KGS Issue - Fischer Discussion

    Keywords: Online Go

From KGSWishlist/rejected, please read this before reading this page

wms: There has been a lot of discussion on Fischer. Some people love it. But at this time I think it would be a mistake, and I do not plan on implementing it. See the fischer discussion page for why.

A heated discussion, moved from KGSWishlist

dnerra: I vote for Fischer Timing as an additional option. (Yes I know it's not well-known among go players, but I'd bet a lot that it would become popular quickly.)

Tamsin: I second that request.

joshual000: Fischer Time would be my time control choice if development time allowed.]

wms: 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. 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 doesn't look interesting enough.

Dave: Thank you for not closing the door completely wms! Please have a look at Timing Systems - Redux. I have given it my best shot. I am very interested in what you think. Hopefully we can look at it once again with less "heat" this time. :-)

Magdirag: just to increase the demand a little bit: I think Fisher Time would be quite a good addition and it would probably be my preferred time system.

mAsterdam: Fisher_Time_preference++ . Oh well I have no right to speak. I haven't been playing at KGS lately. Besides the weekly game at the club I haven't been playing at all!. But I intend to - play more and play at KGS again.

Evand: another vote for Fischer Time here.

ChessWhiz: And here! It's been implemented on the DGS, and it's one of the most popular time controls. My favorite, in fact.

MikkoSaari: I very much prefer Fischer Time on DragonGoServer - I think it's much better than Japanese B-Y for e-mail play. However, I'm not that sure it's the best choice for real-time games.

TJ: Just to be contrary: Fischer time controls, yuck! I don't like Canadian overtime either, but I like normal B-Y overtime in go much more than I liked Fischer time when I used to play chess. Maybe proponents of fast games would like Fischer better, but I wouldn't.:) It at least makes better sense in chess, where openings are pretty much set pieces and you're always likely to gain time if you know your book. Anyways, just wanted to throw in the opposing side, lest it be thought it didn't exist.

jfc: this is a fallacy. You can provide players with lots of time up front using byo-yomi, canadian or fischer time. All 3 of these time systems have a maintime component and an overtime component. Balancing maintime and overtime is a choice made by the players.

If you choose a

  • fischer: maintime = 0, overtime +50 seconds per move

then yes, you will have more time to think at the end of the game than at the beginning but this not any worse than choosing

  • byo-yomi: maintime = 0, 5 periods of 50 seconds
  • canadian: maintime = 0, 12 moves in 10 minutes. Canadian.

The debate isn't "how much time does the system give me during the beginning, middle and endgame?" the debate is "how does the type of overtime force me to play?".

Evand: TJ: is that a vote to not include the feature, or a statement that you wouldn't use it? Is there any reason not to include it other than the fact that it would take wms a little while to implement?

TJ: It's an opinion on and a reason for why I don't think Fischer time and Go go together. Of course, life would be easier for me if there weren't an option for people to play any time controls I don't like, so I do think I'm entitled to expressing that opinion.*grin* So, it's just me saying that there are people out there (at least one) who agree with the opinion attributed to wms: that the utility of Fischer clock for go is, and should be, in doubt. The fact that it would take effort for wms to implement is ALWAYS an argument against anything on this list taking priority over anything else, an argument easily taken up by anyone with a contrary opinion, so I won't argue that. Although I'm sure it's an issue much on wms's mind, I wouldn't presume to prioritize directly on that basis for him.

wms: I'd like to add that it's a common misconception that adding a feature is a fixed cost (ie, once it is added, you are done). It is not a fixed cost. Every feature added to any software system, no matter how well written or how simple, adds to the complexity of that system, and means that all future changes will be a tiny bit more complicated. Ever worse, adding fischer time makes the UI more complex; people have to learn what fischer time is or else they may end up in a game where they don't understand the clock. In addition, I think that something like an added time system are very hard to be removed once added, because some people will be accustomed to using it even though it isn't particularly better than the other time systems provided. All these add up to reasons why I wouldn't add a new time system unless there was really strong evidence that it would be better in some way than the existing time systems, and for Fischer time that evidence simply isn't there. (Plus of course my personal opinion on looking at the time system is that it isn't very good for go, as posted above).

ChessWhiz: You're right, of course. Sorry, I now realise that a feature is not just a feature -- it increases inflexibility, compile-time, .exe size, and complexity. I think Fischer time works well for turn-based go, such as on the DGS, but real-time go is probably better as Canadian and Japanese byo-yomi. Hey, we're still one step ahead of IGS! :-)

Tamsin: They're not very convincing arguments, Bill, if you will pardon me for being blunt. The truth is that some people would like it, others wouldn't. It wouldn't take very long for newbies to learn what Fischer time is (as it happens, I found the way you initially implemented traditional-style byo-yomi confusing when I first met it), and it can't be inherently any less suited to go than Canadian byo-yomi or any other time control system (how on earth can anybody declare it's not suitable for go without even trying it?). If you don't want to add it because you don't want to add it, that's the best reason, as it's your server and your work. But adding Fischer time would only make the server better by increasing peoples' options. People who don't like it would decline games with that kind of control. On a final (still more controversial?) note, I'd advocate adding Fischer time for the purpose of getting ahead of the game -- I doubt it will be very many years before it is tried and accepted in real-life tournaments. Better to be a trendsetter than a dedicated follower of fashion.

