st, 2005, 4:58pm As I was answering Medarch's question about time controls, it reminded me that the list of time controls we get to choose from is somewhat irregular. That is to say, we can choose how long per move, but we can't infer on the basis of move length how long the starting reserve, maximum reserve, and maximum game time will be. (I'll leave out the percentage of excess time transferred to reserve, because 100% is standard and good.) My hunch is that a standard formula would work very nicely in most cases, because it is the proportion of times that matters. A two-minute maximum reserve is perfect for a game at fifteen seconds per move, but very cramped for a game at two minutes per move. The latter offers a very small buffer compared to being marched along making every move at the same pace. I submit that a maximum reserve of eight times the per-move increment strikes an excellent balance between keeping the game moving at a steady pace and allowing the players some control over their own time management. With a built-up reserve, one can think twice as long as normal for eight moves in a row before being required to revert to a normal pace, but one never can slow the game to a crawl by thinking for twenty times the average move length on a single move, as happens at times in chess. If you need doubled think time for more than eight moves in a row, you don't need more control over your time management, you just need more total time. For starting reserve, the present time controls vary wildly from one extra move up to the maximum reserve. I suggest standardizing on a starting reserve of half the maximum reserve. This emphasizes good time management more than either extreme, because a full starting reserve removes any incentive to move quickly, but an empty starting reserve removes any possibility of thinking longer for a few moves. If your reserve starts out halfway between, you have the option of moving either faster or slower, and are therefore rewarded for having good judgement as to which is appropriate. Finally, the maximum game time really only comes into play when there is a bot that is winning by enough that a human can't beat it, but the bot is too stupid to convert its advantage. To take care of this case, a maximum game length of 240 times the per-move increment seems adequate. Admittedly, there might be a rare theoretical win that needs more than 120 moves to convert, but speaking as the only guy who has played out a full eight-hour game against a bot, believe me it is enough. Even so, for blitz games the 8-hour cutoff is silly: I certainly don't want to play on for 960 moves just because a bot is too dumb to finish me off. For blitz a one-hour cutoff suits me fine. To summarize, I propose that the standard time controls be 0:15/1/100/2/1 0:30/2/100/4/2 0:45/3/100/6/3 1/4/100/8/4 1:30/6/100/12/6 2/8/100/16/8 3/12/100/24/12 Noboby plays at four minutes per move, so that time control can be abandoned. Does this set of standard time controls sound reasonable? |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by 99of9 on Dec 1st, 2005, 6:27pm A few comments: 2/8/100/16/8 Be warned that if you make this change, the bots really will take 18 minutes to make a move sometimes (at least Gnobby will). This is guaranteed to drive both spectators and players barmy - especially if they are easily winning ;-). In general I think you have made all the time reserves too long. My optimum would be to cut about 25% off all the reserves you propose. 3/12/100/24/12 Surely no-one would like to play a 12 hour game, even if they were happy with 3 min per move!! Bot-Bot games often go for more than 120 moves. I know it's due to their stupidity, but I'd definitely prefer not to see games settled by time running out. Therefore I suggest NOT using these standard timeout controls in bot-bot tournaments or games. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by omar on Dec 3rd, 2005, 3:45pm on 12/01/05 at 16:58:24, Fritzlein wrote:
The current time controls were setup without much practical experience and did eventually have to be tuned. I like the idea of having a set of standard time controls; that makes it easy to be able to just say "a 30s per move game" and not have to specify the other parameters. I think we have a lot of practical experience now to be able to tune these better. I like the set of time controls proposed by Karl, but I also agree with Toby that some of the long max reserves might be a bit difficult for spectators to handle. Since we are often the players we usualy don't consider things from the spectators point of view. But in the long run the spectators decide the fate of any sport. So lets not forget about them. How about if we trim the max reserves a bit. Fast 0:15/1/100/2/1 -----> 0:15/1:00/100/2/1 0:30/2/100/4/2 -----> 0:30/1:30/100/3/2 Medium 0:45/3/100/6/3 -----> 0:45/2:00/100/4/3 1:00/4/100/8/4 ------> 1:00/2:30/100/5/4 Slow 1:30/6/100/12/6 -----> 1:30/3:00/100/6/6 2:00/8/100/16/8 -----> 2:00/3:30/100/7/8 3:00/12/100/24/12 -> 3:00/4:00/100/8/10 I chose the max game time for the last one as 10 hours instead of 12 because I think any human would already be totally exausted after 8 hours of a long game. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by Janzert on Dec 3rd, 2005, 4:19pm I was going to mention something about the time control during the first round but had decided to wait till after the wc was over. Now that the conversation has come up anyway though. ;) I like the pace watching the games at 90s per move. But the reserve time seems short when the situation get's complicated and someone wants to do a deep think. I was going to suggest going to a 10m reserve, but also wouldn't mind Fritzlein's 12m. 6m seems like it would be really tight. Maybe if adopting Omar's suggestions, going to 2:00/3:30/100/7/8 for the wc would work out ok as well. Janzert |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by 99of9 on Dec 6th, 2005, 6:50pm I think Omar has overreacted a bit with his proposed reduction of the time reserves. I know it's for the spectators, but I think it will force us to use longer time controls than we would otherwise, just to give ourselves a bit of reserve leeway. Arimaa is quite an assymmetric game. The person under rabbit pressure almost inevitably uses more time than the other. So in some phases of the game you need more time, and others less time. Hence the need for a reasonable reserve. I would hate to see us change to rabbit-forward games just to put our opponents under time pressure. