Probably because Bobby was a chess player?
Yes but still how it’s used in chess is not super informative about how it should be used in Go. And it seems there are different “types” of Fischer time anyway. Increment at every move is one, increment after 40 moves is another but that’s not really relevant here.
From the EGF, which is my experience of using Fischer in a go tournament
https://www.europeangodatabase.eu/EGD/EGF_rating_system.php
But even this is a different situation to casual online games.
What I can say is that there is a practical difference IRL between 5s increment and 15s increment even when the main time is adjusted to give the same theoretical game duration (240 move basis)
1 Like
While that guideline may be suitable for a full Fischer game where the initial time is some 15% of the max expected game duration, I don’t think it is suitable for blitz, where the initial time is only some 3% of the max expected game duration. With your proposal blitz would have a cap of only 1 minute, which is way too tight IMO, interfering too much with players’ ability to build up a reasonable time bank.
I think that a meaningful cap should not be less than the byoyomi basic time for the same game category (which is an upper bound for how long you can be kept waiting when using byoyomi instead of Fischer), and it should also be around ~20% of the max expected game duration or 30% of the typical expected game duration.
Those criteria would lead to caps of 20 minutes for full, 10 minutes for rapid and 5 minutes for blitz. To me, those caps seem quite reasonable for games that usually finish within respectively ~60 minutes, ~30 minutes and ~15 minutes.
But to hit those caps you’d need to play nearly instantaneously for some 90 moves straight, right from the start of the game, just to have the opportunity to let your opponent wait for a long time for one move in the late endgame. That seems quite challenging to me. So in practice, I don’t think it matters all that much if those caps are applied or not.
2 Likes
Here’s the breakdown of the max time setting players here use, again for 19x19 games, 7kyu +, games that go past move 90
30s+3s
| Max time | Number of games |
|---|---|
| 1m | 2481 |
| 2m | 101 |
| 30s | 58 |
5m+5s
| Max time | number of games |
|---|---|
| 5m | 179 (100%) |
5m+?s
| Max time | number of games |
|---|---|
| 5m | 8266 |
| 10m | 1408 |
| 30m | 285 |
10m+10s
| Max time | number of games |
|---|---|
| 10m | 647 |
| 15m | 85 |
10m+?s
| Max time | number of games |
|---|---|
| 10m | 4455 |
| 15m | 272 |
| 20m | 212 |
| 1h | 130 |
Those numbers certainly raise into question whether effectively uncapped is really desired or not.
That said, these numbers, aside from the 30s, which seems pretty legitimate, might be heavily influenced by our interface. Specifically, if you first select live with default time settings, that starts you off at a lower initial time than 5 minutes. Increasing it to 5 minutes pushes the max time up to 5m, and it might be that people either feel like we’re indicating to them that that’s we believe it’s an appropriate max time, don’t notice it or understand what it is, or just don’t care enough to change it. Similarly, if you play 5m and then 10m then go back to 5m, that max time is going to be left at 10m.
They could also feel like max_time = initial_time for 5m and 10m is totally reasonable and generally preferable, it’s hard to say. I think it is fair to say though that it’s not unreasonable, since you’d expect to see folks bothering to change it like is done with 30s+3s.
4 Likes
I think there’s a significant difference between desired and theoretically/academically “best”
My take from this discussion is that there is a (probably small) set of people who care very much about settings that capture maximal situations in games like suddenly needing to really think for a good while at the start/middle/end of the game or pacing moves is various combinations of ways. (I think a 4h blitz game cap falls into this category!)
Then there are a set of people (I suspect the majority) who just want someone good enough for most cases without worrying about Lee usual circumstances. And specifically the main concerns here are fast matching and minimising the impact of bad behaviour like escaping. I think this is why 30s byo yomi is popular because it ensures pretty regular pace all game.
My gut feeling is that @anoek 's proposals land pretty well in balancing pacing and practically.
