I played 2 games with same player. I won one game and lost another one.
The stats says:
1 losses vs. stronger opponents
1 wins vs. weaker opponents
I don’t understand why it says ‘1 wins vs. weaker opponents’ instead of ‘1 wins vs. strong opponents’.
thanks,
Sean
I looked at the details of two games.
I may be wrong, because I have no documentation that explains all numbers, but what I noticed is that:
in game 13351886 (the first one) the “historical ratings” are
"white": { “username”: “seanh1999”,
“ratings”: { “overall”: { “deviation”: 350, “rating”: 1500, “volatility”: 0.06
"black": { “username”: “kmss6108”, “ratings”: { “overall”: { “deviation”: 73.29548645019531, “rating”: 1852.3543701171875, “volatility”: 0.060578715056180954
in game 13352034 (the following one) they are:
“white”: { “username”: “seanh1999”,
“ratings”: { “overall”: { “deviation”: 297.5989990234375, “rating”: 1936.086669921875, “volatility”: 0.06002296134829521
“black”: { “username”: “kmss6108”,
“ratings”: { “overall”: { “deviation”: 72.8206787109375, “rating”: 1836.9188232421875, “volatility”: 0.0606355257332325
I assume that “historical ratings” refer to the situation prior of the game, but I can’t be sure.
So actually it seems that between the two games white’s rating dropped below black’s one.
It looks quite strange for me, because white won that first game, so his rating should have raised, not fallen.
Raw data available here:
https://online-go.com/api/v1/games/13351886
https://online-go.com/api/v1/games/13352034
I’d like to call @GreenAsJade and @AdamR on this one.
Oh, wait, I think I’m reading badly rating vs rank: do higher number in rating mean higher rank?
I hear your call, but honestly this is way above my head 
Especially since these were your first two ranked games I personally would say that it is kind of meaningless anyway, since the system can have no idea how to rate you yet…
If there is some reasonable suspicion that there might be some bigger error, I will talk to Anoek about it, but I do not want to bother him with every little uncertainty.
Yes, higher number = higher rank, I am a little usefull at least
So @seanh1999 was given 1500 starting rating, as a new user.
kmss6108 was 1852, so higher.
The first game was won by sean, against stronger opponent.
The result brought sean to a higher rating (1936) while kmss was losing a little (1836).
So, the second game sees sean as higher then kmss. Sean lost, so it’s a lost against weaker opponent.
Oh, no, wait, it’s the opposite of what sean reported! 
The win/loss statistics are calculated from the ratings after the game has finished. The raw data is obtained from:
You can see where the comparison is made in the makeRatingEntry function. Since the handicap is zero in both these games the effective ratings are the same as the actual ratings.
I agree with you that this is probably not what the user expects.
I heard so.
In this case, live games and no contemporary games, it’s the same: you start and end the game at the same rating.
So, if I read correctly your link, it seems I’m right:
initial rating 1500
first game against stronger 1836: win against stronger
rating increases to 1936
second game against weaker 1856 (don’t know why this raised): loss against weaker
The diagram states the opposite: “1 loss vs. stronger, 1 win vs. weaker”.
I must miss something…
This is nice. I never entered the code cave before. 
BTW, which language is this?
Yes. At the end of game 1: ranking = 1936.09, opponent_ranking = 1836.92
So it is actually counted as a win against a weaker opponent.
At the end of game 2: ranking = 1739.51, opponent_ranking = 1856.89
So this is counted as a loss against a stronger opponent.
I think now I can get it.
So the sequence is:
I used to think the other way: first state how it’s finished and then calculate and apply new ratings.
Thank for clarifying this.