dnerra: Tamsin, you know that I have the same opinion with regards to the technical merits of Fischer time. But keep in mind that wms not only would have to do the work to implement it, but also would hold responsible for the decision, as he has been responsible for all previous design decisions for KGS. And that he has been very successful in making KGS a good server, so he has been right probably more often than wrong.

wms: OK, Tamsin, I'll be blunt too. In go, most players think a lot in the beginning. A lot in the middle. Less in the end. So saving up a pool of time to use at the end is stupid, and that's what Fischer time is all about. Byo-yomi and canadian times are all about giving you a block of time you can use up early, then forcing you to think faster once that is used up. This matches the way people use time in go, fischer doesn't, so fischer seems dumb to me. So there we go, it can be argued that it is inherently worse than other time systems, without even trying it out. Furthermore, you say adding another time system wouldn't be hard for users - I disagree, every little bit of extra crap you need to know before you can play comfortably makes the server a worse place. You say you had to learn one time system to play, what if you had to learn 2? What if it was 5? The more time systems you must learn, the worse things are. Right now there are 4 time systems, that's probably too many if anything, so I'm really reluctant to add a 5th, especially if it doesn't seem to match the way people play go.

Tamsin: Steady on, Bill. You don't have to use words like "stupid", "dumb" and "crap". If you don't want to implement Fischer time, then fine, but there was no need to put it like that.
You make out that Fischer time is only one of many alternative time systems, i.e., as though it were just another gimmick or fad. That is not the case: it is now one of the principal systems used in professional chess, and will doubtless catch on in go too before long. It is a superb idea in its own right, and deserves consideration for that, if nothing else.

wms: I apologize Tamsin, I should try to be more moderate. It just irritated me because your answer seemed a content-free criticism; to my reading, after explaining why I don't want to add fischer time, you essintially said "No, you're wrong." You gave no reason why you thought fischer time was good (now you have said because chess uses it, which seems odd to me since clock use in chess and go are so different). And you out of hand discarded my argument that extra time systems make the server worse, even though you agree you had to learn a time system yourself by joining the server. Also, fischer is not the only time system I've been asked to add, it is just the one that has gotten the most attention recently - I've had at least 3 or 4 different time systems recommended, I can think of "asymmetrical time controls," "hourglass time," and "Ing time" off the top of my head as proposed changes to time systems.
As a final note, this fischer discussion is getting way out of hand. Please, the next person to add stuff, move it to a different page if we have to continue it so it won't clutter up the wish list. :-) If nobody feels a need to add stuff, we should just chop it out after a few days I guess.

Tamsin: The great advantage of Fischer time is that it prevents players ever getting desperately short of time and making silly mistakes because of that. In Canadian overtime, for example, I have often been in situations where I have had to make 5 or moves in a minute, which is a stress. No matter how generous the rate given for Canadian overtime, there will be time scrambles, and that is bad news quality-wise for both chess and go. I prefer Japanese byo-yomi, but again it can be a nuisance having to make 1 move every 30 secs or every minute. Fischer time enables one to build up time and use it when you need to, without being forced into a scramble situation (Canadian) or one of constant, unending pressure until the end of the game (Japanese).

mgoetze: Eh? People can play entire games in 10 minutes without breaking a sweat. 5 moves in a minute is not the end of the world. Especially not in yose.

As for making the server worse, don't you feel that you might be underestimating players' abilities a little? I'm not the sharpest tool in the box when it comes to that kind of thing, but even so it would not take me more than a very short while to pick up as many as five extra time systems. Besides, a large number of go players from the West come to go from chess, and will have heard of numerous alternative timing systems as part of that experience. I hope that that was a little more persuasive, but thanks in any case for all that you do to make KGS such a fun place. Good luck with it.

Neil: In my experience playing games with Chinese rules, I'm surprised at how many users don't notice those combo boxes in the game dialog. You could probably add Fischer time, Bronstein time, Time After Time, and Howdy Doody Time, and most users wouldn't even notice, let alone be hurt.

Tim Brent: I'd have to agree that Fischer time wouldn't work for Go, as there is an inversion to how Chess, which the system was designed for, and Go work out (i.e. the quickest part of Go is the endgame, Chess the opening). I think the systems in place now (no time,absolute/no byo-yomi,Japanese byo-yomi,Canadian overtime) are more than sufficent, in hat he great majority of players will either use Japanese or Canadian for their games. I also read the complaint about how byo-yomi, Canadian, and of cousrse absolue cause time pressure, that's part of the game. I mean, some moves have to be instinctive and time pressure is good at developing instinct as well as forcing you to learn to read and diest the board in a fairly quick manner. The current systems are fine as they are and don't need to be added to. -- Tim Brent

dnerra: I respect and understand wms's decision not to include Fischer time in KGS, so I am replying here only to discuss the benefits of Fischer time in general. I really don't understand why "playing slowly in the opening, quicker later on" runs contrary to the conception of Fischer time. In a Fischer time game of go with, say, 15 mins plus 10 seconds a move, you can in the beginning take much longer than 10 seconds per move, thereby mostly living from the 15 mins, and later you speed up more and more as your time is running out, until you play the endgame just at the 10 seconds per move (but still having the option to play out some sequences quickly, so that you can read out some harder problems in detail). It seems a much smoother transition to me than, say, in 15 mins plus 30 seconds Japanese byoyomi (which would approximately take equally long). Plus Japanese byoyomi has the disadvantage that the player who runs out of the regular time first gets a benefit, by getting more total time than his opponent for the complete game.

