Fischer Timing / Discussion

Sub-page of FischerTiming

The Little Golem turn-based game server uses Fischer Timing plus vacation days: 240 hours initially. Each submitted move adds 36 hours to that player's time limit. (Up to a maximum of 240 hours.) Additionaly each player has 20 vacation days (per calender year). If a game clock reaches 0 hours, that player's vacation counter is decreased by one day, and 24 hours are added to each of his current games.

mdobbins Even though it is called Fischer Timing it is not, as it is limited by the initial time limit. It is also not Bronstein Timing as that is limited by the time the previous move used. Dragon Go Server uses a similar scheme as Little Golem so we need a new term for the time scheme used by these turn based servers. How about "Limited Fischer Time"? Capped Fischer Timing seems to be the preferred name.


Dave Sigaty: One of its principal features is that (like byo yomi) the players never completely run out of time. There is rather more flexibility involved in deciding how to use one's time than with byo yomi.

Andrew Walkingshaw: Fischer Time, like Bronstein Time, is an alternative time system for chess designed to get games over with in a single session without having to resort to a sudden-death finish.

To understand this, chess games, like two-day Go title matches, used to be adjourned after one playing session (of six hours); the time control was thus two hours for each player's first forty moves, and twenty moves in each subsequent hour (apart from the restriction on the first period, much like Canadian overtime.) However, computers have killed adjournments by being plain too strong. Thus, chess moved to a time control of 40 in 2 hours, then 20 in one hour, then half an hour to complete the game: but some players objected to this last, sudden-death period.

Fischer time was suggested to counteract this. As Tamsin said, it works by each player starting with an initial allocation of time (say, 90 minutes) and receiving an increment (say, 30 seconds) *after* each move. The net result is that the players have the same amount of time (three hours total) for their first 60 moves, but will never get into the situation where a player has to, say, make twenty moves in five minutes: they will always have 30 seconds to make their next move.


Tamsin: I did ask the Blessed Admin Tweet whether IGS had any plans to introduce Fischer time. The answer was no. One can only hope that dashn and KGS will soon offer this excellent time system.


William Shubert (KGS admin): I personally don't plan on implementing fischer time, because it doesn't seem well suited to go. In fischer time, you can move fast in the beginning to build up a time pool that you use late in the game. For Chess, this makes sense, because as the game goes on you get into less-well-known areas and need to think more. In Go, most strong players think the most in the early and early-mid games, and move faster and faster as the game progresses because the decisions become easier[1]. Fischer time does not work well for that style of time use, so I do not think it will lead to better games.

PS - Just thought I should add, I've been proven wrong before on things like this, so if somebody demonstrates to me that fischer is indeed more fun than Byo-Yomi or Canadian, I may implement it later! But for now, it just does't look interesting enough.

Tamsin: If you want people to play on KGS, you should give them what they want. I'd like the option of playing Fischer time, please, and I'm sure I've spoken with others of like mind. So, I'm adding it to the KGS Wish List if it's not there already.

dnerra: To William Shubert: Let me try to explain why I think it is also very well suited for Go. You are probably thinking about the typical "1 min full time, rest byoyomi" game setup on Go servers. I agree this doesn't make so much sense with Fischer time. Instead, think of "30mins + 15 seconds per move" Fischer time replacing the "5*30 seconds" Japanese byoyomi for a typical 1.5-2 hour game on KGS.

I think it's very typical for Go that you often have whole sequences that are automatic, and then you have to think a lot about one or two moves. With Fischer time, you can accumulate the time during the former periods (instead of either throwing away your time, or spending it on counting the game or whatever), and use it for the latter. And still you know that the game will be finished in a reasonable time frame. And you avoid the unreasonable rushes at the end of Canadian byoyomi periods. (I have often enough seen 10 moves in the last 12 seconds of a 25 moves/8 min period, causing mistakes even by highest level players.)

So in summary: Fischer time allows you to spend your time on the interesting parts of the game, without forcing you to plan ahead your time usage (this is not s.th. I want to focus on during a Go game), without giving advantage to the player reaching byoyomi early, and while still guaranteeing that the game will be finished in a foreseeable time frame.


FHayashi: A time limit system used in Chess, presumably invented or championed by controversial chess genius Bobby Fischer. Not commonly seen in Go.

