1s per move ranked games

There is lots of sense in homogenising what a ranked anything looks like in a certain system, almost all sports have unified agreement on how long a game lasts within certain formats. However, I have no intention of laying out those arguments here. So you can either take me at my word or dismiss it, I don’t mind which.

IF 20s/move and 50s/move both allowed and both counted to single main rank, there is no sense in forbidding 35s/move ranked


I think I agree with you, especially after the 10 games I tried yesterday. But they can’t be compared to “normal” games IMHO.
I’m probably not the best player at this speed, but what I realized is that I play very bad moves, sometimes. 1 second is very very short :sweat_smile:
In the 4 examples I posted, for example, I missed some killing sequences that I would have played with 3+ seconds, IMHO. I won them anyway, but if I can I always go for the kill.

I think the games I played yesterday against zhangpingchuan are quite normal and similar to what we would have played with longer time settings, except maybe this one (see my d4 or their f1)
Here white played h8 instead of h2. It’s very hard to make the right choice in one second. I’m sure they would have played the best move with more time.

No blaming intended, of course. I used these games because they’re the only examples I have.

I agree, of course :slight_smile:


Maybe there could also be a maximum amount of time per move that would trigger the warning sign for live games? Like if the total estimated game time is greater than 4 hours or so? (see topic Slow live time setting)

To compute an estimation of the (soft) max total game duration, you could use something like (per player) softMaxTotalTime = basicTime + 120 * increment. And perhaps apply a reduction factor for byoyomi (maybe 0.5?), because in general byoyomi gets partially lost, which doesn’t occur with Fisher increment.
I haven’t figured out yet how to account for Fisher maxTime.

Minimum absolute time setting can be determined. It should be ( different for 9x9 and 19x19 ) or ( 9x9 should have minimum time of 19x19 )

possible idea: if main time of byo-yomi is equal or more than minimum absolute time, then 1s/move byo-yomi may be allowed.
But: people which choose byo-yomi instead of absolute time do it for reason and may often incorrectly expect for it to be significantly different from absolute time. So I think 1s/move shouldn’t be allowed in ranked anyway.

1 Like

I absolutely agree. I’d use that not only for super long time controls, but for approximating Primel time controls with, for example, increments of 8s. I would love to retain the flexibility of ranked too, though.


A suggestion to compute time estimates for different time settings and board sizes:

numberOfMovesPerPlayer = round(boardWidth * boardHeight / 3)

Fisher: totalTimePerPlayer = intitialTime + min(fisherMaxTime, numberOfMovesPerPlayer * increment)
Simple: totalTimePerPlayer = numberOfMovesPerPlayer * timePerMove * 0.75
Byoyomi: totalTimePerPlayer = mainTime + (numberOfMovesPerPlayer + periods - 1) * byoyomi * 0.75
Canadian: totalTimePerPlayer = mainTime + (numberOfMovesPerPlayer / stonesPerPeriod) * overtimePeriod * 0.75
Absolute: totalTimePerPlayer = totalTime

(time units of all parameters are the same)

And the timePerMove = totalTimePerPlayer / numberOfMovesPerPlayer
Then maybe the game should be unranked/signaled when timePerMove is less than 4s or more than 60s for live/blitz. And maybe the boundary between live and blitz is at about 10s timePerMove?

You can play (and maybe find bugs) with this google spreadsheet. Edit the bold numbers (note the units, m = minutes, s = seconds) to see numbers change in the F and G columns.


I was trying to remember what we discussed last year on Christmas eve.

Comparing to chess (taking numberOfMovesPerplayer = 80):

Bullet would be about 2 minutes + 1 second increment, computing to a timePerMove of about 2.5s (~5 minutes absolute in go).

Blitz ranges from about 4 minutes absolute or 3 minutes + 2 second increment to about 10s per moves. That would compute to a timePerMove ranging from about 3-4s to about 10s (~6-20 minutes absolute in go).

Rapid range is 10-60 minutes + 0-10s increment, computing to a timePerMove ranging from about 10s to about 60s, which is similar to what I would call “live” time settings in go (~20-80 minutes absolute in go).

1 Like


as we can see, number of games that needs 300 moves is not little.
if 2s/move is minimum, then 150*2s=300 seconds absolute time per player or 5 mins

so absolute time surely shouldn’t be less than 5 mins in ranked

all except max blitz options for absolute time are less that 5mins!
This is absolute absurd for 19x19


Yes, the average game may be (a little) longer than 240 moves due to ko fights and captures in general. But ignoring (ko) captures, I think 240 moves is a decent estimate: both players placing about 120 living stones and making about 60 points of territory is quite normal IME and it adds up to 360 intersections.


If you think but don’t read, a blitz is ok between 8 and 10 mns/player.

Some related debates here


Ah yes I cropped it too much and missed that one.

Was trying to avoid the many variants ratings.

Something like 15s+0 absolute.

¼+0 • Rated • UltraBullet

It could be something like adding back some fraction of a second for general latency?


Probably; that’s one of the cool things lichess does.


I’ve seen people make upwards of 20 moves with less than a second on the clock in lichess bullet arenas… I get the impression their clocks run differently to ours… Almost like every second is its own mini byo-yomi period where it gets restored if not exhausted… I’m not sure if we do that?

Premoving? Like conditional moves here, but a much faster input method as they aren’t actually conditional on what you’re opponent did, so can lead to mistakes if opponent plays something surprising.


Yes certainly pre-moving is part of it, but even apart from that, I believe I’ve seen milliseconds be refunded to restore a full second and I’m not sure if we do that or not…

Besides premoving, isn’t that just a 1s increment they’re often using in bullet?
If you use less than a second to move and the software records the actual time used client-side between showing you the opponent move and you responding (compensating for various sources of lag), you would be building up your time reserve.

1 Like

nah almost all bullet tournaments I’ve seen on Lichess are 1+0 not 1+1