Life In 19x19

John Fairbairn wrote:I have never used Fischer timing or seen it used, so I have a couple of questions.

1. Is it taking over in the chess world?

2. Where it is extensively used, have there been any complaints about abuse of the system or the need (if any) to plan time-management, such that newcomers to the system may have a significant disadvantage compared to those who have plans in stock, as kit were?

1. I am not deeply involved in the chess world (although I used to be). As I understand it, Fischer time is the dominant system used on servers, but not so in live tournaments.

2. I am aware of no complaints/abuses, although I probably wouldn't be, since I am not really involved in chess much anymore.

The prevalence of Fischer on the servers suggests to me that this would also occur in Go. I know wms won't implement it, but my nickel says that if he did, it would become the most popular choice.

John Fairbairn wrote: But I have never used Fischer timing or seen it used, so I have a couple of questions.

1. Is it taking over in the chess world?

Yes, it is becoming more common every year, including at the highest level. Even when the earlier time controls are non-Fischer (e.g., 2 hours for your first 40 moves), the last time control (e.g., 30 minutes for the rest of the game) now often have a modest increment attached.

2. Where it is extensively used, have there been any complaints about abuse of the system or the need (if any) to plan time-management, such that newcomers to the system may have a significant disadvantage compared to those who have plans in stock, as it were?

Not to any great extent.

Note that chess timing at the highest level has basically gone through three stages, historically:

I. The equivalent of Canadian byo-yomi (but time carries over from block to block), e.g., 2 hours for 40 moves, then an additional hour for every 20 moves thereafter. This required adjournment to the next day once the game went over 6 hours or so, so players and their support groups would often spend all night analyzing the endgame in question.

II. Sudden death for the last stage, largely because the availability of computer analysis made adjournments more silly.

III. Increments during the last stage.

Most of the complaints have been more about the transition from I (endgames are now played with far less precision than they were in the middle of the 20th century, purely because of these factors) than the transition to III.

John Fairbairn wrote:1. Is it taking over in the chess world?

Yes. In Finnish chess tournaments Fischer is dominant time control. About half of the weekend tournaments uses Fischer and practically all longer (championship) tournaments. Typical settings for weekend tournament is 90 30f. Although Fischer is dominant, absolute time control is also used. (E.g. 90min/40moves + 30 mins for the rest)

zinger wrote:The prevalence of Fischer on the servers suggests to me that this would also occur in Go. I know wms won't implement it, but my nickel says that if he did, it would become the most popular choice.

This is reasonable assumption, because Fischer suits go better than chess, IMO. Although I am little doubtful if it would be most popular in Kgs. In FICS 3 0f is my favorite and most popular settings. Also my personal opinion thinks that in Kgs blitz I would not change readily my preferred timing system from 3x10sec, although in IRL I blitz using delay or Fischer. For slow games it is very difficult to argue that Fischer is not the best choice, for both chess and go.

Matti wrote:However if I had chance to program the clock's for a tournament with more than one round a day, I would set the increment to decrease when the game progresses. It might go down in steps or alternately the increment might be T/n. Where T is the increment for the first move and n is the number of the move to be played. If the player plays over 180 moves the increment could be 2-5 seconds for the remaining moves.

This is very bad idea for number of reasons.

1: First we can adjust this by changing the ratio of increment and time given at the start. E.g. 80 min T270 adjusted Fischer time is equivalent with 12 30f, 35 20f, 58 10f and 71 4f.

2: Second problem is that balancing time usage is difficult and is fixed to assumed mold. Not everybody want to use time lots for the beginning and middle game, but I find it very difficult and time consuming to play small endgame when I know that game is really close.

3: Third reason is that we do not have and never will have game timers for such.

4: this kind of timing would not be any better than absolute timing without any increments. At least absolute timing is predictable and people do learn to control it if they practice.