Specifically, I think it is a bigger problem for most people to have someone playing something like a move every 10s or so and then suddenly taking a minute or two to make a move.
IRL that’s ok because you can see your opponent deep in thought but online you have no idea what’s going on so a drastic change in pace feels wrong.
If people are going to want a “serious” game that allows for all this then there’s custom settings.
2 Likes
In a “full” game there is just no way to prevent your opponent from thinking for 2 minutes straight, be it with Fischer or byoyomi. They could even think for 10 minutes straight. Why should we try to fix that with Fischer specifically?
If it’s unacceptable to you (general you) that your opponent might think for 2 minutes straight even just once in the whole game, I’d say that you should perhaps avoid rapid and full games, and only play blitz or bullet games (regardless of your preference for byoyomi or Fischer)
6 Likes
Yes, I too don’t understand the notion of preventing your opponent of thinking for too long. I specifically pile up my Fischer time on simple moves to make sure I will have enough time when I need to read some sophisticated position. There is simply no way to distinguish between bad faith actors and genuine reading. If you go into a 45 minute main time / cap game, you implicitly agree to take this risk.
3 Likes
One indicator is whether opponent plays a move at the end of their thinking time.
1 Like
I don’t think 10s/move is a full game. Maybe we could scale. 30s/move typically and then 6 minutes thinking time.
But in practice a 5m+5x30s is practically a “full” online game and sure, you can’t prevent your opponent from taking 2 mins to think but it’ll only happen once.
My understanding is that in go played online globally (i.e. not just OGS) these kind of games are far more popular than 30m or 45m games. My hypothesis is that one reason (the main reason?) is that people don’t want to hang around that much and especially they want to know that their opponent is engaged in the game.
Because it can be easily fixed with byo yomi (5m 5x30s or even 10m) and while I favour Fischer and believe it’s the future, if it can’t encourage consistent pace and discourage escaping almost as well as byo yomi then people will get sad.
I feel that anoek’s proposals are a good balance anyway with pretty small increments compared to byo yomi periods.
Maybe this is a question of terminology. I think online full is almost like IRL blitz more or less. Certainly it’s faster than IRL rapid.
1 Like
I do play on lichess, and I can say that at 5s increment (let alone less) I get in trouble simply by not being able physically to make my move. (I’m slightly handicapped as I have already noted here a couple of times.)
–
Ian
1 Like
However, if the opponent is already yose, or even filling in the dame, and makes you wait for more than ten minutes, it is difficult for you to think that this is not intentional.
Difficult but not impossible. May have a break for estimating the scoring.
2 Likes
I agree maybe take a minute or two break.
But if a ten-minute break seems too much for me, that’s no excuse.
And the break can also be done when confirming the score, and it lasts five minutes.
The clock will also be reset every time you change life or death, so you won’t face time attacks like absolute time.
So no matter what, it sounds like an excuse and an unreasonable behavior.
Unreasonable behavior can be considered as running away.
It’s like saying that you need ten minutes to think, but in the end you didn’t notice that the clock timed out.
Now the moderator determines that this behavior is an escape because it is unreasonable.
Below I’m analysing some scenarios of “boredom” attacks in full games byoyomi (20m+5x30s) versus uncapped Fischer (10m+10s). I’m assuming a typical game takes not much longer than 240 moves (120 per player) until scoring.
First I consider the scenario where the boredom attacker plays at a fairly quick pace of 10s per move up to the point where they execute their attack:
-
In the byoyomi case their boredom attack could keep you waiting for some 20 minutes in the early game, and this attack potential shrinks as the game progresses. At around move 120 (early endgame) their attack potential will have dropped to some 10 minutes wait time and this keeps dropping towards the later stages of the endgame.
-
In the Fischer case they could keep you waiting for some 10 minutes at any point in the game.
So the byoyomi case is worse in the earlier stages of the game and Fischer is worse in the later stages of the game.