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by Fritzlein on Dec 7th, 2005, 9:14am I was too hung up over exact proportions in my first post. Now that I think about it, a little bit longer max reserve and max game time are in order for blitz games, and a little bit shorter max reserve and max game time for slow games. For computers the ideal proportions might be uniform, but for humans there are issues of attention span and exhasution. My revised proposal would be 0:15/1:30/100/3/1.5 0:30/2:00/100/4/2.5 0:45/3:00/100/6/3 1:00/4:00/100/8/4 1:30/5:00/100/10/6 2:00/6:00/100/12/8 3:00/8:00/100/16/10 I agree with Janzert and 99of9 that the max reserve times in the current WC games and in Omar's proposals are a little tight. To some extent it is fine with me if people complicate the situation hoping for a mistake under time pressure, but we don't want to encourage unsound play too much. If the objective in having a smaller max reserve is to benefit spectators by keeping the game moving steadily, then I propose a different solution for WC games, and any other games that will draw live observers. Rather than reducing the maximum reserve, reduce the percentage of excess time that is banked, e.g. try 2:00/6:00/75/12/8 for the finals of the WC. The smaller percentage of excess time going to reserve has a double benefit for spectators: Not only are players slightly discouraged from thinking longer on any given move, they are also slightly discouraged from moving quickly, since they will bank less time for doing so. The net effect is to encourage the game to proceed at exactly two minutes per move, which means specatators will suffer neither long waits, nor a flurry of quick moves where they don't have time to kibitz in the chat room about what is going on. On the other hand, if players earn it by moving faster most of the time, they will still be able to compile an adequate reserve to deal with delicate situations. Having a relatively large stockpile of time would discourage intentional surprise "cheapos" that only work because the opponent didn't have enough time during a couple of critical moves. To exhaust a larger reserve, it takes a pressure that lasts for more moves, i.e. it takes strategic pressure rather than tactical tricks. (By a tactical trick, I mean something which can be resolved in two or three moves with careful play.) Now that I think about it, reducing the percentage of banked time makes sense to me, not just something for games with lots of spectators, but as a standard for all games. You always have at least one spectator, namely your opponent. With that in mind maybe the standard time controls should be 0:15/1:30/100/3/1.5 0:30/2:00/100/4/2.5 0:45/3:00/100/6/3 1:00/4:00/100/8/4 1:30/5:00/90/10/6 2:00/6:00/75/12/8 3:00/8:00/50/16/10 |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by Fritzlein on Dec 7th, 2005, 9:31am on 12/01/05 at 18:27:54, 99of9 wrote:
It's true that bots are more likely to need more than 120 moves to settle a game on the board than humans are. Also we don't need to protect bots from exhaustion. On the other hand, bots are also more likely to get stuck in a nearly infinite loop where nothing is happening. I think it makes sense to impose a move limit somewhere, because there is no guarantee that a game which has gone 250 moves won't go 2500. Eventually a tournament director is going to need to start the next round. I can see allowing bot-bot games to have double the move limit of games with a human involved, but having any game be totally open-ended is asking for trouble in my book. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by MrBrain on Dec 9th, 2005, 9:54am I have noticed that nobody mentions the other effect of having a max time reserve in that when you hit it, you actually begin thinking longer on some moves than you ordinarily would. There are many times when I have my reserve maxed, and think I have a good move almost immediately. But then why move immediately (it won't go to reserve)? Use the additional time to make sure the move is right. This actually slows the game down. In short, I am in favor of as much control over how to spend your time as possible. I think the spectators have to come second when it comes to Arimaa. If the game is not fun to play due to being under time pressure all the time (which is how I always feel), then the spectators won't have anything to watch, because people won't play. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by Fritzlein on Dec 9th, 2005, 11:36am Mr. Brain, I know that you prefer slow games, and you would support anything that gives you more time than is currently allowed. However, the issue of how much time to give the players is somewhat distinct from the issue of how much control to give the players over time management. We're concerned about game pacing as well as total game length. In addition to your opinion that anything longer is better (which I respect) I'm also interested in your opinions on tradeoffs within a given time control. For example, Janzert proposed compensating for a smaller reserve with a longer time per move. As an extreme example, we could play the WC games at 2:30 per move with zero reserve or at 1:30 per move with a ten minute maximum reserve. Would you prefer the extra time per move even at a loss of control over time management? I personally value some degree of time management. I find that I don't need lots of time on every move. As you say, sometimes a good move is quickly apparent. And, like you, I want to have more time on the moves where I really need it. I can get by with less total time if I'm given more control on each move. But (perhaps unlike you?) I almost never feel unprepared to move after four minutes of thinking. If I haven't found a good move after four minutes, it's not because I think too slowly, it's because I'm losing! I agree that the needs of the spectators should be balanced with the needs of the players. However, as a player, I actually support restricting time management so my opponent can't blitz for twenty moves and then think for half an hour on one move. Huge pauses are annoying to me as a player even more than as a spectator. I like the game to move along at a fairly steady pace, and I'm willing to give up some of my control over my own time to make that happen. At the moment I'm still enamored of moving away from banking 100% of time in all games with longer time controls. The slower games are exactly the ones where one is likely to not need all the time on some moves, so if you banked everything without a maximum, the potential for overly long thinks would be too great. Yet if you solve the problem by installing a low maximum reserve, then players will quickly exhaust it in a complicated situation, although they could use a bit of extra time for a few more moves. As you point out, when the max reserve is low, players will often reach it, but when maxxed out they have no incentive to "throw away" extra time, even on obvious moves. This increases total game length without increasing the standard of play much. In contrast, a higher reserve with a percentage discount on banked time would actually speed up games, given the same time per move. Players would less often be at the max reserve, and therefore more often choose to throw away extra time (when appropriate) in order to build a reserve. To the extent that players manage their time rather than moving at a steady pace, they choose to use less total time. Yet if they don't manage their time, they will move steadily, which satisfies Omar's concerns about keeping spectators engaged. Omar proposed 2:00/3:30/100/7/8 as the standard for two-minute games, but I would be more comfortable playing 2:00/6:00/75/12/8, which I think meets his concerns just as well. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by PMertens on Dec 9th, 2005, 3:35pm Quote:
yes I always hated that about gnobby, but I have not ever seen a human to play like that (without good reason) Due to my personal playstyle I would definitively prefer longer reserve with shorter moves ... and it also solves my problem of constant interruptions, possible network problems and anything like that ... So I love the reserve and would hate to do without it. Sure, I play fast, so my reserve is quickly filled up - mostly quicker than for my opponent but with no/low reserve the "slower" has no advantage (apart from less interruptions or a better network ;-) ) since he will time out far more easily. I like the reserve because it is a pressure-meter. Therefore it cannot be to small because its value would decrease. Most of the times it is not only one move that needs time but a series of moves because playing to fast brought you into a stupid position ;-) I think 5 minutes is a healthy mininum, but I could live with more (so 1:30 with 10 is better than 2:30 with 0) In addition I am well aware that many players just need their time and that anything below 2 minutes would make no sense since we do not want to see who plays best under pressure but who plays best. I know it is unrealistic, but I want breaks like in tennis :-) If it rains I want to be able to stop without timeout. Anytime a game goes longer than 4 hours I am in extreme risk of timing out - be it because of timezones, work, sleep, food, wife, child, restroom, phonecalls or any combination of the above or what else you can think of ;-) |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by omar on Dec 10th, 2005, 1:21am This issue of ideal time controls is not something which has a definite right or wrong answer. If it did I would definitley want to find it, but this issue is tangled with personal prefernces and different people will want different things. We could talk about this till we run out of disk space and we may still not all agree on the same thing. I hate dwelling on such things especially when in the end it won't even matter too much whether we go with A or B as long as we're in the ball park. Without getting into too much details, let me just state some things that I beleive: * No matter what the time control is there will always be someone who thrives at that speed. * The time control has nothing to do with wether or not players will enjoy playing the game, but has a big impact on how the game appears to spectators. * The true spirit of a player should be: I don't care what the time control is; I'll just consider it part of the game rules; as such I'll just practice till I get good at it. And practice does make a difference: http://abcnews.go.com/Technology/story?id=99513&page=1 * Besides no matter what the time control is, your opponent has the same. So a player has no real reason to complain other than personal preferences. * The main reason for increasing the time controls is so that there will be less errors and the games will be of higher quality. But at some level of analysis even the best played games contain strategic errors. So there's a limit on how much can practically be acheived by increasing time controls. * The main reason for decreasing the time controls is to reduce the maximum wait time for a move and the maximum time it will take the game to reach a certian number of moves. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by Fritzlein on Dec 10th, 2005, 8:56am Omar, I actually started the thread with no intention of reviving the slower/faster debate. I mostly wanted to talk about good values for the other parameters when the number of seconds per move is given. I think there is some room for improvement there, but I agree that as long as we're in the ballpark it is fine. I also believe you have other important ways to spend your time, especially in tournament season, and that this thread could be back-burnered until June when nothing else is going on. It's even no big deal to me if you ignore it entirely. I do, however, want to particularly comment on one thing you said. Quote:
This is too strong. The time control does affect how much players enjoy playing the game. Many people have expressed this in many different contexts. It may be based in pesonal preference, and it may be difficult to satisfy everyone simultaneously, but that doesn't mean it makes no difference to anyone. That said, I will of course try to adapt to whatever time controls you decide on as standard for the gameroom, and I accept the time controls for tournaments as part of the rules. Anyone who isn't satisfied with your decisions has the option of not playing, and for the record I am quite satisfied, even though I am still suggesting minor adjustments. I like the way things run around here, and I appreciate all of your efforts. If I ever stop playing Arimaa, it won't be because you didn't go the extra mile to create a great environment for playing, in addition to having created a good game. Thank you for everything you do to run this server. Please take my comments as an expression of my desire to help make things even better, rather than as a complaint that they are unsatisfactory at present. You are doing a great job, which is another thing I have heard many people express in many different contexts. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by Ryan_Cable on Dec 11th, 2005, 10:28am on 12/10/05 at 01:21:23, omar wrote:
I mostly agree with this, but I think there are a few opportunities to make improvements that nearly everyone will agree with. For instance, I think there is a broad consensus that Arimaa's system of Xm per move is more fun than Xm per Y move even though the latter gives more flexibility. I think that Fritzlein's suggestion of using larger reserves, but having only part of your increment banked could be one of these situations. Does anyone not think that 1:30/5:00/90/10/6 2:00/6:00/75/12/8 3:00/8:00/50/16/10 are better than 1:30/3:00/100/6/6 2:00/3:30/100/7/8 3:00/4:00/100/8/10 ? I can't think of an objection other than saying "Well my opponents play more worse with a short reserve than I do?" which I don’t consider legitimate. I have done much more spectating and kibitzing than I have playing this tournament, and I have had at least as much fun kibitzing as playing. I think the same is true for many players. The only times I have been bored were due to the players playing slowly on the board rather than playing slowly on the clock. Thus I feel qualified to say that adding 3 or 5 min to the max reserve would not decrease my enjoyment of spectating at all. In fact, it would probably make it more fun, by giving a longer chance to look at interesting positions and by marginally improving the quality of play. Personally, as a player, I would like to have an infinite reserve with full banking, but I agree that a reserve greater than 10 times the increment is potentially harmful to the spectators unless there is a large banking penalty. However, I think increasing the reserve to ~7 times the increment would be a good thing all around. In my experience, Arimaa has runs where the position is sharp for several moves, and other times it is fairly dull for several moves. While the time control for a tournament is fundamentally tied to the structure of the tournament, there are two other sets of time controls that are at least as important to discuss. The first is the list of time controls that we have available for inviting humans. For this I think Fritzlein has made a good proposal that should be considered and perhaps refined further. The other is the set of times that Omar sets up the championship bots to play. For the blitz and fast bots I would like to see the reserve times doubled, but maybe some people see the short reserve times as part of the thrill. I have gotten pretty good at bot bashing at 30s per move, but a year ago I would have liked to have a bot at 1m per move or else have the fast bots play 45s per move. The P2 bots fill this gap somewhat though. on 12/10/05 at 01:21:23, omar wrote:
You haven't played bot_GnoBot2004CC lately have you? ;-) Seriously, while I can enjoy playing games at 45s or even less, I know I would like the tournament much less if I didn't have at least 1m15s to think. I think the frequent rapid, blitz, and even sudden death games is a notable contributor to the FIDE knockout tournament being such a farce. on 12/10/05 at 08:56:27, Fritzlein wrote:
Yes, I would take Omar over FIDE any day. But maybe Omar will put the WC time control up for a vote if MrBrain makes a $1m deposit. ::) http://chessbase.com/newsdetail.asp?newsid=2781 on 12/10/05 at 01:21:23, omar wrote:
Indeed, practice at blitz has helped me far more than I expected. When I quit playing 'nator back in May, I could only beat it 2/3 of the time, and I struggled badly in games that were 1m per move or less. However, the two games I have played with 'nator since Omar put it back online have felt like a cat playing with a mouse. And most of my games since May have just been Bait and Tackle rating inflation. on 12/09/05 at 11:36:09, Fritzlein wrote:
So true! I lost a 4d/4d/100/8d/0 postal to 99of9 on time because of this. I got into a position where I could not find a good move in 30m of looking, so I put it off to look again another day. I repeated this several times never finding the miracle move I was looking for. Then I went a couple of days without logging into the gameroom, and I lost on time without ever even noticing I was in time trouble. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by omar on Dec 13th, 2005, 12:44am When I re-read my last post it sounds kind of harsh. Im sorry I didn't mean to come off that way. I was just a little too blunt. I guess in my mind I've been re-evaluting the time controls situation completely even though Karl was just proposing some adjustments to a few of the parameters. Im a bit unsatisfied with our slow time controls. The faster time controls Im not too worried about, but the slower time controls like the ones we use in the WC and challenge match have been bothering me and I had put it on the back burner and avoided it. This topic keeps reminding me about it :-) The thing that bothers me about our slow time controls is mainly the potential for having to wait for 10 minutes or more for a player to move. Yes, it's true that in some situations even that much time does not seem like enough, but I also agree with Karl that: "If I haven't found a good move after four minutes, it's not because I think too slowly, it's because I'm losing!". Likewise for spectators, I think waiting about 5 minutes for a move is the limit of what can be endured in a live game. Beyond that it begins to enter the zone of watching paint dry. Another problem with our current slow time controls is that players need to build up the reserve by making some moves more quickly than the allocated time per move. If a player has not yet built up the reserve and hits a tough point in the game they are sure to get into time trouble and possibly lose on time because they may only be getting about two minutes to make the move (much less than the four minutes Karl mentioned). The bad thing is that this could happen even when the game is relatively young and there are still many hours before the max time limit for the game is reached. This afternoon I was re-reading the time control specs on the "Match Rules" page: http://arimaa.com/arimaa/learn/matchRules.html and rediscovered the T parameter. I had thrown that in thinking it might be useful sometime, but had never really used it. Now that I am looking at it again, I think it might help us fix the problems I mentioned. Consider this time control: 0/2h/0/0/0/5 This is basically a sudden death time control where each player has 2 hours to make all their moves. So the whole game will be over in 4 hours. Some people like such time controls because it gives the player full control over managing their time. I also like that part of it, but I don't like the fact that a player could take 30 minutes on one of the moves and put everyone to sleep. However there is twist in this time control. The 5 at the end says that a player may never take more than 5 minutes on any one move even if they have lots of time in reserve. The more I've been thinking about this kind of a time control (incorporating the T parameter) the more Im begining to like it. Not only is it conceptually simple and easy to understand, but it provides a way to solve the major issues I've been unsatisfied about. * There is a clear, definite and well controled limit on how long the spectators/opponent needs to wait for the move. * There is a clear and definite limit on how long the game will take. * A player does not need to make moves quickly before being able to use more thinking time and can also use sufficient thinking time on a series of moves without fear of immeadiate time trouble. Thus the quality of play problem is solved. * No need for a game result to be determined by score. If we are worried about the sudden death ending we could even do something like this: 15s/2h/0/0/0/5 Thus you will always get at least 15 seconds per move. In any case Im begining to think that we should make use of the T parameter. It really adds a new paradigm to the time controls. Im really glad I threw that in when I was writting the time control specs. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by PMertens on Dec 13th, 2005, 6:06am Naveed and I played several 15min games and I must say that they do play absolutely different even more stressful than Blitz in some ways. Assuming one player plays slow with an average of 120s per move and some other player (like me) only needs 60 then I have a totally new way of threatening a goal ... I just need to make the game long enough: 60 moves (and the time pressure around move 50 will be far worse than today) It does actually pay to play defensive for the faster player ... it is (for me) easier to move the mouse faster than to play on the board and outsmart the opponent (mostly I outsmart myself) Is this really what you want ? I would hate to see to many game ended because of this absolute time pressure. Today a great move can help resolve my pressure and save me. I guess 15s/2h/0/0/0/5 is a little bit to cruel, but maybe 1m/1h/0/0/0/5 is more acceptable ? (also over at move 60 with 2 minutes per move but far easier to stretch for the slowmoving player) |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by Ryan_Cable on Dec 13th, 2005, 9:17am 0/2h/0/0/0/5m is certainly an interesting time control. If you added 15s/30m/0/0/0/3m, 15s/45m/0/0/0/4m, and 15s/1h/0/0/0/5m and made them count for PotM, I think some people would play at those controls and enjoy it. (Currently the PotM rules say: Games with only a reserve time and no time per move are not counted) I would try it out at least. However, I would oppose making this our WC time control. With the Xm per move system, time management is tactical. You only have to decide if considering an alternate candidate move / investigating a line further is worth the time it will take from your reserve. Basically this is just a simple guestimate of how sharp the position is. With Zm per Y moves, one must always be thinking of how many moves are left until the next increment and whether the position is likely to get sharper or duller in the future. This makes time management harder for newbies and people who just want to focus on playing the game. Still, for Z = X * Y, Zm per Y moves dominates Xm per move (by which I mean you have the flexibility to do anything you can do with the other and more), so it can be a good time control for professionals and serious amateurs. However, I think Wm per game goes much too far, because you must estimate how many moves are left in a game to decide how to spend your time. Moreover, the main advantage of Wm per game is being able to think for a long time about a move or two at a really critical spot, which you explicitly forbid. Still, I think the T parameter is probably a better way to keep the game moving along for the spectators than the L factor. I suggest the following for a potential WC time control: 1m30s/10m/100/0/100t/6m The key to making this control work really well is to alter the specification so that when G is reached the game doesn’t end but instead the increments stop. (I think this is a good idea independent of other time control decisions.) This time control is dominated by 0/2h40m/0/0/0/6m, but normal length games will take half as long, and for all but record length games the players will always have at least 1m30s per move. Personally, I would prefer to increase the T factor to 6 times the increment (9m). While, 5m or 6m is pretty good for even the sharpest of tactical situations there are often strategic situations for which it is reasonable to think for much longer. Look at my WC game against MrBrain around move 19b or 28w: http://arimaa.com/arimaa/gameroom/replayFlash.cgi?gid=22244&s=w&client=1 Yes, it would be a pain for the spectators to have to wait 9m for a move, but if it only happens once or twice in an otherwise fast paced game, I don’t think it would drive people away. I think the worry about long pauses is clouded somewhat by chess, where games will often only have 15 to 25 moves that are thought about at the board. Arimaa doesn’t have long opening books, frequent short draws, and a culture of resignation brought on by massive endgame study. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by omar on Dec 13th, 2005, 4:03pm on 12/13/05 at 06:06:05, PMertens wrote:
In my last post I was just trying to bring to light the possibility of using the T parameter. The specific values I proposed were just an example and not intended for any specific tournament. We still need to think about what values might be good and try them out. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by Fritzlein on Dec 13th, 2005, 4:30pm I think it is a great idea to have a five-minute limit per move. I actually thought about proposing that myself, but then shied away of creating more programming work. Given that it is already coded, I think it should be the standard for the WC, and maybe for lobby games too. The reason I was arguing for a larger maximum reserve was that someone may need a slightly longer think for several moves in a row, not a super-long think on one move. The maximum time for a single move solves that problem nicely. Could we use all six parameters at the same time, e.g. 1m/1h/75/1h/5h/5m? If so, then let me rethink my proposals for ideal time controls. Between the extreme of "All reserve and no increment" on one hand and the extreme of "All increment and no reserve" on the other hand, I would lean towards most of the total time coming from increments. Arimaa is not like chess, where the position simplifies towards the end. Chess players actually need less time per move to play at a high standard towards the end than they need in the middlegame. Arimaa, in contrast, often has tactically complex and tense positions late into the game. It would be a shame for those positions to be blitzed because the reserve is gone and the increment is small or non-existent. I buy PMertens' argument on this issue. With a large starting reserve, some players are going to approach the per-move maximum almost all the time until the reserve is exhausted. This probably isn't in their best interest, but the punishment could be a long time in coming. For example, in a game at time control 15s/2h/0/0/0/5, someone could (unwisely) think for four minutes every single move, and not feel the pain until move 32, when they suddenly have to blitz every move. I wouldn't like this as a spectator, and I wouldn't like this as the opponent. Yes, as Omar argues, there may be some early moves that require a longer think before there has been much time to build up reserve, but (parallel to my earlier comment) if you need a long think