So why are the two lists you made so different from each other, and why does the first one look much more… sensible?
I…what? I’m that high up? That feels so wrong.
I’ve made 3 lists… so, I’m not sure which you mean. The very first list (top 10) was the lifetime average Elo. That is, the average over the Elo calculated after every match you played. As such, players that are consistently high Elo will be higher in that list, even if they’ve lost games more recently and thus have a “current” Elo that’s lower. However, the average Elo calculation will also tend to discount players that have improved recently (because they won’t have a long period of time at the higher Elo score to negate their earlier lower scores).
I can’t really comment on the reasonableness of any of the lists, though, since I don’t have a lot of history. The last list I generated (current Elo for players who’ve played w/in the last 6 months) seems reasonable based on current tournament performances that I’ve seen.
You’re better than deluks! I’ll start polishing the trophies.
I got plenty of space for them!
I thought I might be of help, since you did the hard bit of calculating the elo numbers
This plot shows a single player (me) as a blue line over the top of a quantile plot of the entire player database. The dashed red line is the median elo value.
EDIT: better quantiles employed, the middle darkest area is the middle 50% of yomi-ers so this area means you are in the middle of the pack. The slightly lighter area means you are in the upper (or lower) 25% of yomi, while the lightest area means you are in the top (or bottom) 5% of players at each timepoint.
I created a pull request so @vengefulpickle should be able to merge this in to his existing work
This is all well and good but what’s the best way to use this data to make me look amazing? This is the brass ring, people.
congrats to my boy @Fusxfaranto for being the best active player and second-best of all time. you deserve it, #gwenmeta
Since its you
The illustrious career of @Niijima-san began in mid 2015, when they proceeded to rocket into top 5% status in less than 3 months with a run of sick wins. After this promising start followed a quarter known as “the great yo-yo” where everyone’s favourite san would oscillate back and forth from middle of the pack to a top ranked player and back again.
By the end of 2015, Niijima had risen to the top of the upper 25% and held this position for several months, from here their efforts soared up-up and away into the echelons of the top 5% where they remained for almost all of 2016. In fact, since November 2016 they haven’t dropped out this bracket a single time.
With a lifetime average elo un-equalled by any other player known to yomi, I give you all, the one, the only IamNiijima-San.
May their legend never die
P.s. I hope the royal purple pleases your royal yomi-ness
Wow this brought a tear to my eye. I don’t even care that I specifically asked for it. I’m printing this out and using it as my obituary.
That’s what confused me about that list at first, because pickle said “of all time” but it looks to me like it means “currently”, and the of all time is the original list he posted. Or maybe im still misunderstanding something.
The first list is lifetime average (which is one way of talking about all-time). The second list is talking about all players, looking at their Elo at their “time of retirement”. The third list is for players who have played in the last 6 months, and gives their current Elo.
Comparing your trajectory to @Fusxfaranto’s, he’s been on a tear in the last few months, and has run by you, but you’ve been more consistently highly rated over the past several years (so your average rating is higher).
fine, i’m back. who’s good these days. let’s play
Cool, I’m in Top 20 lists these days! Feels good!
You missed the two biggest tournaments of the year, brah. Lum’s Lucky Lottery and 19XX are coming up though.
I was inspired recently to make an attempt at this. I’ve put the results in a separate sheet, so as to not slow down the Historical MU sheet.
It live updates from Historical MU sheet, so it should stay correct. (Also, it uses a k-factor of 40 for the first 30 tournament sets played, and 20 after that).
Well, I’m out again.
Does anyone here have a good sense of whether these numbers make sense? From looking at various players trajectories, I’m beginning to thing that the k-factor after some number of initial games should be smaller (because Yomi is relatively high variance, so individual sets shouldn’t affect your score as drastically). I’m leaning towards perhaps a continuous scale that starts at 40 and then gradually descends towards 10 after you’ve played 10-20 tournament games… But I don’t really have anything but gut feeling guiding me on it.
Don’t know enough about ELO to offer an opinion. Is there a game of approximately equal variance that uses ELO that you could look at for a gut check?
Good suggestion! I found http://www.mtgeloproject.net/, which describes how they set up their Elo calc, and it sounded sensible for Yomi as well. The sheet is updated. Now, being ahead 200 points over your opponent means you should expect about a 60% win rate, and for games played between players with more than 15 games under their belt, that more or less actually holds true.