Bill: This is not to suggest that wms should do anything, one way or the other. But I find that my thinking is much in line with dnerra's. Back when I played in tournaments, I steeled myself to play in Japanese byoyomi at 30 sec. per move. I would spend most of my time on the opening and enter byoyomi in the large yose stage. One problem, I felt, with Japanese byoyomi is that you tend to use up the whole 30 sec., even if your move is obvious. With Fisher time, I think I would be comfortable with, say, 10 min. with 10 sec. added per move, in casual play. It is not that I would build up a time bank by rapid early play, and if my opponent did so, I would probably be pleased. ;-) With Fisher time I would not worry about losing time if I did not use it, nor would I face the additional strain of budgeting a set block of time that comes with Canadian byoyomi. Instead, the whole thing sounds rather smooth, like Japanese byoyomi without the drawbacks. I think Fisher time would be fun to try. :-)

jfc: exactly! With byo-yomi (and to a lesser extent) you must learn to adapt your playing rhythm to the clock. Fischer timing overtime does not penalize you for playing obvious moves immediately.

While all time systems force you to divert some of your attention to clock management, Fischer Time requires the least amount of unnatural changes to your playing habits to allow you to get the full use of your time.

Of course I also agree that, on the whole, wms's benevolent dictatorship has been a stunning success! Fischer Timing or no Fischer Timing, KGS remains my realtime Go server of choice!

blubb: Fischer time mainly ensures a minimum average speed, but with additional xx minutes "joker time" (main time, that is) to be used arbitrarily whenever needed. I was about to write a few more things on this soon, but since I nearly completely agree with you, just my "me too".

Joshual000: I also must respect wms and his decision not to include Fischer time, I must however argue that Fischer time is superior to Canadian time. The *only* difference between the two is under Canadian time you lose unused time when entering a new period. Under Fischer time, the unused time is conserved for future moves. Whether or not this store of time is beneficial most players, I don't know. I do know I've lost games on time using the Canadian time control.

That said, I develop software for a living. If the time control portion of the KGS server code (C/C++ if I remember correctly) is inflexable as it appears, it could potentially be very painful to add new time controls. I believe wms is fully qualified to assess the cost of adding such a feature versus the potential benefit. The manner in which he is defending against additional time controls tells me the cost of such a feature is not trivial. Maybe when the source is ported to Java :) -- josh

yoyoma: My opinion is with wms. Every feature added is not only a cost for the programmer, but also for the users. Already with the 4 there are now I only like one, Japanese. None is great for people who don't want a formal time. But for the other 3, it takes experience to get a feel for how the time works. Absolute is the largest challenge to me, how do I account the amount of time I need for the entire game? Canadian is next hardest to me; during a hard problem I find myself looking at the clock and trying to do some sort of fuzzy division in my head to figure how much time I can afford to spend on the current move. With more experience I would be able to get a feeling for that and do it without effort, but it costs several games to get this feeling. So I prefer Japanese because to me it's the least taxing to figure how much time I can spend on a move. Even this time system required some time to get a "feel" for. I played many games before getting a feel for when I can afford to use one of my periods. When looking for a game, I often bypass the games offered with Canadian simply because I would prefer Japanese. And I don't want to see more time systems introduced to further fragment the player population. So my argument is not so much based on Fischer being better/worse than any one time control. Instead I argue that the 4 we have are sufficient, and Fischer is not so much better than the others that warrants it being added or replacing an existing one.

Rich: I have to side with wms too; an extra feature is extra work and extra maintenance, and should not be added on a whim. I also by no means shar Tamsin's optimism that it is merely a matter of time before Go Tournaments start using it; timing is far more game-specific than, say, tournament structures (that have crossed over). For example, you could have players who use up their time, then play several forced sequences, ending up with a lot more time than their opponent, who now has fewer sequences left to play out. Going into penalty time should increase the pressure, in my opinion; people who want leisurely, thought-out games can play to different time settings than people who like the scramble of lightning Go; players who dislike the fact that they make suboptimal moves when time is short should either learn to ration their time better, or learn to accept that this happens for everyone, up to pro level, and work on how they deal with the pressure. In short, I see Fischer as an unnecessary addition.