5: if several games are scheduled for the day, maximum increment is 20 secs. That means that T360 game for 40 20f will last 200 mins where as T270 will last 170 mins. And schedule can always bend 30 mins if such a rare event would occur that somebody plays 360 moves AND uses all their time. This kind of bending of the schedule is more likely with regular slow overtime than similarly slow Fischer.

6: sixth reason is simplest that it introduces extra variable and thus increases complexity of the timing system and people do need to use integral calculus to calculate how much there is time left.

Liisa wrote:

Matti wrote:However if I had chance to program the clock's for a tournament with more than one round a day, I would set the increment to decrease when the game progresses. It might go down in steps or alternately the increment might be T/n. Where T is the increment for the first move and n is the number of the move to be played. If the player plays over 180 moves the increment could be 2-5 seconds for the remaining moves.

This is very bad idea for number of reasons.

1: First we can adjust this by changing the ratio of increment and time given at the start. E.g. 80 min T270 adjusted Fischer time is equivalent with 12 30f, 35 20f, 58 10f and 71 4f.

Do you find all these equally desirable as a player?

2: Second problem is that balancing time usage is difficult and is fixed to assumed mold. Not everybody want to use time lots for the beginning and middle game, but I find it very difficult and time consuming to play small endgame when I know that game is really close.

No one forces you to spend your time earlier you want. If you want you can save time for the endgame.

3: Third reason is that we do not have and never will have game timers for such.

I wrote if I had a chance to program the clock. If you choose to discuss, this is not a reason against.

4: this kind of timing would not be any better than absolute timing without any increments. At least absolute timing is predictable and people do learn to control it if they practice.

Do you means that this kind of timing is unpredictable? Why?

5: if several games are scheduled for the day, maximum increment is 20 secs. That means that T360 game for 40 20f will last 200 mins where as T270 will last 170 mins. And schedule can always bend 30 mins if such a rare event would occur that somebody plays 360 moves AND uses all their time. This kind of bending of the schedule is more likely with regular slow overtime than similarly slow Fischer.

Are you comparing my idea to Fischer or somethig else.

6: sixth reason is simplest that it introduces extra variable and thus increases complexity of the timing system and people do need to use integral calculus to calculate how much there is time left.

Naturally the clock would show how much time there is left.

On average the time needed for each move slowly decreases during the progress of the game. Of course there a fluctuations, some times players play quicker are somtimes slower. It is good to have a fixed amount of time to begin with. At the end to fill dame one needs a couple of seconds for each move. In next Kido Cup they have Canadian byoyomi where first period is 20 stones in 5 minutes, next 30 stones, and then 40 stones. This is one way of increasing the speed. But we cannot predict when exactly players enter yose or start filling dame, so increaseing the speed gradually would be another method.

Matti wrote: On average the time needed for each move slowly decreases during the progress of the game.

I do not think that this assumption has any grounds to stand on. The only data on time usage patterns I know of is Timing Systems Redux on sensei's library. There were 10 games analyzed, where the players did not enter byoyomi at all, and the conclusion was:

• A typical game consists of a majority of plays made using very little time and a minority of plays made after thinking for longer periods of time. This probably surprises no one familiar with the game. However, the extent to which the distribution of plays is skewed toward the short end probably will surprise many. Typically players use about 10% of their time for the shortest half of their plays and 90% for the longest half. Typically players use 40% to 60% of their time for their longest 10 plays.
• The plays that use more time are randomly scattered throughout the game. This contradicts the "common knowledge" that Go players think longest early in the game and require less time later.

I think that the idea to strangle the time bonus comes from the misconception that a game under bonus time could "last forever". However, this is not the case at all: for example, a 30/15 game of 300 moves cannot take longer than 2:15 h, and even 360 moves do not extend that much (2:30 h). You can comfortably schedule the round starts three hours apart, maybe even two and a half hours. This schedule security is one of the main advantages of Bonus time.