In Fischer Time, you start with a set amount of time, and you add more time for each move you make. Thus, you can build up your thinking time by playing moves quickly.

A very reasonable time system, as you can never run out of time as long as you are playing reasonably briskly, i.e. you never get in the situation where you have to make an impossibly large number of moves in a limited time span. -- Fhayashi

Hu: As dnerra points out, it is only in the Canadian system that one has to play an impossibly large number of moves in a limited time span, at board positions that should have nothing to do with the clock. In Byo-yomi play, there is a rhythm for each move, which encourages reasonable briskness.

dnerra: Yes, but byo-yomi is a little too regular for my taste - there are far too many moves on which you don't want to spend 30 seconds of thought at all.

Tamsin: You don't have to use all of the byo-yomi time on each move! If you like, you can move straightaway. The idea is that you get up to 30 secs or 1 minute or whatever on each move before that time is subtracted from your remaining byo yomi time. That said, I still think Fischer Timing would be good for go because it does allow the greatest flexibility in how you use your time. Come on Bill Shubert, give the people what they want!


kaiou: The only Go server using Fischer time is the Dragon Go Server, but I have never tried it.

uxs: I have. It's actually a mix between the Fischer and Bronstein systems: the increments do stack, but only up to a maximum. The maximum time you can have is your initial main time. There are other systems, but this one just feels right.


Benjamin Geiger:

I just bought an Excalibur Game Time II clock, and it offers a "delay" option, supposedly to support FIDE rules. Basically, an N second delay means that your time doesn't start running down until N seconds after your opponent presses his plunger. Is this Bronstein time? It's not referred to by that name...


dalf: "Fischer" time is the time system implemented on (almost?) all chess servers. Initially, it started there as a way to counter the lag (this was when Internet backbones were like 1.5 Mbps), but now real client time is used, so it is used as a real feature. In real life, it is now also used in serious tournament chess to finish the game (which used be 2 byoyomi periods: for the first 40 moves, the following 20 moves, and then adjournement), now that chess computers may ruin the idea of game adjournements.

It is still an excellent way to play fast/blitz/lightning games: with an increment of 1 to 5 seconds per move, there is no way to really build up a vast amount of time. That is unless your opponent is playing moves that don't worry you the least, in which case you are likely to play fast anyway to quickly finish the won game. In general that's why a player is not building up vast amounts off time he spent later: if the players are playing on equal terms (handicap etc...), both should be spending about the same amount of time on the moves - until the game is more or less decided. I think that time usage is similar in go as in chess: except for say, the 4 first moves in go, and 10 first moves in chess, the goal is to get an edge or to get back in the game, so all thinking power is needed to play cunning moves.

The big advantage over byoyomi is that there is less artificial move boundaries, so you don't need to do the following
  • see that you have plenty of your byoyomi time left for one obvious urgent move, so you start thinking for the move after, because anyway that time will be lost.
  • see that you have many moves to play and little byoyomi time left, so you avoid to play the best complex move, and play a more restrained move.

[1]: I disagree with that statement. Most amateur high dans (5D/6D) I know try to have plenty of time available until at least the early endgame. And this despite the Candian byoyomi (used exclusively in tournaments here) giving an advantage (in total time allowed) to the player reaching byoyomi earlier.

It's true that many low-dan players spend a lot of time early on, because they think they won't lose many points in the endgame anyway. They are wrong. -- dnerra

Neil: Perhaps he was referring to pro games?


OferZ: The nice little server that used to be on playgo.to used this time system and it was my favourite thing about it, only just now did I learn this time system's name. I never played on those servers with the extremely long games, but I liked this system for short games quite a lot. So I agree with everyone who supported it, especially with Dnerra's arguments. Personally, even with one minute per move byo yomi I sometimes find that I need more time in the endgame and it's quite depressing, since the endgame can be crucial and defending just cause you didn't have enough time to read when you don't have to can lose you the game. I know it's an old subject, but since there's still no such time system on KGS I guess it's not too late to ask again for one, as one who mainly plays on KGS and loves this server.


PeterHB: OferZ, you are the 20th person today, and I just keep telling them, there's no demand for it.


Pledger: I think we need some players of at least intermediate strength who have experimented extensively with Fischer time to share their impressions here. Only then can we begin to judge the system fairly.

axd: TBP fundamentally needs some way to cope with lags due to time zone differences - see also time systems and below.