DragnSlayr: Every feature added to any software system, no matter how well written or how simple, adds to the complexity of that system. Furthermore, having another time system is hard for users, every little bit of extra crap you need to know before you can play comfortably makes the server a worse place. What if you had to learn 2 time systems? What if it was 5? The more time systems you must learn, the worse things are. Right now there are 4 time systems, that's probably too many if anything, so it would be better to remove one and have only 3, especially if it doesn't seem to match the way people play go. Therefore Canadian should be removed. It is not disigned for Go, in Go, most players think a lot in the beginning. A lot in the middle. Less in the end. So having a set amount of time for a set amount of moves, regardless of the progress of the game is stupid, and that's what Canadian time is all about. Byo-yomi and Fisher times are all about giving you the time you need when you need it, not forcing you to think faster at arbitrary points. This matches the way people use time in go, Canadian doesn't, so Canadian seems dumb to me. So there we go, it can be argued that it is inherently worse than other time systems. And allmost all of us have tried it out. Canadian should be removed.

uxs: I actually agree with the previous editor: after all, the Canadian system was only invented because: a) clocks that were made for chess couldn't do japanese byo-yomi and b) most normal people can't afford to hire time-keepers to count for them. (And while normal (chess) clocks can't expicitly do Canadian byo-yomi either, it's easily rigged by taking X stones out of the bowl and adding Y minutes to the clock. The Japanese system isn't so easy to simulate.)

mgoetze: You forget that Canadian overtime only steps in after a (possibly quite long) initial time. I could copy your arguments word for word to argue that Japanese byoyomi should be abolished, and it would be just as silly. (The set amount of moves would be 1.) It always seems to me that the people who dislike Canadian overtime are a very small and very vocal minority - in my poll of league players so far, a large majority have voted for Canadian time. [limited, unscientific poll]

DragnSlayr: So Japanese Byo-yomi, Canadian and Absolute are basically the same time systems. This being Canadian with 1 move per time unit and Canadian with no time per x moves. This would mean that removing Canadian time will not solve any complexity problems. Fisher can start with a possibly quite long initial time to. And thus most of the until now not marked as silly arguments against Fisher time are indeed silly. The only serious argument would be to keep an eye on how many people want Fisher time and compare that to the added complexity of the system.

Stormer You know, I just really hate time pressure, so absolute time, and canadian really put me off. I end up using regular byo-yomi of 1 minute, but that tends to scare a lot of opponents off. Since the moves speed up towards the endgame, I can understand why fischer time might not be considered ideal. But what about just making up a whole new timesystem?

It would work similarly fisher time, except you could never "gain" more time than you originally started with. For example, lets say you start with 10 minutes main time, and also get 20 seconds a move. You play through part of the opening rather slowly and drop down to 5 minutes time. Eventually you get into yose and the game speeds up somewhat, you manage to catch up and eventually end up with 10 minutes main time again. Then suprise! your opponent plays some unexpected tesuji, and you actually have 10 minutes to think about how to play it out.

Seems to me this would have some of the advantages of canadien time without the "whoops 1 minute to make 10 moves" drawback of canadian.

Mef: Allow me to preface this by saying I agree that 4 time systems is enough and KGS has no need for Fischer time. However should they add Fischer time I think it could be and interesting learning tool for weaker players. I know people like to think more in the opening, but in all honesty I'm never really sure that the extra thinking does any help for me at all. But later in the game, about the time I would be entering byo-yomi, when I start playing faster, is the time when more time could be good because it would encourage me to start using this extra time to do things like count the game. That's one of the few things that I at least know is being helpful when I use my extra time, and it's something I never take the time to do when I should...

Also as a random sidenote to the people who don't like budgeting their time in an absolute game: The key to playing absolute time isn't having enough time to finish, it's just having more time at any given point than your opponent...which is why I too hate it...

Alex Weldon: I wanted to say something somewhat related to Mef's comment. Although what wms has to say about Fischer time being inappopriate for go is logically correct, he's making an assumption that needs to be clearly stated, which is that we're talking about reasonably strong players. Up until about 15 kyu, most players find that the game gets more and more confusing the more stones there are on the board, rather than less. Also, for double-digit kyus fuseki is of comparatively little importance, as any advantage reaped there is lost in the "noise" of middle-game fighting. Therefore, although Fischer time would probably not be used much by single-digit kyus and dans (except maybe for blitz games... I could see it being useful there), it might be useful to beginners.

Also, Fischer time would greatly reduce the use of "time tesuji" kikashi, since you'd be adding time to your opponent's clock as well, each time, and not gaining that much yourself.

(Sebastian:) Another supporting argument: Fisher time is good for playful games. Byoyomi unduly punishes a player who takes more time on a particular move, and there can be many such situations (particularly in mid-kyu non-tournament games): Netlag or disconnection; roommate or cat needs some attention; phone rings; territory needs counting; maybe you just type a question to your opponent or you're just awk for two minutes. Why punish players who don't spend the same time on each move? Fisher time is just fairer; it treats lost time the same, regardless of when it is lost.

Robert Pauli:

  • Agree to dnerra's "benefits of Fischer Timing in general": FT meets Go as well. Sorry, wms, but as far as I see you ignore its fixed time block, which indeed can be used in the fuseki.
  • Keeping the feature list small is an argument, yes, but this can't mean that old stuff blocks new. Guess what most want isn't necessarily Fischer, but simply a "no time spill" option.

(Sebastian:) What's that?

