It's difficult to guess how long a game will take

You could display expected game length in the “Play” tab and when creating a custom game.

Even EGF has heuristic rules for how long a game will take. For a Go web site with lots of data about games, it is not difficult to do better in predicting how long a game will take based on the game’s time settings, there are plenty of simple predictive models in e.g. scikit-learn that will probably do just fine. It doesn’t need to be perfect, just a ballpark estimate. Adding some kind of credible interval would be a nice further improvement, but that would probably require help from a data scientist.

I think the main point of use for something like this would be an estimate / warning on correspondence tournaments…

1 Like

I like this idea for live games but virtually impossible to give a meaningful estimate for correspondence games because of the vacation feature. Would anyone mind if we scrapped the vacation feature? :wink:


I am using it durring my trip to China right now. No regular internet.

I found out about my trip when all my games were in progress.

Very nice to have, but I could see it where it could be abused (particularly with Canadian timing). I think it should stay, but could be worth looking at Max avalible as well as how fast it regenerates.

Very bad idea, all my games would end by timeout xD

For ballpark estimate for any correspondencegame, i would say: “up to a year”

Can’t see this ever happening lol

If your opponent agreed to and respected your request to pause the game for a long span of time, would the pause you initiate stay in effect until you disengaged it, or does it time out after X amount of time has passed?

Because if that was the case… what a beautiful world it could be if everyone was kind and considerate, both in allowing others to pause and in players not abusing the kindness of others :sob:.


Yes, all pauses are infinite unless otherwise acted upon. The pause itself won’t end unless one of the players removes it, however, the game can be terminated by a moderator if it is deemed necessary. (I believe tournament games can also be ended by the tournament director, but don’t quote me on that…)

1 Like

I just quoted you on that. :smile:

Tournament directors can’t end individual games. They can only disqualify a participant manually which results in the loss of all their games. Something not to be done lightly.


You don’t really need much. A “stats for nerds” page showing, among other things, mean (± 2SD), median and mode would be a nice place for this kinda stuff.

If I had to pull a few numbers out of my hat, I’d say
1m main, 5x10s blitz: ~30min (±15m)
5m main, 5x30s live: ~30-90min
30m main, 5x1m turtle: ain’t nobody got time for that

You can roughly approximate the time it takes by using best/worst case scenarios. The typical game takes 250 moves, so we get

n Moves s seconds T time
50 5 4m 10s
50 30 25m
150 5 12m 30s
150 30 1h 15m
250 5 20m 50s
250 30 2h 5m
400 5 33m 20s
400 30 3h 20m
1 Like

If you use Fischer time in a fast (non-correspondence) setting, you can calculate a pretty good estimate quite simply: add the basic time for the two players and the bonus time for 300 moves. 300 moves is a rough upper bound for the number of moves in a game (even if it actually gets to 360 moves, you’re only 20% off in the bonus time part), and 300 seconds are 5 minutes which makes it easy to calculate in your head.

Examplle: a 20/10 game (20 min basic, 10 sec per move) can take up to about (2 × 20 + 5 × 10) min = 90 min = 1:30 h.

Of course, any extra time rules complicate the matter.

1 Like