Salvaged from [ext] http://www.hackershack.co.uk/go/fischer.html with permission of [ext] James Taylor:

Intro

Of the three time systems offered on the Dragon Go Server, only Fischer time should be used. The other time systems are designed for real-time play and are not suitable for Dragon's turn based play. Unfortunately, most people are not familiar with anything other than Japanese byo-yomi, so they just use it by default without giving any thought to whether it is appropriate. I hope the following explanation will help bring much needed enlightenment. How it works

Under the Fischer time system the "main time" is a pool of available time which is increased after each move by the amount given in the "per move" setting. The pool of main time starts at its initial setting and reduces with time, just like the other time systems. However, after each and every move the pool has the "per move" time added to it. This is limited to a maximum such that the "main time" may not increase beyond the initial setting. This allows each player to play more often or less often at whim as long as the average frequency of play is better than the "per move" time. Furthermore, if a player has been playing within that "per move" frequency they will have the full "main time" in reserve to cover them in the event of needing to leave the game for a few days. You can think of the "per move" setting as meaning "average move frequency" and the "main time" setting as meaning the "maximum period of inactivity".

Timezone fairness

The beauty of the Fischer time system is that, when the "per move" time is a whole day or more, it no longer matters what timezone each player is in because, regardless of what time of day they typically get online, they will always be able to keep their main time at maximum. In contrast, the other time systems penalise a player that, for example, moves once a day at 7:00am in favour of an opponent that responds at 8:00am each day; this is because the second player only loses one hour of main time each day whereas the first player loses 23 times as much main time each day! Japanese byo-yomi is fine for a real-time game, which is what it was designed for, but totally unfair for Internet turn based play on Dragon. It baffles me why so few Dragon members have woken up to this yet.

Abandoned games

In addition to the issue of timezone fairness, Fischer time is far better at terminating an abandoned game without simultaneously restricting total playing time. With the other time systems it is necessary to allocate a large amount of main time in order to allow leisurely play, but this has the unwanted side effect that an abandoned game hangs around in the system for many months! In contrast, Fischer time does not restrict total playing time at all, it only ensures that players move regularly. Fischer "main time" can be set to something quite short, such as four days, and an abandoned game would terminate after four days of inactivity without restricting the total playing time of active players in any way. I believe this is exactly the behaviour one wants from a turn based time system.

My own preferences

Personally, I choose a "per move" time of one or two days for the sake of timezone fairness, and a main time of at least a week to allow for the short trips away from home that I frequently make (usually two or three days at a time). Dragon's stupidly inflexible vacation system simply cannot cope with these short trips because it does not allow a setting of less than a week, does not allow cancellation of a vacation once started, and only allocates 30 days per year which would cover only four of my trips. With opponents that I know and trust I tend to select a very generous two weeks of Fischer main time which allows for a longer holiday, or several of my short trips in quick succession as long as I play frequently in between.

Feedback

Readers may contact me as James Taylor on Dragon. I would be keen to hear about any mistakes I've made or any way I might improve the above explanation. Thanks.


PeterHB Copied from [ext] L19x19 with permission of [ext] Charlie:

There is a formula for finding good Fischer Time parameters:

  • Identify a desired total game time. For example: 1 hour or 60 minutes or 3600 seconds.
  • Decide on a main-time length: perhaps 10 minutes or 600 seconds.
  • Subtract twice the main-time length (there are two players): 3600 - 600 - 600 = 2400
  • Divide the remaining seconds by the rough average number of moves in the game -- about 250 for 19 x 19: 9.6 seconds per move.
  • Do some rounding: 10 minutes + 10 seconds per move should split an hour fairly between two players.

This, too, is a big pro for Fischer Time in my opinion: you can easily employ the time system to split an allotted expected total time fairly. If that is not the very reason for existence of clocks in the first place, I don't know what is!



Fischer Timing / Discussion last edited by PeterHB on January 30, 2018 - 22:58
RecentChanges · StartingPoints · About
Edit page ·Search · Related · Page info · Latest diff
[Welcome to Sensei's Library!]
RecentChanges
StartingPoints
About
RandomPage
Search position
Page history
Latest page diff
Partner sites:
Go Teaching Ladder
Goproblems.com
Login / Prefs
Tools
Sensei's Library