RP: For instance Canadian Overtime, but if you've done your load of stones, instead starting next slot and load, thereby spilling the unused time of the current slot, you continue on that slot until it's used up -- without a charge of stones now, OC. Takes away "let me think twice - have to exploit my time".

uxs: A comment about the Fischer time control on DGS: it isn't. It's more a mix between Fischer and Bronstein, in that you can end up with more time after a move than you had before your move (= Fischer), but you can never have more time than you initially started with (= not really Bronstein, but close). Incidently, I like this version of Fischer time on DGS, and specifically with a main period of a few weeks, and increments of a few days. If you play reasonably, running out of time is unheard of, but if you stop playing for weeks, you get timed out, in a few weeks. For real-time play, I like Japanese Byo-Yomi best.

Zinger: Okay, a short summary: A decent portion of players seem to like Fischer time. However, it isn't really necessary, and it's burdensome to implement. Ergo no Fischer time. This is all the info I really need about it.

As for the merit as applied to Go, I also reject the "think slow early" argument, as the inital time could be user-set. One drawback might be: incentive to waste ko threats. You get short of time in a tough reading situation, so you fire off several ko threat-type moves to gain time. These moves are obviously bad aji-keshi. In Japanese byo-yomi, there is less incentive to do this since you can never gain more than 30 sec (or whatever) for a move.

jfc: regarding waste ko threats. But player's who manage their byo-yomi and/or canadian well do do this. Even pros do this. Pros also play kikashi as the sealed move for two day games. The difference is that because byo-yomi and canadian are harder to adapt to you may have to waste more threats under these systems to buy the amount of thinking time you need.

Calvin: Actually, there are a lot of games where extra time in the endgame would be useful. Close games are one case. But in particular, it would be handy when playing against an opponent who has the habit of playing speculative invasions in the late endgame when behind. It takes more time for me to refute bogus invasions than for my opponent who is just messing around playing random kikashi in my territory, hoping for me to either run out on time or make a mistake under time pressure. So in games where there is main time, I like to get to the endgame long before byo-yomi starts, so that I can have more time to deal with this rather rude habit. Fischer time would of course make this easier.

Mark Galecki?: I think the strongest argument for Fischer clock is that:

  • It is elegant/simple (like Go itself!) and logical
  • Other popular systems, in my opinion, are less so.

Let me explain. In Fischer clock:

  • you get a certain amount of time for each single move (this is the elegant and simple rule)
  • unused time is not lost but is added to the time available for later moves (this is the logical part)

(What I described is indeed the simplest and most elegant form of Fischer clock, where there is no initial extra amount of time, just the time alloted for the first move. If more time is desired for the first moves in Go, which may require more time, one can start with some extra initial time.) In other clock systems:

  • either there are various amounts of time for various moves or moves are artificially combined into blocks (in my opinion, not elegant or simple)
  • unused time is wasted (in my opinion, not logical - does not make common sense, and, if the player wants to use all their time, which makes sense for competitive games, they have to artificially slow down their game during certain moves).

Cyclone001?: Add me for a vote for Fischer Time. It makes the most sense to me too. The very idea behind having a clock at all is to keep the game moving. Time pressure, I think we can all agree, is unpleasant for players. Fischer time solves these issues better than other systems:

  • Byou-yomi - let's face it some moves take longer than others. If you have 30 second byoyomis for example, the 30 seconds is way too long for some moves, and way too short to read out a difficult life/death problem.
  • Canadian - It is common that complicated fights have 10 or more moves that require extensive thought in the difficult fights, then speed up significantly. This is a big disadvantage when the majority of a tough fight takes place in one time slice. I don't know how many times I've seen people have to play 7 moves in 10 seconds after a tough fight - it really takes away from the game.

By allowing you to, in essense bank time on easy moves, you can take your time on tough sequences and still finish the game in a resonable amount of time. As for the argument that the game should get quicker at the end, that can be mullified if the players wish, by setting the added time to be short, and the main time longer, in essense turning it into a better version of a long normal time /w short byou-yomi game.

dnerra: Let me add one more statement here, as I have since played a couple of blitz games (real-life) with Fischer time (5 min + 3 sec/move). I have played many blitz games with sudden death (10 min), byo-yomi (usually 5* 10s/move), Canadian (25 moves in 2 or 3 min), and a few in Fischer time. I definitely liked Fischer time best by a bit margin; it is the one where I can focus almost only on playing go, instead of having to worry about using my time well without running out of it. IMO a go server could well do with Fischer time and no other time system.

Harleqin: I also strongly support Fischer time. It is the most elegant system that solves the problems each time system seeks to solve:

  • The players shall be restricted in the time they use, so the game doesn't last forever.
  • You do not know how many moves the whole game will take (in go, between 40 and 400).
  • Moves have wildly varying complexity and thus also the appropriate considering time varies.

Fischer time uses a formula in which the overall thinking time is linearly dependent on the number of moves while allowing the players complete and effortless control over how much of their time they use for each move. Most important, they do not lose time when they play faster. This is completely independent on the type of game, and how the time usually is used.

The only argument I see against using Fischer time on KGS is the aforementioned increase in the software's complexity. But I'd advocate scrapping all other time systems anyway ;o).

Ectospheno: I'll just put in my $.02 and say that Fischer time rocks. Its about the only time system I use on my chronos clock.

