Right.
“Should”, "maybe , “probably” and so on.
Every time someone asks “how does this work?” these are common answers.
That’s what the meme is about.
My problem is that I was reading some variables and they don’t seem to be valid anymore.
Other variables, that seemed to be out of use, now are the most up-to-date.
There are plenty of nodes that can be read in api, but I just can’t say which is good and which is bad. The ratings system change affected also those variables and I’m still trying to figure out how to settle my toy again.
I know all the “volunteers” and the “one dev” stuff and I sympathise with you all, but it’s frustrating when a piece of code that was working, suddenly doesn’t and you can’t know why.
This week I am also experiencing issues on my side (MSAccess has become slow and unstable and again I don’t know why) so I am frustrated twice.
I’m looking more closely at the 6d vs 3d game that got 4 stones. I realized that you can find the rank at the beginning of the game by looking up the historical rank from the first game each player finished after the start.
EDITED using the API endpoint recommended by @Lys, below, which appears to show the new rating after each game. Note that the questionable game ends at time 1614538833, so you can look for games that ended close to that time in the history.
The highest rating I see for Sadaharu anywhere close to that time is 2435.07. This gives a rank of ln(2435 / 525) * 23.15 = 35.519..., or 6.5d.
Here is the same thing for minxing43. The lowest rating I see around that time is 2084.29. This gives a rank of ln(2084.29 / 525) * 23.15 = 31.918, or 2.9d.
That’s a difference of 3.6 at most. Using my best guess about which line in the API represents the starting time, I get ~3.1. So I’m very skeptical that the handicap of 4 for this game was actually computed from the formula @flovo presented.
On the other hand, I did a bunch of tests (1, 2, 3, 4) with unranked challenges and they all worked out correctly according to that subtraction formula. Does this mean something is different with the handicaps used in tournaments?
I don’t know if you are right with that. Probably you are. But isn’t that weird?
The game record has place for ratings at the beginning of the game but they are unusable so we must check other games.
Did you check also rating history for that player?
That’s another way to pick so’s rating back in time.
Excepting mod actions, your rank only changes after a game ends, right? So the first game that ends after your game has started should indicate it. Yes it’s weird.
Is there a rating history API that shows the details when it changes several times during a day?
My understanding is very limited. But I think the game API data includes the rank at the end of the game (and this is what is displayed in the player cards forever after), before it is adjusted based on the result of the game.
I didn’t play with it yet.
I don’t know how it looks.
The older one was presenting one record per game with new rating calculated after the game… probably…
Is that the old cache refresh problem if you’ve looked at Sofiam recently? I’m completely untechnical with all this stuff but that often seems to be the answer with incorrect display!
Now I should edit all my queries to collect also ratings (I just took rankings so far) and check if it’s according to what happens in the tournaments I’m following.
Is there any cap on handicap stones number?
Could they be more than 9?
Apparently for 13x13 games, players with nearly identical ranks get a game with handicap 1 and komi 0.5! Example: this game, in which the players had a difference of only 0.03 kyu at the start of the match. This is unlike 19x19 and 9x9, in which the handicap can go all the way down to “None” (zero).