Secondly I consider the worst case scenarios, where the boredom attacker prepares a big attack by playing bullet (less that 2s per move) up to the point where they execute their attack:
- In the byoyomi case they could keep you waiting for some 20 minutes at pretty much any point in the game (and after their attack they will be in byoyomi for the rest of the game, which would be a much more relaxed pace for them than the preparation phase).
- In the Fischer case they could keep you waiting for some 20 minutes only when the game is almost done (after move 240).
So I’d say that a worst case boredom attack is harder to pull off with uncapped Fischer than with byoyomi. With byoyomi it is progressively easier to execute earlier in the game.
Overall, I’d say that boredom attack potential is (at least) as bad with byoyomi as it is with uncapped Fischer, although the timing is different: with byoyomi the scenarios are worse for the victim earlier in the game, while with uncapped Fischer the scenarios are (possibly) worse later in the game.
But perhaps getting boredom attacked might feel more frustrating in the late game than in the early game, because you already invested more time and effort in the game?
One other note: using a 20 minute cap for 10m+10s Fisher is just about the same as no cap at all for games up to 240 moves. Like, a 4 hour cap will only be hit after move 1500 or so. I have never heard of a game of go that took that many moves.
3 Likes
Yes.
It’s also important to know that this usually happens when the attacker knows he’s going to lose.
Usually only late in the game can you tell that you are going to lose.
In byo-yomi, the main time in the late game is usually running out.
Performing such an attack requires the attacker to be present, and the attacker has to spend his or her own time performing such an attack.
But in fisher, the attacker only needs to drop the game.
Maybe watch Youtube, maybe go have dinner, take a walk, etc.
The victim needs to wait for several minutes or even ten minutes without being able to leave because he does not know when the attacker will come back.
This is the main reason why I am against fisher.
Of course, time caps can solve some of these problems, but I don’t think they solve much.
Because if the time limit is set too low, it will definitely affect the normal game.
And if the setting is not low enough, the attacker can still make people wait for such a long time.
2 Likes
As I said in my previous post, if in a 20m+5x30 byoyomi game your opponent plays at a moderately quick pace of 10s per move, they can still let you wait for 10 minutes on move 120. On move 120 the game is usually already in the (early) endgame and they can probably already tell when they are losing.
So I still don’t see why 20m+5x30 seems so much better/safer to you than 10m+10s in regard to potential boredom attacks. Shouldn’t you play rapid or blitz instead to exclude the possibility of needing to wait for 10 minutes?
A potential solution to this concern is to use a different time system instead of (or on top of) byoyomi or Fischer: applying a time cap for each move. FlyOrDie uses a move time cap of 5 minutes (in combination with 40 minutes absolute time for the whole game). I don’t know if that system of move time cap has a name.
2 Likes
Only if it’s optional. The Arimaa Gameroom does this (as an option). In addition to letting you set initial time, increment (optional), and cap (optional), it also lets you set a max time one can spend on any given move (or no limit: everything except initial time is optional). While I think it’s a quite niche situation where it’s useful (namely for broadcast games where you want the audience to be guaranteed to not need to wait too long before moves), and I would much prefer just simple main time and increment fischer as default, I would enjoy having the option to set this, though there would be a UI problem of how to represent it
It probably sounds like it’d be better for an audience until it turns out players aren’t used to it and then one of them times out because of the move cap timer 
It’s been used in the Arimaa WC for years with no issues. Nosteps (a 3rd-party client which connects to the Arimaa Gameroom) has a very nice UI for the clock which shows in big blue numbers the amount of time you have left for this turn (or will have for your next turn), while smaller black numbers next to it show the current total time on your clock. With a good UI, it’s no more difficult than byo-yomi
That said, I firmly believe both caps and max-move-time, while useful in certain situations, are not good defaults, as for most purposes just choosing a good main time and increment is all you need, and poorly set caps or max-move-times can have an undue negative impact on gameplay
2 Likes