Matti wrote:In next Kido Cup they have Canadian byoyomi where first period is 20 stones in 5 minutes, next 30 stones, and then 40 stones. This is one way of increasing the speed. But we cannot predict when exactly players enter yose or start filling dame, so increaseing the speed gradually would be another method.

I have decided against attending the Kido Cup this year because of this, even though I live in Hamburg and enjoyed last year's tournament very much.

This so-called "progressive byoyomi" is one of the worst ideas ever. In fact, it is just a hidden sudden death. At some speed, driving the opponent over the time becomes possible; this system thus fails to prevent this clock abuse. The only difference to sudden death is that you get some time warnings towards the end.

The problem that is attempted to address with this "idea" is that round schedules were quite insecure with main time plus canadian byoyomi. Many people, even including experienced tournament organizers, do not realize that main time plus byoyomi is completely different from absolute time. "One hour plus byoyomi" is not at all similar to "one hour sudden death". In the latter case, all moves must be made within one hour, but in the former, you do not even need to play a single move during that time! So, the scheduling problems from the "main time plus byoyomi" do not come from the byoyomi, but from the main time.

Some tournaments have recognized this, and simply reduced their main time. Many others have not, and fool around with always faster byoyomi. They try to get back to the sudden death, and they got it in hidden form.

Matti wrote:

Liisa wrote:1: First we can adjust this by changing the ratio of increment and time given at the start. E.g. 80 min T270 adjusted Fischer time is equivalent with 12 30f, 35 20f, 58 10f and 71 4f.

Do you find all these equally desirable as a player?

No. The bigger the Fischer bonus the better. That is my opinion. However there are some arguments that can be issued that 35 20f is best compromise of these.

One usual misconception with Fischer is that people think too much how long is the theoretical game length. Much more relevant issue is to consider what is the pace of the play that prevents time from decreasing.

With Fischer these values are as compared to Canadian overtime: 20 sec fischer = 15/5min Canadian, 15f = 20/5min, 12f = 25/5min and 10f = 30/5min. From these it is easy to see that 10 second Fischer is too fast pace for most purposes, because time pressure is too severe if player gets short on time too early phase of the game.

Liisa wrote: With Fischer these values are as compared to Canadian overtime: 20 sec fischer = 15/5min Canadian, 15f = 20/5min, 12f = 25/5min and 10f = 30/5min. From these it is easy to see that 10 second Fischer is too fast pace for most purposes, because time pressure is too severe if player gets short on time too early phase of the game.

I think that basic and bonus time should be balanced. 10 sec bonus time is quite OK when the basic time is also just 20 min. Bonus time also feels more comfortable than the corresponding canadian overtime, so 15st/5min canadian overtime feels actually more close to 18 sec bonus time, for example.

The bigger the Fischer bonus the better. That is my opinion. However there are some arguments that can be issued that 35 20f is best compromise of these.

Since you acknowledge the need for a compromise, I would like to point out that the factors that influence it are:

• forced pace to prevent time trouble,
• flexibility in time management for the player,
• and schedule security for the organizer.

My experience is that players do not like it when they do not have enough flexibility. This has happened in 20/30 settings, but also in 30/20. The time schedule is also more secure the smaller the time bonus is: 60 extra moves take 10 min with 10 sec time bonus, 15 min with 15 sec time bonus, and 20 min with 20 sec time bonus.

One usual misconception with Fischer is that people think too much how long is the theoretical game length. Much more relevant issue is to consider what is the pace of the play that prevents time from decreasing.

The overall game length is the first consideration, and the basic reason to play with a clock at all. Dividing it into basic and bonus time is an important measure to remove the sudden death problems. The forced pace resulting from this should have a relation to the overall game length.

