Hey
I would like to know how long on average a correspondance game 19x19 lasts.
With some common setting like for ladders, automatches, AST or your personal preferences.
Anyone interested in statistics (if this is possible), or anything already published somewhere?
Estimation or personal experience as well as statistics on datas are all welcome!
Thanks
For me anything from a few days to about 6 months. There would be outliers beyond that. I tend to only play “fast correspondence” tournaments. But even my “through the ages” games only lasted about 6 months.
For me, between 1 week and 3 months.
For me, using +1 day up to 7 days with weekend pause, it takes 3 or 4 months.
I’d say mine last roughly between two weeks and six months.
When I did some +4hr increment ultra-fast correspondence games, I got 7 games that all finished without any time outs. Here are their lengths:
I’ll look into the 27M database, but I didn’t save details about time settings.
I should have some flag about ladders and tournaments though.
Wonderful
Don’t ask me 
Asking me? IIRC, my longest corr. game so far lasted for over two years 
Good pick, thanks.
Thanks for everyone own experience too.
From the 27M OGS Dump, these are games that lasted more than 24 hours.
1’096’544 ranked games on 19x19 board.
Longest game was 1811 days.
As you can see the curve is very flattened after about 200 days.
Here is the detail for first 30 days.
This is when we pick only ladder games (game name contains “Ladder challenge”): 146’352 games.
This is for tournament games: 393’297 games.
The big picture is the same: a peak at 3 days.
I’m quite suprised by this result. It seems like there’s a lot of fast correspondence players which are able to complete a correspondence game in three days!
I’m asking myself if I’m doing it wrong… 
Did you exclude annulled games? Pretty sure 3 is when non-moving games time out?
These are the time controls for a ladder game, meaning if you play enough moves to get it ranked on the first day and then abandon it not realising it was corr, they would all end after 3 days too…
Correct!
When I keep only scored games and resigned games (excluding cancellations, timeouts and so on), only 683’028 games remain and the curve becomes even smoother!
The overall average is 38.7 days but the range spans from 1 to 1658 days!
This is just tournament games:
This is just ladder games:
My correspondence tournament/ladder games at +1<7d would usually take several months.
I can easily change scale to log but it won’t change the curve shape. It will just shrink it horizontally.
I thought about the bump in your chart and I believe it’s due to the clusters you applied to data in order to make them visible like a histogram.
Each bar has a different time span. So games that ended within, say, “1 week to 1 month” are more than games that ended within “1 day to 1 week”.
My chart instead shows that games that ended after exactly 30 days are less than games that ended after exactly 7 days.
Technically they are both correct, but I wonder which is more meaningful. I feel like yours is a little misleading, since at first it brought me to think that more games end after a month rather then a week, which is false.
Yeah, I of course meant redo it with logarithmic bins. As I said, it makes sense to group together 400 and 410 while keeping 1 and 10 separate.
If you mean games that ended exactly on the 7th day vs games ended on exactly 30th day, I don’t think the question is good. When we say, oh, the game lasted an hour, we might mean actually give or take 10 minutes. Game took a week, it could be maybe 6-8 days actually. If we say a year, we could generously give or take even couple months. We think in relative values.
And the area is preserved, I think. So as you said, it’s easy to compare ranges, especially with helpful red lines.
To be honest, haven’t thought about it in a while.
Maybe math person could help. @Vsotvep, linear vs logarithmic histogram?
Registered finally 20 ladder games as main dishes and a handful of rengo for snackies. Hope won’t be too much strain.
