PanAtlantik vs. PlayOk - relation of ratings
Daniel Toebbens
players that are active in both PlayOk and the PanAtlantic Elo system.
Nearly all of them are from Europe, both Eastern and Western. Therefore,
the following might only be valid for the FESA part of the PanAtlantic
system. I will split the comparison into a number of mails, so that the
subjects will be easier to search in the archive.
The most important point is of course the relation between the PlayOk
rating and the PanAtlantic Elo rating. I cannot post a picture here, but
you can find a plot of the data at
http://www.sanitaetshaus-winkler.de/shogi/Downloads/Toebbens2009_PlayOk-vs-PanAtlantik.pdf.
It is in German, but the figures should be understandable.
There is very strong scatter in the data. This is hardly surprising, as
a number of players have played only a few games in any of the systems.
In both systems players normally start with a low rating, and work their
way up. So I would assume the higher of the ratings to be more correct.
Of course, as has been pointed out to me, the PlayOk rating can be
falsified quite easily by sandbagging.
There seems to be a slight tendency for players from Eastern Europe to
have higher ratings at PlayOk in comparison to Western players with the
same Elo rating. But the difference is not large enough to be significant.
Among the 50 players are 14, which have established and actual ratings
in both systems. They have played at least 16 tournament games, 8 of
those during the last year. At PlayOk they played at least 30 games, 15
of those during the last 6 months. In the plot these players are marked
with a large circle. These data points give a very nice straight line.
The equation is
*/PanAtlantic/** = 1.75 * /PlayOk/ - 1400*.
Standard deviation from this line is only ±100 Elo points, or ±60
points at PlayOk.
Daniel Toebbens
systems the probability P to win over an opponent with rating OR depends
on your own rating TR as P = 1/(1+10^((OR - TR)/W))). In this equation,
W is the width of the win-loss distribution. If the opponent's rating is
W points above the own rating, the chances of winning the game are less
than 10 %.
This equation is used to update the ratings, so the distribution of the
players ratings depends on W. If one sorts the games of a player
according to the ratings of the opponent into a number of bins, and
calculates the winning ratio for each bin, one can calculate the actual
value of W in a rating system. I did this for my own tournament games,
and the result was W = 404. This is both interesting and hardly
surprising, since in the PanAtlantic Elo system W = 400 is used.
When I did the same for my games at PlayOk, I got W = 247. For those
interested, this is the second plot at
http://www.sanitaetshaus-winkler.de/shogi/Downloads/Toebbens2009_PlayOk-vs-PanAtlantik.pdf.
Since the rating at PlayOk increases 1.75 points for each Elo point, I
had expected a value of /W/ = 400/1.75 = 229. This is again a very good
fit. But the PlayOk system sems to be slightly skewed. I did the same
calculations for a number of of other players, and the result is not
constant. At a PlayOk rating of 1200 the width is about 200 only. And
for very high ratings (around 1900) the width increases to about 300. So
the PlayOk system does not seem to work like an ideal Elo system.
However, the differences are too small to be relevant.
Daniel Toebbens
post their nicknames at PlayOk:
Adrian Woloszyn, adrianwoo, www_shogi_pl, PL
Michele Borassi, borassi, IT
Carl Johan Nilsson, calvinb, SE
Alexsander Shcherbina , cfytr, UA
Karl Wartlick , charlesgo, DE
Daniil Kulin, daniil, RU
David Rockwell, drockwell, USA
Gergely Buglyo, gbuglyo, HU
Martin Hershoff, hershoff, DE
Steven Cain, ippusenkin, GB
Sergei Lysenka, marfey, BY
Benjamin Briffaud, Nivlinch, FR
Oliver Orschiedt, oorschiedt, DE
Peter Hingley, peterhingley, GB
Patrick Arnold, rick7, DE
Erwann Le Pelleter , seikkon, FR
Manabu Terao, takodori, JP
Richard Rödel, temporalen, DE
Bart De Schepper, thecreator, BE
Daniel Többens, toebbens, DE
Thomas Pfaffel, tom2700, AT
Thomas Majewski , trithom, DE
Mikhail Volfson, volfson, RU
Yury Shpilev, YuryShpilev, RU
Adrian Woloszyn
Nicolas.Wiel
country you come from, and it looks like you proved that players from
central Europe are quite underrated.
I had a look at the last European championship results (unfortunately
with very few players from central Europe), and I have an additionnal
remark on the win-loss distribution, which also has national
discrepancies. On the 133 games played, only 5 games where won by a
player whose ELO before the tournament was lower than his opponent’s
by more than 400 points. Among these 5 games, 4 where won by French
players (who totaled 3 participants out of 32 players in the
tournament) and one by a player from Sweden (from a pool of 15 Swedish
player).
Finally, we were all surprised at the end of the tournament that the
European champion Jean FORTIN (2 kyu before the tournament), whose
performance rating during the tournament was above 2000 ELO points,
didn’t get enough points in this tournament to become shodan.
Daniel Toebbens
> FESA ELO seem to reflect different realities depending on which
> country you come from, and it looks like you proved that players from
> central Europe are quite underrated.
>
all of Eastern Europe in the sample. And those with reliable data are
not underrated. On the other hand, those with the highest discrpancies
have played no or few FESA games during the last year. To me this does
not look like an overall underrating. But there are certainly a number
of players that have grown in strength, but have played only few FESA
rated games since then.
Gergely Buglyo
On the Hungarian web, there is an article about shogi and the 2nd Yingde Cup recently held in Shanghai. The text is in Hungarian, but there are a few photos. Here's the link for those who are interested:
http://www.haon.hu/hirek/Sport/cikk/a-debreceni-egyetemi-sogiklub-csapatanak-eredmenyei-sanghajban-a-2-yingde-kupan/cn/news-20090902-06265809
check out the rest of the Windows Live™. More than mail–Windows Live™ goes way beyond your inbox. More than messages