I think that setting time controls to keep a schedule has become exceedingly simple since the advent of Bonus time: divide the overall time in minutes by 9 to get the time bonus in seconds, then multiply this by 2 to get the basic time in minutes. More elaborate steps:

• Set how long a round should take.
  Example: the rounds shall start 3 hours apart. I'll set 20 min apart for administration, so there are 2:40 h left for the games.
• Convert this time to minutes. This is the time you want a 300-move game to take at maximum.
  Example: 2:40 h = 160 min.
• Divide by 9. This value in seconds is an estimate of the time bonus.
  Example: 160 / 9 ~= 17.8
• Multiply this time bonus estimate by 2. This is an estimate of the basic time in minutes.
  Example: 17.8 * 2 = 35.6.
• Set the basic time and time bonus close to this estimate.
  Example: 36/18.
• Check the overall game length for 300 moves.
  Example: 36 * 2 + 18 * 5 = 162 min.
• Adjust to your liking. For each 2 seconds you change the time bonus, you have to change the basic time by 5 minutes to keep the same overall time.
  Example: one might like multiples of 5 better. This would lead to 30/20 (160 min), 40/15 (155 min), or 45/15 (165 min).

Note that you can divide by 9 quite quickly in your head: 160 / 9 ~= 16 + 1.6 + 0.16 ..., you really only need two elements from this. I do not think that forcing the values into multiples of 5 is useful, by the way.

I had claimed a few days ago that there is no time accumulated under Bonus time. Here is an explanation:

At any point during the game, a player has a certain time on the clock. I call this the time reserve. A player's overall remaining time is this time reserve plus the bonus time of his remaining moves.

Example: A player has a time reserve of 8:00 min, the bonus time for this game is set at 15 sec per move, and there are 40 moves remaining (half of which he has to play, half his opponent). This player's overall remaining time is then 8:00 min plus 20 times 0:15 min = 13:00 min.

At each move, the time the player has used is subtracted from his time reserve, then the bonus time for this move is added. His overall remaining time is then the new time reserve plus the bonus time of his remaining moves, which are one less.

Example: Above player uses 10 sec for his move. His time reserve is now 8:05 min. His overall remaining time is 8:05 min plus 19 times 0:15 min = 12:50 min.

The overall remaining time is always reduced by the time actually used. There is no accumulation and no spill.

Harleqin wrote:I do not think that forcing the values into multiples of 5 is useful, by the way.

Depends on the clock used. Excalibur2 has presets only for nearest 5 minutes. So it is pretty much necessary to use multiples of 5 as a basic time in the beginning. For time bonus this does not matter and bonus can be freely chosen.

---

But I would like to add some notion for those who like to estimate probabilities for actual game length accurately. With longest games there is typically increased the move count greatly, because there are lots of exchanges of last ½ point kou and many dames are played before counting. In most cases people do not use much time for these trivial but numerous small endgame moves. That means that often case in the end of long game there may be accumulated unused time more than 5-10 minutes for both players.

I think that this, that typically there is more time allocated for players for small end game than they need, is the reason why actual Fischer games tend to be slightly shorter than what can be calculated with using T300 value. Game might last 300 moves and even more, but there is spilled time left that has accumulated in the end. This is the reason why it is safe to estimate schedule 10-30 minutes too tight, what should be the theoretical value for the maximum time span of game. In large tournaments however, where law of large numbers will apply, this effect might be slightly diminished.

Liisa wrote: But I would like to add some notion for those who like to estimate probabilities for actual game length accurately. With longest games there is typically increased the move count greatly, because there are lots of exchanges of last ½ point kou and many dames are played before counting. In most cases people do not use much time for these trivial but numerous small endgame moves. That means that often case in the end of long game there may be accumulated unused time more than 5-10 minutes for both players.

I think that this, that typically there is more time allocated for players for small end game than they need, is the reason why actual Fischer games tend to be slightly shorter than what can be calculated with using T300 value. Game might last 300 moves and even more, but there is spilled time left that has accumulated in the end. This is the reason why it is safe to estimate schedule 10-30 minutes too tight, what should be the theoretical value for the maximum time span of game. In large tournaments however, where law of large numbers will apply, this effect might be slightly diminished.

