First thing: in the previous message I was posting a wrong picture.
Here is the shape for games that ended by resignation, having ranks for both players on the axes.
As you can see, there are a lot more points than before, since the previous one was just a test on a subset of data.
Ranks span from -67 up to 45.
This is a little confusing for me, since I’m used to this conversion table:
- 38 is for 9 dan
- 30 is for 1 dan
- 29 for 1 kyu, down to
- 5 for 25kyu
I’ve seen before numbers lower than 5, even negative ones, but not down to -67!
Similarly, I’ve seen before high numbers such as 39 or 40. I was classifying them as 9d+
Now I see that we have PRO players on OGS and their rank goes up to 45.
Of course there’s a lot of bots in the higher ranks. I don’t have an easy way to filter them out. I should pick them one by one, based on their names, which would be too hard for me.
There’s also a strange number of “zero” ranks which don’t seem correct.
As an example, we have a couple games by Yeonwoo as black, where her rank is 0.
But overall the “zero rank” players draw some stripes in the above chart, which seem strange.
To answer @_KoBa 's question, let’s take a look on how many games end by resignation, compared to the other possible outcomes, by year. These are ranked games on a 19x19 board.
Resignation is the highest part of games outcome. Then we have respectiverly scoring, timeout and cancellation.
If we look at percentages, the resignation share was increasing through the years. I’ve drawn a reference line at 50% value.
I then placed on the X axis the rank of the losing player instead of years.
This way we can compare game outcome against the rank for the player who possibly resigned the game. I removed ranks below 5 (25k). The chart goes up to rank 45, but there’s a very few games on the highest rank levels.
I drew reference lines for ranks 5 (25k), 30 (1 dan) and 38 (9 dan)
Looking at percentages, it seems that higher kyus tend to resign more than lower kyus, but the overall range is pretty consistent, spanning from about 40% to less than 60%…
In the dan zone the decreasing number of games is affecting a lot the chart. The trend is not clear. It seems that scoring tends to disappear. Resignation is still about the 50% of cases.
For the highest rank we have only few games and the chart becomes highly affected by accidents, such as cancellation and disconnection. This information is not very significant.
The above charts are based on all games available (ranked, 19x19).
I wonder if we could do any analysis based specifically on players:
- are there players that never resign?
- what is the incidence of resigned games for people who do resign sometimes?
People may have very different habits and approaches, but maybe the big numbers could draw some more general shape