for more than half a dozen moves in a row, you are not fighting through a tactical trick of the opponent, you are under strategic pressure. I therefore suggest a starting reserve of six times the increment. Any time beyond that, you have to earn. This places a premium on correct time management from relatively early on, as opposed to allowing lousy time management for most of the game before dropping the hammer. However, there is no longer any need for a maximum reserve. The maximum time per move takes care of long waits for spectators and opponents alike. I stand by my earlier ideas for percentage of unused time banked. At longer time controls, extreme time management should be discouraged. This is also in the best interest of the spectators. In a two-minute per move game, moving in ten seconds is not only rarely wise, it also gives spectators insufficient time to follow developments, and merits a discount on time banked. Thus my revised revised proposal for standard time controls is: 0:15/1:30/100/0/1.5/5 0:30/3:00/100/0/2.5/5 0:45/4:30/100/0/3.5/5 1:00/6:00/100/0/4.5/5 1:30/9:00/90/0/6/5 2:00/12:00/75/0/8/5 3:00/18:00/50/0/10/5 |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by omar on Dec 13th, 2005, 5:30pm on 12/13/05 at 16:30:23, Fritzlein wrote:
I think these look pretty good and the 1:30 time control happens to be pretty close to what Ryan suggested for the WC. Lets give this set a try and get a feel for it; we can always adjust it more if we feel the need to. I'll added it to the invite form. I did not test out these kind of time controls very much so if you notice any problems, let me know. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by PMertens on Dec 13th, 2005, 5:50pm on 12/13/05 at 16:03:23, omar wrote:
I agree and am always open for a game to try new things out ;-) |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by omar on Dec 13th, 2005, 6:46pm I justed tested out the T parameter with Paul's help. Unfortunately I had added it to the specs, but did not implement it yet. Bummer. I'll try to add it soon. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by Ryan_Cable on Dec 14th, 2005, 6:03am Omar, what do you think about my proposal to alter the specification so that when G is reached the game doesn’t end but instead the increments stop? The only downside I can see is that infinite move draws would be won by the player who blitzes fastest. I don’t believe infinite move draws have ever been much of an issue in actual play, but PMertens v. bot_Clueless2005CC does look close to an infinite move draw after 55w with perfect play: http://arimaa.com/arimaa/gameroom/replayFlash.cgi?gid=22232&s=w&client=1 In fact, Arimaa could very well be an infinite move draw with perfect play. A workaround could be to say if after G is passed, the game ever goes 50 moves without a capture then the game is declared a draw, and score is used only for tournament purposes. If we made these changes, I would then want to change Fritzlein’s time control set to 0:15/1:30/100/0/150t/5 (max game length 1h18m) 0:30/3:00/100/0/150t/5 (max 2h36m) 0:45/4:30/100/0/150t/5 (max 3h54m) 1:00/6:00/100/0/125t/5 (max 4h22m) 1:30/9:00/90/0/125t/6 (max 6h33m) 2:00/12:00/75/0/125t/8 (max 8h44m) 3:00/18:00/50/0/100t/12 (max 10h36m) I also increased the T factor to always be at least 4 times the increment. I think it is mostly a waste of people’s time to play at 2m or 3m increment per move if they must always move within 5m. on 12/13/05 at 16:30:23, Fritzlein wrote:
True, but the T factor makes it significantly harder to deal with strategic pressure because you can’t sit and think strategically for 10m or 15m at the first of those half dozen or more moves. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by PMertens on Dec 14th, 2005, 10:14am on 12/14/05 at 06:03:49, Ryan_Cable wrote:
good question ... Infitite move draws are usually not for humans because games would just take to long .... (The only reason I messed that one up was because I did NOT want a draw ... or actually a victory by points at the end of time ... and because I was simply to tired ;)) The whole speedtrap experiment and the rabbit pulling with no piece crossing the line are both fine examples that infite moves are more than just a possibility, but a very real thing to consider. Unfortunately most bots would play blitz eventually and never run out of reserve (unless bugged) :o |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by omar on Dec 14th, 2005, 12:56pm on 12/14/05 at 06:03:49, Ryan_Cable wrote:
So after the game time expires the players no longer get the time per move and can only use up the time they have in reserve, right? I think this is an interesting possibility. But keep in mind that the real purpose of the G parameter is to allow tournament organizers to be able to fix a definite time on how long the game will take, due to practical considerations. For example people might have to catch a plane or have an important appointment and so the game must finish in 5 hours. If after the G parameter expires and there is a lot of time in reserve it could become a problem. Of course if that is expected than the G paramter could be set to a lower value; but might involve some guess work. Fortunately all the important games so far have ended without hitting the game time limit since we set it high enough to allow for about 120 moves. But still I plan to discuss in the future alternatives to what happens when the game time runs out. Currently the result is determined by score. But other solutions have been proposed and we eventually need to hash through them. So lets hold this tought till we get into that discussion. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by omar on Dec 14th, 2005, 12:58pm on 12/13/05 at 18:46:17, omar wrote:
It's working now. Give it a try and let me know if you encounter a problem. Many times I end up breaking something else whenever I make a change :-) |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by Fritzlein on Dec 14th, 2005, 1:52pm on 12/14/05 at 06:03:49, Ryan_Cable wrote:
