The top 28, six recent winners and/or the runner-ups as well as the current last place (for reference) are shown. Players with 0% rating reliability are excluded. A golden background of a player’s rating indicates that the player is currently at their personal career high.
My code for the calculations slightly modifies the Glicko system (link to paper).
Glicko Rating [97.5%] = r - 1.96 * RD
i.e., the true strength (rating) of a player is higher than the respective value in the ranking table with 97.5% certainty based on the system methodology
Each player has two values that determine their Glicko Rating at any point:
r
true strength (rating) predictor
every new player starts at 1500
changes only when player participates in a game, based on various factors (e.g., RD)
RD
rating deviation
every new player starts at 350
decreases when player is involved in a game of a tournament week, based on various factors
increases for player before each tournament week, up to a maximum of 350 (based on c = 35)
reliability = (350 - RD) / 350
Please let me know if you have any feedback, questions, or suggestions for improvements. Also, feel free to discuss the results.
For further details on the games considered, refer to the replies in this thread.
I’d like something like that too, but it’s really hard to do. I don’t think @JayKyburz is keeping a huge data dump of all past games and their results.
Even then it’s hard to make it comparable - the multiple player aspect is just one of the issues. Players may also have acted differently if all games are ranked. Don’t get me wrong, I’d still give it a try because so much data not being used in some way would be saddening haha.
Even though my match with @Solfyre of Round 39 is still ongoing, I just updated the rankings counting it as a draw for now. This is so we have a mostly up-to-date ranking that will be used for seeding the December tournament (sign up now!).
With Round 37 and 38 the April tournament is now wrapped up. @HULK took the win and keeps raising his personal career high rating at the top of the ranking. At 1770 GR now, he’s getting closer to @SuperDave’s all-time record of 1808.
Round 39 is comprised of the second round of the August tournament as well as the first round of the August 2CC. Many players boast new career highs - @DarkDragonwing, @Kaine and @BOsshOgg had an impressive rating increase of around 200 points. Also, @Karmadrome catapulted themself back into the Top3 with their latest victory.
Certainly. There’s still three games running in Dec Round 1, but I would say there’s a clear leader in all of them. Might publish new preliminary ratings in a few days based on my assessment which I’d correct in case of changes.
There’s still fairly open games in the August tournaments, so this Round 40 update considers the first round of the December tournament.
64 unique players have now participated in at least one official 1v1 game since 2018, meaning that if all of them participated in one tournament we could set up perfect brackets for it
@Kaine managed to break into the Top10 with his second win, surpassing seven other players to land at 8th. Six newcomers achieved victories in their first game, providing them with a good foothold in the rankings. The more veteran 1v1 players @DarkDragonwing, @panblanco and @BelSon secured wins and increased their rating points by up to 149 which lifts them into the highly contested “upper middle area” of the ranking list.
Actually saw that, but to keep things simple and consistent the whole ranking calculation currently follows a round-based approach, meaning games are only considered once a tournament round is fully finished (or its end is in sight).
While keeping games bundled that “belong” to each other, it comes with certain drawbacks, most importantly some finished games waiting in the backlog. This problem has become somewhat more prevalent now due to bigger and parallel tournaments and vast differences in bracket pace.
Your reply made me think some more about that and alternatives. I may switch the whole thing to a week-based approach as I’d like the steady time component, but need to look into my code first as I’ve written it quite some time ago.
Totally reasonable. Yeah, Good example- I’m waiting on a round three game against the winner of a pending round two game that’s still waiting for a winner in round one.
Adjusted my dataset, changed the methodology and re-ran the calculations - 40 tournament rounds were transformed into 62 tournament weeks.
A tournament week is defined as any week where a tournament game comes to a conclusion. If no game concludes in a week, the RD (and rating) of all rated players stays the same. Limitation: Only one game per unique player per tournament week is allowed. In case a player finishes more than one tournament game in the same week, they get split into different tournament weeks (9 adjustments of this kind had to be made for the 115 past games). I adjusted the c value down by 7 to account for the generally increased interval between games, making the RD increase and reliability loss a bit less punishing for player ratings.
Will upload the new ranking tomorrow, including all finished games up until now
Update - the new ranking with tournament week (T-Week) methodology is live now, including all 116 tournament games that have concluded since 2018. Rank and rating changes are measured on the last 28 days to get a better picture of trends (sometimes a T-Week only has one game).
With two victories each, December has been the month of @panblanco and @GluteusMaximusII - the former rose by 305 points and 16 positions in the ranking to a personal career high of 1,101 points, while the latter jumped straight into the Top5 with their first two games, leading to a rating of 1,316 points. @DarkDragonwing scored their second victory in a row and rose to over 1,000 rating points for the first time in their career.
Several new faces joined the December tournament and put themselves on the map - or more specifically into the Top25 of the ranking: @Cryptious, @Keeves, @Dr_Gaming, @plug and @kinghuang already moved on to the second round.
@Dysp is also new in the Top10 as one of their won games waited in the backlog with the previous methodology, the inclusion of which helps them to an impressive 1,284 rating points at the moment.