Yes, people usually have at least 5 min left at the end of the game. I propose to use T300 because it is a reasonable upper bound of the game length; I actually feel confident to guarantee that a schedule will work based on it. The schedule is always dependent on the longest game of the round, and in big tournaments, there can be 150 games per round. The probability that one of the games actually goes to significantly more than 300 moves and the players use almost all the time is not negligible then. Having a conservative estimate then helps to keep the schedule.

I think that these estimates will be revised with further experience, so that less buffer can be chosen (it needs not, since a little pause between rounds is not necessarily bad). It might even be possible to schedule rounds three hours apart with 40/20 (T300 = 3 h). Since T300 can be so easily calculated in one's head, I think that it is a good guide value in any case.

Harleqin wrote:I have decided against attending the Kido Cup this year because of this, even though I live in Hamburg and enjoyed last year's tournament very much.
This so-called "progressive byoyomi" is one of the worst ideas ever. In fact, it is just a hidden sudden death.

I just noticed that main tournament in Kido cup is not qualified to class A, because adjusted time is less than 72 minutes. So, as a class B tournament these time settings are ok. There should not be reasons for boycott tournament, because people can consider it to be class B tournament with 70 min sudden. This is very fair time setting for class B.

But I agree that this application for progressive Canadian is not the best idea. My understanding thinks that progressive is like first 15/5min, 20/5min, 25/5min, 25/5min, 25/5min... This prevents people from playing half of the game in overtime like it is the case with 5/15min, but transition is smoother than to go directly to 25/5min.

henric wrote:* Availability of clocks is and will remain a decisive factor.

i just noticed and I find myself a little silly that I did not notice it earlier. That actually modified Fischer timing will suit for analog chess clocks even better than Canadian overtime!

There is just needed to do one compromise, that time is added to both players clocks after every 2×25 moves. The amount of time added should be 5-10 minutes, depending how sluggish time control we want.

And because we know that it is very rare that game will last less than 200 moves. First increment can be added on move 200. That means that in the beginning there is counted 100 stones for each and time is given some 30-60 minutes. Then on move 200 5-10 minutes is added to both players clock and it is counted additional 25 stones for both. And again on move 250 and 300 this is repeated.

That means that because it is very rare that games are longer than 350 moves, In practice there is only 2-3 times needed to add more time. With slow Canadian, if both players do enter on overtime early, there may be up to 12 times needed to add 5 minutes and count 15 stones in turn for both.

This will give same benefits for schedule regulation as regular Fischer. Also players may freely allocate their time and because time additions are accumulating, there is no spilled time and there is always possible to let safe buffer to flag.

Drawback is that there need to be made a compromise how often time is added to both players clock. This by definition makes time control less smooth than regular Fischer and active time management is required. But since keeping buffer is an option, time management should be easier and less stressful than with Canadian overtime.

If 200 stones feels as too many stones for the first period, this amount may be reduced to e.g. 2×75, but this also means that there is more need for adjusting clock.

How silly it can be to not to grasp such a simple idea. Too bad that since even Ing clocks do support Fischer timing, there is not much need for playing with analog clocks. Too bad that I did not grasp this 4 years ago!

As far as I know, the Ing clocks only have sudden death and "japanese byoyomi".

I'm going to have to find out if any tournament I go to will expect me to count out 100 stones at the start of each game as part of Fischer timing. Knowing the space in which tournaments are usually played, the thought of every participant counting out 100 stones is going to be a recipe for disaster. There is usually a clock, various cups and containers, recording materials and now a scattering of 100 stones on both sides of the board? Now I know that someone will come along, a second Mr Ing perhaps, with some ingenious device for counting and retaining 100 miscellaneous-sized go stones which every player will be required to own before they can take part in one of these tournaments, but count me out.

Best wishes.