From my experience with 2-minute games, I definitely don't think a 5-minute limit per move makes the reserve pointless. An "I'm in trouble" pace of thinking, where I take twice as long on every move, is still extremely useful. Obviously it gives me less control than an 8-minute limit per move, but the control I do still have is nothing to sneeze at. I think a 5-minute limit per move is a reasonable concession to the needs of the spectators. For 3-minute games, I don't feel strongly. The WC, CC, and Challenge tournaments will never be played this slowly anyway, nor will I likely ever play a lobby game at this speed, so my opinions should count less than those of someone who would actually use the time control. Looking in my database to see what speeds people actually choose to play, I find over 4000 games played at 3 minutes per move, but that is misleading because the lobby bots played at this time control until near the end of 2004. If we look just at 2005, and exclude games played against Bomb2004CC and Gnobot2004CC, I find excatly six games played at four minutes per move and three games played at three minutes per move. If a time control is used less than a dozen times in a year, we could leave it out with justice, but then again, why not keep it around just in case? I guess what the lobby should have is one "long" time control of more that two minutes per move, for the very few people who want it. Furthermore, because no tournament games will ever be played at such a slow speed, we might as well loosen up the restrictions. With that in mind, I propose scrapping the 3-minute time control (which has been chosen less often than the 4-minute time control anyway), replacing it with 4:00/24:00/50/0/10/16 True, it will hardly ever be used, by why not leave something ultra-slow for the people who actually want it? |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by PMertens on Dec 14th, 2005, 3:52pm to make games longer than 10 hours playable I think the reserve needs to be much larger ... in fact I doubt that reserve makes any sense with that kind of game to the near certainty of interuptions. But then I certainly am not interested in that timecontrol :-) |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by Fritzlein on Dec 14th, 2005, 5:42pm Actually, I'm having a hard time putting myself in the shoes of someone who wants an ultra-slow game, but doesn't want a postal game or a totally untimed game. You must want the game to keep moving at some pace, and must want it to end at some time, so half-hour breaks don't seem reasonable, yet at four minutes per move you also really don't want to be rushed, and are letting yourself in for a ten-hour game without anything extraordinary happening. It doesn't make sense to me! |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by 99of9 on Dec 14th, 2005, 7:41pm on 12/14/05 at 17:42:55, Fritzlein wrote:
Here's a comparable situation that might help: It's Sunday, a boy and his father both have the day free, they enjoy playing arimaa, and plan a game for the day, but it needs to finish by the end of the day (in say 12 hours time). Both will do other things during this time, for example eating meals, taking phone calls, and even maybe having an afternoon nap. Both want to win the game, and would even consider using sly time-control tactics to win if they were available. Neither wants that kind of exploitation to be possible, because they'd like a fun game. What time control should they use? I'd say something like this: * Each person starts with 6 hours in the bank. * No increment per move. * No move limit. That's it, simple! I don't think any of the other fancy time control measures would help here. Quote:
Not really, you may be happy for it to progress in fits and starts. Quote:
It will. Quote:
They do to me. Quote:
You won't be. If you *really* want to think, you are welcome to spend the entire day at the board. Quote:
You can do plenty of other things at the same time, it's quite relaxed. Quote:
Hopefully now it does. I agree this kind of situation is less likely for pairs of internet players, but it's certainly still possible. For example if there were 3 back-to-back WC games going on, and the two of us were not involved (having been knocked out in an earlier round ;-)), we might be willing to have this kind of game whilst watching the champions battle it out. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by Fritzlein on Dec 14th, 2005, 11:45pm I wouldn't want to play a game with a six-hour starting reserve, knowing that I might have to wait an hour for your move if you went to eat lunch. I'd rather play at four mintues per move, and that's saying something, because I never play at four minutes per move. But not everyone wants the same things I want, so I should just shut up. Anyone who will actually use a long time control can make the case for it being included in the lobby, and I will make no objection. :-) |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by 99of9 on Dec 15th, 2005, 12:47am on 12/14/05 at 23:45:33, Fritzlein wrote:
I'd probably give you a heads up ;-), then you could fit in an extra game of lightning! |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by 99of9 on Dec 15th, 2005, 12:54am on 12/14/05 at 23:45:33, Fritzlein wrote:
I personally think it would be great if we could have the option to type in weird and wonderful time controls of our own construction. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by PMertens on Dec 15th, 2005, 2:41am on 12/15/05 at 00:54:49, 99of9 wrote:
That might indeed be the best solution. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by Ryan_Cable on Dec 15th, 2005, 5:23am on 12/14/05 at 12:56:38, omar wrote:
Exactly. on 12/14/05 at 12:56:38, omar wrote:
If G is given in terms of time, you are right; the game length would be mathematically unbounded. The faster people played the more increments they would get before G is reached and thus the greater reserves they would build up. However, if the G parameter is given in terms of moves then there is a strict upper bound on how long the game could last: Max length = 2*(G*M+R) Thus we have bounded the time a game can take, but the number of moves the game could last remains mathematically unbounded (unless we add some workaround for infinite move draws). Additionally, any game that lasts g<=G moves will take at most: Length = 2*(g*M+R) Thus for the two likely WC times from my list we get: 1:30/9:00/90/0/125t/6 (max = 6h33m, 45 move max = 2h33m, 90 move max = 4h48m) 2:00/12:00/75/0/125t/8 (max = 8h44m, 45 move max = 3h24m, 90 move max = 6h24) If those max times seem too long we could decrease G to 90t, and 95%+ of games would still not run out of increments. I picked 125t just to make sure that people don't have to adjust their time management for the possibility of running out of increments unless they are already in a record length game. on 12/14/05 at 13:52:12, Fritzlein wrote:
I definitely agree that the reserve is useful, and I greatly prefer these new time controls to our current ones even if we keep the T=5m. What I meant is that under this new system, I don't think I would play noticeably better at 2m per move than at 1m30s per move if I must always move in 5m. I am sure everyone is different, but my experience thus far in the WC is roughly: 0s: 10% of moves I had anticipated my opponent's move and chosen my response on his clock 30s: 20% 30s to 1m: 30% 1m to1m30s: 10% 1m30s to 2m30s: 20% 2m30 to 4m: 10% 4m+: 0, 1, or 2 moves per game I have really wanted to think for a very long time and not been able to. All of that adds up to about 1m10s per move, which leaves 20s for watching the move come in, entering my move into the client, opening a plan window, etc. Yet I consider myself to be a slow player, and I don’t like to play humans at less than 1m per move. The 7m reserve limit often caused me to move slower than I thought was necessary and to move too fast at a few critical points. On the whole, I think I will be quite happy with the new style 1m30s time control. And I am sure, if I got to play at the 2m time control, I would find something useful to do with the extra 30s per move, but by far the thing I would most want to do with them is store them up for a 5m to 8m think. I understand the desire of the spectators to keep the game moving along and not be bored. About 30 moves into megamau v. Belbo, I wanted to slap Belbo and yell "Attack man; attack already!" :-) Still, I think T=8m is much better than T=5m. It is only 3m extra of boredom for the spectators, and the need to store up time for it means it can only happen a few times per game. But for the players, it could be the difference between winning and loosing or at least the difference between loosing and getting crushed. Still, getting rid of the reserve limit would solve 80%+ of my problems with the time controls, so I would be happy to wait and debate the T factor after the 2007 WC. on 12/15/05 at 00:54:49, 99of9 wrote:
Yes, that would be nice. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by omar on Dec 19th, 2005, 12:15am Ryan, you are right that if the G parameter is specified as max number of turns then if the game continued after reaching this limit using only the reserve time there would be a definite limit on the total time for the game. In fact even if the G parameter is specifed as a max time and there is a max reserve limit (L) there would also be a definite limit on on the total time for the game: G+2*L. The only time it is a problem is if G is specified as a max time and this no limit on max reserve. So an event organizer could always limit the total time for the game if they really want to. Your proposal should definitely be considered when we discuss what should be done when the game limit parameter (G) is reached. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by omar on Dec 19th, 2005, 12:26am I added another set of experimental time controls to the invite form. I don't know if I really will like them yet; I need to try them out. But they do look clean, elegent and easy to understand. 15s/30m/100/1m/0/1m Blitz 20s/60m/100/2m/0/2m Sprint 25s/90m/100/3m/0/3m Run 30s/120m/100/4m/0/4m Jog 35s/150m/100/5m/0/5m Walk 40s/180m/100/6m/0/6m Stroll |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by omar on Dec 19th, 2005, 12:28am on 12/15/05 at 00:54:49, 99of9 wrote:
I'll try and add that. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by Fritzlein on Dec 19th, 2005, 9:08am on 12/19/05 at 00:26:23, omar wrote:
Er, am I reading that right as a starting reserve of 30 minutes and a maximum reserve of 1 minute? |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by omar on Dec 21st, 2005, 1:43pm on 12/19/05 at 09:08:14, Fritzlein wrote:
Yep, it's allowed according to the specs: http://arimaa.com/arimaa/learn/matchRules.html An upper limit (L) can be given for the reserve so that the reserve does not exceed L when more time is added to the reserve. If the initial reserve already exceeds this limit then more time is not added to the reserve until it falls below this limit. The upper limit for the reserve is optional and if not given or set to 0 then it implies that there is no limit on how much time can be added to the reserve. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by omar on Dec 22nd, 2005, 6:20pm I've changed the page to start bots Bomb2005CC and Bomb2005Fast to allow having the bots join a game which is already created. We can use this feature to test out the experimental time controls against these bots. The way to use this is to first use the "Start Game" button to create a new game with the time control that you want to use. Then from the bot's start up page enter your username in the "Join game against" field and click the "Start Bot" button. You should see the bot join your game. If you make the game rated, please use the faster time controls with Bomb2005Fast and the slower time controls with Bomb2005CC. Thanks. Have fun :-) |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by Fritzlein on Jan 9th, 2006, 12:06am Now that I've had a chance to try out the experimental time controls of 0:45/4:30/100/0/3.5/5 1:00/6:00/100/0/4.5/5 against Adanac and Robinson, I must say I like them very much. The unlimited reserve brings intelligent time management into play more than it was before, yet being restricted to 5 minutes on any move was no hardship, because I always wanted to move in under four minutes anyway. In each game I temporarily built up a large reserve and needed it all later when I hit a patch of tricky moves. It seems that this type of time control, compared to previous ones, is better for the players and better for the spectators. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by Ryan_Cable on Jan 27th, 2006, 1:54am As I expected, I have come to greatly prefer the new style time controls. I guess the CC bots will be set up with the official CC time control, but I would prefer to have the Blitz and Fast bots setup with 0:15/1:30/100/0/1.5/5 and 0:30/3:00/100/0/2.5/5. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by Fritzlein on Jan 27th, 2006, 11:04am Yes, I'm ready to make the experimental controls standard. They aren't very different, but insofar as they are different, they are simply better. Does anyone not like them? One problem with retroactively applying the new time controls to old bots is that some of the bots might not be able to handle it. Some don't make use of accumulated reserve, and others won't be aware of the per-move limit. Actually, I wouldn't be surprised if none of the bots (not even the 2006CC bots) could handle the new time controls, in which case for them the older, simpler time controls would be best. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by Ryan_Cable on Jan 27th, 2006, 2:27pm Good point, the T=5m would be a big problem for GnoBotCC, and maybe an occasional problem for others. For the Blitz/Fast bots I doubt it will be as much of an issue. I don’t think the T factor is that big an issue for HvB games anyway. Also, this is a good reason for us to pick our 2007 CC time control relatively early, so the bot developers can prepare. |
||||||||||||||
|
Title: Re: Ideal time control proportions Post by 99of9 on Jan 28th, 2006, 7:50am on 01/27/06 at 14:27:29, Ryan_Cable wrote:
I think you are right, but I can't recall for sure whether I put in T control. I'll check sometime. |
||||||||||||||
|
Arimaa Forum » Powered by YaBB 1 Gold - SP 1.3.1! YaBB © 2000-2003. All Rights Reserved. |