SirLyric: As a double-digit kyu I have recently found myself wishing for a Fischer-like time setting as well, and am looking forward to picking up a Chronos so I can try it out in some RL games. My impression is that while strong players think a lot in the opening and middle game, and relatively little in yose, players like myself run something near the opposite. At my level I tend to play a quick opening unless things get complicated, because I don't see the variations and potentials that occupy stronger players' minds at that time. But I play a very slow yose because I'm trying to build my skill in that area and I'm still slow at counting the moves. For me Fischer time would feel a little like an adaptive Canadian system. The quick exchanges generate time on the clock to be used when thinking on your decisions, so the time naturally divides itself into periods that mirror the locality of exchanges on the board. I find it rather elegant.

Harleqin:I'd like to stress one point here: Fischer time is not necessarily about building time pools in the beginning. For example, in a normal real-life tournament setting, I'd use something like 30/30, meaning that each player has 30 minutes in the beginning, and adds 30 seconds after each of his moves. Assuming a game of 240 moves, it will last three hours at maximum. No player will ever have less than 30 seconds left for a single move, and that is the only restriction on how he may use his overall thinking time.

IanDavis I don't see the point of this system. The game is supposed to be played in maintime. Overtime is a nicety, the game is not really supposed to be played in it. Although having said that internet players often prefer systems like Canadian because they can dictate how long the game will last. Fischer time sounds to me to be the sort of system that can make the game drag on forever, which i loath. I actually prefer the Ing way, where you get penalised for each overtime period you use.

dnerra: In my view, exactly the opposite is true. In Canadian or Japanese byo-yomi, you get an ADVANTAGE by getting into byo-yomi very early in the game -- you will be allowed to use more time overall. (And in high-dan amateur games, I have often seen people getting into byo-yomi around move 60-70. These games loathe forever.) Of course, if you replace say 60min + 20 moves/5min Canadian byo-yomi by Fischer time, you should reduce the main time. About equivalent would be 40-45 min + 15 sec/move Fischer time -- this gives the same total length for games where people would enter byo-yomi around moves 120-160, but avoid the extra length of games among the specialists who start byo-yomi shortly after the first joseki. Anyway, I hope the number of the few enlightened people will continue to grow, the revolution is coming, around 2020 all server games will be played Fischer time, around 2030 most professional games, around 2050 the Japanese big title matches, and by 2060 historians will dig up this page, and it will be famous as the first public effort to make Fischer time more popular in Go.....

SirLyric: While I'm not quite that hopeful, I agree with you - heavy use of overtime is very common in pro games. I often see commented pro games saying things like 'Black, as usual, is in his last minute of byo-yomi...' and it's move 100.

Harleqin: I strongly object Ian's notion that overtime is a nicety. It is not, it is a necessity. Without overtime, players have to speculate about the remaining number of moves and their complexity. If you penalize players for taking overtime or give no overtime at all, this can (and most probably will) be abused by artificially prolonging the game, e.g. starting speculative invasions. Time should never be a weapon. There are players who reserve a few frivolous overplays for playing when their opponent has not much time left. I can hear "then don't get into time trouble", but this attitude leads to the destruction of the game, since the only way to win under such conditions is to find the harsher overplay, the trickier hamete, and with that, luck. Mind you, it is a "mind" game, not a "not mind" game. OK, sorry for the slightly offtopic rant, but I hope that I made clear why overtime is necessary. "Overtime" is also not a good word for it IMHO, since the distinction between main time and overtime leads to misconceptions. Fischer time does not use this distinction.

Fuchsnoir: Neither does [ext] Byo-Yomi

Harleqin: True, at least for the original japanese time keeping system. On KGS, a non-granulated "main time" phase is used before the second "byoyomi" phase.

That said, I think Ian might be pleased by the fact that with Fischer time, the end is "hard", i.e. a game with 30/30 Fischer time and 240 moves can never take longer than three hours, with 30/15 Fischer time it could never take longer than two hours. Just to repeat, the main advantage over other time systems is the high flexibility in time management while giving a hard end time which is linearily dependent on the total number of moves.

Chris Hayashida: Where is Fischer time supposed to have advantages over the other time systems? First of all, if the basic time was large enough, all of this would be irrelevant.

Harleqin: Without a dependency of the total time on the number of moves, absolute time is either too small or meaningless.

So is it better for blitz games? Or am I missing something? I sort of feel that blitz games that are decided by the clock aren't "real Go" in my opinion anyway, so it just sort of seems like another way to get a time win. ;) It also seems to sway the game in favor of the person already leading on time... If you play faster than your opponent, you'll already have more time left, but your time remaining will also be increasing. It doesn't seem balanced, to me. The winner wins by more.

Harleqin: Since both players play alternating, both also get the same time added (you can never play more moves than your opponent). The fact that the player who plays faster has more time on his clock is unavoidable and obviously reasonable :) .

If it was implemented, I doubt I'd use it. And since we're worried about finite time, I'd worry more about wms's finite time to work on KGS. I'd rather have him spend his time coding for the server for other things besides Fischer time. I definitely wouldn't want him to waste time defending his decisions on Sensei's. :P

Harleqin: Chris, as I see it, this page is about finding a conclusion, or at least bringing all relevant arguments with respect to Fischer time, so wms can look at it without having to discuss himself.

