What happened to tournaments.players.ranking?


looks like in this forum I am the only one who talks about “playerranking”. :grin: (EDIT: yes, I am. I checked that at a later time)

Well, it also looks like no-one (not a single one!) out of 2233 players in the Through the years tournament had changes in his “playerranking” since last february 1st. (my previous download for tournament data).
It seems quite unbelievable, doesn’t it?

If you look at a single player you could overlook that… here is my chart:

Looks the same as before changes in rating system. And actually it shows my rank as 7k, while on the main site I am 4k today.


But if you look at the whole picture…

These are all changes in rank for each player in the “Through the years” tournament since my first download (16th of july 2020). 2233 players.
In the last week ranks didn’t change at all…

I was expecting bumps in this chart too, but nothing happened of course:

Please please please, fix that!
It’s SOOOOO useful!

BTW, I noticed something strange also in in-game chat. Ranks don’t look consistent:

Here is an example from this game: https://online-go.com/game/28887902

Here is the chat:


Look at the evolution in rank between these two dates.
Lys: 2k to 4k
EverG: 1k to 3k

Here are rank graphs for both players.



Love your modern art.


No change is the new volatility…


While I admit it is unlikely for no one to change rank over the course of a week, don’t you think players in the “Through the years” tourney would have the most stable ranks?

1 Like

Which API do use to query the ranks? The rating history does show newer games as well.

In this case I use “rankings” from tournament players list.

Well, I have to double check that, but maybe the word “playerranking” is only in my head! :crazy_face:

Maybe I created that monster while dissecting some poor json.

Anyway “ranking” is a thing and the one in “tournament \ players” json didn’t change in a week.

I don’t think so.
Surfing through the data I collected, I found players with stable rank and players with huge fluctuations. The panel is so wide that you could find anything inside.

Moreover: we can see from my “modern art” that after few days there were always many changes… except for the last week.

Was it a medical fiction we should be screaming: “we’re losing him!”

1 Like

Rankings are also in “game” json, in multiple occurrencies.

I discussed it here: https://forums.online-go.com/t/salt-cat-mad-handicap-9x9/31157/25

I didn’t check that, but I should be doing it soon, since “Salt cat” tournament’s second round is about to finish (only 3 games left) and we would discuss handicap and ranks… what’s going to happen?
Will historical rankings be the “old way”? Will them stay like that? Stay tuned…

1 Like

Well, not exactly: it’s also in my database.
While processing data the “player.ranking” attribute was stored to a field losing the dot.


We are losing them all!!!



It is a funny enough thing to spot. In the initial graph I didn’t know what I was looking for, but then you highlighted the flatline.

I think unless you happen to have a player with particularly volatile ratings or who finishes a lot per day, it could be the case, since ranking is rounded in some way to an integer, unless it changes significantly it’s not going to move.

I think wins against even-ish players are only giving about 10 rating points for a win/loss, it’d probably take a few games to push to the halfway point between ranks?

@anoek the user ratings on some APIs do not update


Except that it does.
We have about six months of data and you can see from the first picture that changes are very likely to happen.
Some player’s ranks are more stable, some have big changes, but the overall picture is brownian chaos. The opposite of the rigor mortis seen in the two last weeks.

1 Like

I’m wondering if that’s an effect of the way the calculations were done retroactively?

As far I know, it started off recomputing ratings of active players first and then worked back to inactive ones?

I actually have no idea how a calculation like that worked. Let’s imagine I was the first person whose rating was being updated for no reason other than needing an example. Would it need to update, for each player I’ve played, their ratings until the point I’ve played them so that it knows their rating as accurate as possible so that mine can be updated accurately?

In that case it one would think, updating from the oldest games first to the newest would be better. I don’t think that’s what happened

What I wonder then is that if you ran the computations again would you get different ratings, different volatility or different deviations for players.

I’ll need to look at my own rating history out of curiosity to see how it put me where I was, how many points I was gaining in old games prior to the update vs now etc.

Or maybe a simpler explanation, that parameter just stopped being updated? :slight_smile:

Reading back a tad

1 Like

Just curious- has anyone noticed this manifest on the site itself (AKA tournament pages)?

I think so.
The ranking coming from the tournament API is way different from my present rank as displayed in the site and it’s stuck to the value it had before the changes in rating system.
I imagine it’s the same for all other players.

I really don’t know what data you all are looking at… when I go to https://online-go.com/termination-api/tournament/59567/players, I see @Lys’s ranking as 4k, just as on your profile page:

"308797": {
  "id": 308797,
  "username": "Lys",
  "country": "it",
  "icon": "https://b0c2ddc39d13e1c0ddad-93a52a5bc9e7cc06050c1a999beb3694.ssl.cf1.rackcdn.com/5ccab3d1d5a7f9e6ecef1437a840a10d-32.png",
  "ratings": { 
    "overall": {
      "rating": 1678.9264106565283,
      "deviation": 63.950846297381744,
      "volatility": 0.059807983530599795
  "ranking": 26.91214271252403,
  "professional": 0,
  "ui_class": "supporter",
  "points": -1,
  "net_points": 3,
  "sos": 0,
  "sodos": 0,
  "rank": 0,
  "disqualified": 0,
  "resigned": 0,
  "eliminated": 0

26.9 == 3.1k, right?

I feel like I’m going crazy here!

EDIT: JSON formatting

1 Like

OH. I see this is a termination_api vs api/v1 issue. I don’t know what the difference is, but is there any reason not to switch to the former? That’s the one that is used in prod…

1 Like


Me neither.

The only one I see is that my workflow already has the latter. :grin:
I must check if the jsons structures are the same. In that case it would be very easy to switch.

It would also be helpful if someone told us whether api is the right one to use!