Its interesting that even by 5 games, the spread is very flat, which means that people joining up are very well spread, not biased in the direction of beginners.
The two peaks after 1 game makes sense because one person wins and the other loses, so the higher peak is those people who won their first game.
Rating increases toward the right and the leftmost blue peak is higher than the one to the right, so there are more people who lose their initial game. This spread is pretty much what I imagined it would be. I’m also a little happy to see a trimodal distribution for the first time.
The left peak are player who lost their first game, the right player which won.
In the last year 69.3% lost their first game, the other 30.1% won it.
For the last month 72.5% lost and 27.5% won their first game.
PS: It just came to my mind. This includes every ranked win of the first game. Including resignations because opponent got bored or realized they are playing against an inexperienced player.
That statistic is amazing, but it includes a distorting factor that is hard to quantify. I have seen several cases where it seemed that someone had created two accounts and was using one to rank up the other. In other words, playing against themselves and using one account as the losing account.
Since we are in this wishing mode… Can I ask for the distribution of final ranks, for players who never got to their 15th game? I.e. who are the players who could have stayed with us, but didn’t.
Basically I filter nothing. If the rating history contains at least 1 game, they go in the 1 game group, if it contains 2 games they go in the 2 games group as well, and so on.
I added the total number of player to the table in the first post.
I won’t do this in the near future.
There are some problems to that:
To me it looks like the rank settles after at least 10 games, therefor inferring the rating of players who quit early wouldn’t work. (2/3 “quit” after less than 5 games. The rating hasn’t settled at this point)
I only have ranked games in my sample. If they still are here, but decided to play unranked games until they are better, they appear in my sample as quieted early.
What I have in mind is a breakdown like:
how many players who won their 1st game stayed
how many players who lost their 1st games stayed
lost their 1st and won their 2nd, lost 1st and 2nd, …
I just don’t have the time to do that right now.
Edit: This would probably don’t show anything interesting. If you take a look at the table in the initial post, you can see that while the number of players which have played at least 15 games is a fifth of all players with at least 1 game. The percentage of players weaker than initial rating stays almost the same (≈70% → ≈64%).
If my estimation is correct 19% of the weaker players and 24% of the stronger players stayed for 15 games.
I took a look on how long it takes the average player to finish 15 games if they are new on OGS:
fastest 20% to reach 15 ranked games in 1 day
fastest 50% to reach 15 ranked games in 5 days
fastest 60% to reach 15 ranked games in 9 days
fastest 70% to reach 15 ranked games in 15 days
fastest 80% to reach 15 ranked games in 28 days
fastest 90% to reach 15 ranked games in 66 days
The histogram for September hasn’t settled yet completely.
So we can infer from your data that, out of 63025 new accounts on OGS:
18426 (29,2%) played just one ranked game
18962 (30,1%) had less then 5 games (2 or more)
9215 (14,6%) had less then 10 games
3577 (5,7%) had less then 15
4627 (7,3%) had less then 30
2349 (3,7%) had less then 45
1226 (1,9%) had less then 60
while 4643 (7,4%) had 60 or more games.
In other words: the visible part of each colour in your first histogram is the number of player that didn’t get to the successive group.
If we assume that 60 games are enough to think that a player is willing to stay undefinitely, those players are only 7,4% of new subscribers.
Also about 60% of new players won’t get to play 5 ranked games.
It’s quite impressive, isn’t it?