Perhaps you can see at Fischer time this way: the total time, whose exact amount is unknown at the start, unveils at a certain amount per move. At the same time, players are forced to play at such a pace that their remaining time per move can never fall under a certain amount.

Mef: In essence, Fischer time is just Byo-yomi, only it takes longer since all the little bits of time you don't use add up, whereas in byo-yomi they are discarded.

Harleqin: Why should you use the same nominal settings? Also, when the time you don't use is discarded, it is always an advantage to use all the time. When playing with Fischer time, there is no benefit from using more time than needed for each move. Since in go the last few moves are usually played rather quickly, I'd expect the players to always have several minutes left at the end of the game. This should reduce the overall time about to the same amount as the beancounting effect of discarding bits and chunks of the players' time during the game.

Mef: Indeed, and if it's comparable to Byo-yomi, is it really a necessary addition?

Harleqin: That is a good question, and my answer is "yes" :). You can find a mean total time for all time systems and thus, they are all comparable by that. The important difference lies in the details of how players have to use their time:

  • In japanese-style byoyomi, the player will always at the nearing end of one period have to ask himself "is it worthwhile to invest another whole period?". You can never invest less time than a full period. If the move is obvious, you have to use the period anyway for other calculations, otherwise you lose a chunk of your total time.
  • In canadian-style byoyomi, the player always has to keep track of how many seconds per move he has for the running period. In real-life tournaments you can see players make the "time-bow" (to look at the analogue clocks more precisely) quite often. Additionally, when you have only one move left to make in a period and already know where it shall go, you have to use the remaining time of the period for other calculations, otherwise you lose a chunk of your total time.

These are all distractions from the real game. With Fischer time, all you have to do is look at the clock sometimes to see how much time you have left. There are no additional calculations and no time use optimizations to overcome the flaws of the time keeping system.

Mef: Indeed, it might be slightly more convenient on occasion, but the question comes down to is it big enough of a difference to merit adding a feature to KGS, and if so where in the order on the long list of future features does it get placed. True, all other things being equal, it might be a nice addition to have for some people, but I think with the tradition of go being with Byo-yomi, it's pretty far down on the list.

Harleqin: Just to stress my point - it is not only "slightly more convenient on occasion" but always better in every game. The tradition of byo-yomi (both japanese and canadian) stems from the inabilty of the early analogue clocks to overcome the sudden death problem in a simple way. With modern digital clocks, and especially on internet servers, the introduction of the simple and elegant Fischer system should only be a matter of time ;).

Mef: Alas, to say it is always better in every game is quite an overstatement, many games never reach byo-yomi, and many that do, the players do not feel pressured by the clock, in these games any of the time systems would be equal. Also this is avoiding the main argument of is it worth it to add to KGS, does the benefit outweigh the cost? At this point, especially with the other systems being the ones used for go. Also on a historical sidenote, while I can't speak for the Canadians, I do believe Japanese time has been traditionally kept by a timekeeper, so I doubt it was designed around the shortcomings of an analogue clock.

Dave: Actually it was designed around the shortcomings of a timekeeper with a single analogue clock :-)

Dave: I am one of those slow players who can (and do) run out of time no matter what the settings are. As a result, I would really like to try Fischer time. In practical terms I see as the different methods translating into the type of time trouble you can get into.

  • With Canadian of course it is the horror (I guess thrill for you blitz lovers) of suddenly realizing you have 15 seconds left for your last 9 moves :-(
  • With Byo Yomi it is the inability to deal reasonably with reading issues that require more than xx seconds (it goes without saying that I manage to burn through my extra periods pretty quickly too). In my own case for example, I can not accurately count the board within 30 seconds (I am just not good at it). [1] So playing under a 30-second time control is painful because I inevitably play the last half of the game without understanding the score. :-(
  • With Fischer I am sure that I will be able to run out of time in my usual feckless manner. However, I am interested in testing the ability to store up a little something with a few quick plays later in the game. Not to mention the fact that I can feel less guilty due to the 'hard' time limit - an added benefit that I did not appreciate until I read this page. :-)

Bill: Reading this discussion, I am struck by the differences in how chess players and go players might typically use Fisher time. In go it does not make much sense to build up time by playing the opening rapidly. Go endgames can usually be played relatively rapidly. Chess endgames that can be played rapidly can normally be resigned or a draw can be agreed. Go endgames can also be prolonged indefinitely, which is why having a constant time per move is important.

Earlier I suggested playing with 10 min. plus 10 sec/move, for a game lasting about 1 hour. Now I think that 20 min. plus 5 sec/move would work out better.

Looking back, I see that wms asked if someone could demonstrate that Fisher time was more fun than byoyomi or Canadian. I think that Dave's point is telling in regard to Canadian time. Canadian time can lead to a horrific time scramble. What fun is that? As dnerra says, with Fisher time you can just play go (mostly).

Which is more fun, 5 sec/move byoyomi, or 5 sec/move Fisher time? The average time per move is the same, right? Under byoyomi, say that I am sure of my move after 1.5 sec. How can I use those precious 3.5 sec. that I have left? Hmmm. What if my opponent plays there? Oh, God, how much time do I have left? Better move now. What a contrast with Fisher time. I can just make the play and then use those extra 3.5 sec. to think about go. :-)

Tamsin: Well said! Fischer time should be more fun than either traditional byo yomi or Canadian-style byo yomi: you need never get into a situation when you have no more than, say, a minute to think about every single move until the end, no matter how difficult, and you need never find yourself with only 30 seconds in which to make five or more moves, which can be very unenjoyable, especially when you're playing online and you are not sure how quickly data is being transmitted back and forth.

We love you wms. We love your server. Please give us Fischer? time controls!

Harleqin: Perhaps you would like to add your + vote on the KGS Wishlist/Game Handling page :)

Harleqin: Update, wms has moved the Fischer item from KGS Wishlist/Game Handling to KGS Wishlist/Rejected :(.

After all this discussion, it would be nice if wms could add a conclusory remark to this page telling what the final reason is for rejecting it.


jfc: Here is a summary of the debate as I read it.

wms perception of how Go players should spend their time

  • Lots of time on opening moves
  • Lots of time on middle game moves
  • less time on endgame moves

wms perception of Fischer Timing

  1. Fischer Time gives the players little time to think in the opening
  2. Fischer Time gives the players little time to think in the middle game
  3. Fischer Time gives the players lots of time to think in the endgame

Comparison of 3 similar time limits

Nothing helps clarify a discussion as a few good examples, so lets compare some concrete examples of Japanese byo-yomi, Candian byo-yomi and Fischer Time.

Example 1: 15 minutes plus 30 seconds a move overtime

  1. byo-yomi: 15 minutes primary, 5 periods of 30 seconds
  2. Canadian : 15 minutes primary,20 move per 10 minutes (30 seconds a move average).
  3. Fischer: 15 minutes primary, +30 seconds added to the clock for each move.

Discussion:

Alternative fischer timing parameters If we assume a game lasts 250 moves then the fischer time limit above will give each player 77.5 minutes of thinking time.

"Hey, wms is right! Fischer timing is too heavily weighted towards endgame play!" you say. Not so fast ... if our approximate game length goal is 75 minutes per player but we want more thinking time for the opening and middle game then the solution is to use these paramters:

  • Fischer: 45 minutes primary, +15 seconds for each move

This give each player about 76 minutes for the entire game and 5 minutes for every 20 moves.

A big difference between the fischer timing 20 moves in 5 minutes and a Canadian Time of 20 moves in 5 minutes is that the Fischer timing doesn't dictate when you stop to count or think deeply.

With Canadian's use it or lose it policy you must be careful not to play the last move of the overtime period when there is an excess of time left on the clock. Instead you must stop and use this time to review the board. What if you, in the excitement of the game, play that last move of the period and then on your next move want to stop review the entire board? Well, your carelessness (of clock play, not game play) has just cost you a few minutes of thinking time.

With Fischer timing, if your opponent plays 4 sente endgame moves that don't require thinking to respond to then you've just built up a nice reservoir of time that you can spend when you want.

abstract comparison

Fischer timing blends the best of Canadian and byo-yomi:

  • you never get caught in the dreaded Canadian trap of having to play 5 moves in 2 seconds
  • you are not stuck in the byo-yomi trap of having no time to pause and think deeply

miscellaneous

jfc: The most compelling argument against implementing fischer timing is wms saying "KGS is my Go server and I do not want to implement fischer timing. End of story". wms has, in so many words, already said this by moving fischer timing to the rejected portion of the KGS feature wish list.


[1]: Calvin: Thanks, Dave. This comment just made my day because I was thinking just the same thing. I assumed all strong players could count the score in like 10-30 seconds, and I'm struggling to get it under 3 minutes without gross oversights. If I'm in 30-second byoyomi, I just don't count, so I suppose I'm playing randomly. I want to play faster games, but I also want to count more often.

MarkGaleck?: wms says that adding Fischer time would be adding another time control. This is not quite true. One can simply add an option (a checkbox) to Canadian time control, that says "carry over unused time". It is simple, and now we have Fischer as a special case of that.

blubb: Personally, I would prefer a clear set of seven rather than four time systems, including

+ Fischer,
+ Bronstein and
+ hourglass

beside the current ones. Such a set would cover the very most timings I've ever seen requested. Asymmetrical time can be realized already by the "add time" option at the beginning of the game.
Anyway, a little checkbox for carrying over unused time certainly would be better than nothing. Cute idea!


Fischer overtime

Jouni: I think that there is one very important point missed in this discussion. Because correct way to implement Fischer to go is not Fischer added to main time like in chess (because as Ian stated that go game should be played mainly on main time), but introduce Fischer byouyomi. That is that we have e.g. 30 minutes main time and 5mins + 10 secs per move extra time. This kind of Fischer byouyomi would translate roughly as 30 stones in 5 min Canadian in speed, but it is hugely more humane way to allocate overtime than insane Canadian stone count. Thus I would like to suggest very strong argument that we really should include Fischer overtime option to KGS. We could even discard altogether Canadian overtime option. (2010-2-17)


KGS Issue - Fischer Discussion last edited by tapir on October 24, 2010 - 22:15
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