Rating comparisons
Bob Newell
A previous thread prompts me to ask this question: how do ratings compare
from site to site? (They don't--- but can we quantify it?)
For instance the previous poster said, subtract 200-300 from Dailygammon
to get an approximate GamesGrid rating.
Does anyone have an idea, even if a rough guess, how site ratings compare?
For instance, I would suspect you'd have to subtract a *lot* from my
GoldToken rating to get a FIBS or GamesGrid rating. So if we look at, say
GoldToken
BrainKing
Dailygammon
FIBS
GamesGrid
and any others you might care to mention... any idea?
--
Michael Sullivan
> A previous thread prompts me to ask this question: how do ratings compare
> from site to site? (They don't--- but can we quantify it?)
>
> For instance the previous poster said, subtract 200-300 from Dailygammon
> to get an approximate GamesGrid rating.
>
> Does anyone have an idea, even if a rough guess, how site ratings compare?
> For instance, I would suspect you'd have to subtract a *lot* from my
> GoldToken rating to get a FIBS or GamesGrid rating. So if we look at, say
>
> GoldToken
> BrainKing
> Dailygammon
> FIBS
> GamesGrid
A guy on dailygammon (moorg) ran up a page that tried to do this
comparison for DG, FIBS and GG based on a set of players who played on
at least two servers a bit over a year ago, and came up with GG - 150 ==
FIBS - 90 == DG for those three. But he himself said that DG has
shifted weaker in that time, and the experience of a few players on more
than one server corroborates that. The top 5 or 6 players at DG are now
rated over 2100 and none of them has the profile of a ratings
manipulator (it's almost impossible to do without obvious cheating since
you can't pick your opponents for rated games). At 150 differential,
that would put them at 1950 GG, which is a rating only bots world top
players see, AFAICT. Since none of the top DG players are famous names
to my eye, I'm guessing that they'd typically be mid-high 1800s on GG.
One of them (Rob Adams) posts here and I think he'd agree with this
assessment. He may even have a meaningful GG or FIBS rating for
comparison.
There's also the problem of different strengths for real-time vs.
turn-based play. People are drawn to turn-based play who want to be
able to think hard about their play. Some of the difference in DG
ratings may be that the same players actually play at a higher level
given as much time as they want to think about difficult plays.
I have no clue how the ratings on GoldToken or BrainKing compare, but
they'd have the same kind of turn-based effect as dg.
The other problem with turn based servers is more timeouts. I believe
that the DG ratings float upward gradually as people turnover, because
most players who leave the site do not stick around to finish all their
games. On a real time server they normally don't start games that won't
get finished. At worst they may drop their last game, and on many rt
servers dropped games don't count anyway. So the field is strange at dg
-- the most common rating is in the low-mid 1400s, but the experience
weighted average rating is a bit over 1600. I think that a large number
of folks try the server and leave after not too many games and their
timeouts end up leaving them with a below average rating. And that this
happens often enough to gradually shift the ratings of players who stay
on the server upward with time.
Michael
Bob Newell
>Bob Newell <bne...@linux.chungkuo.org> wrote:
>
>> A previous thread prompts me to ask this question: how do ratings compare
>> from site to site? (They don't--- but can we quantify it?)
>>
>> For instance the previous poster said, subtract 200-300 from Dailygammon
>> to get an approximate GamesGrid rating.
>>
>> Does anyone have an idea, even if a rough guess, how site ratings compare?
>> For instance, I would suspect you'd have to subtract a *lot* from my
>> GoldToken rating to get a FIBS or GamesGrid rating. So if we look at, say
>>
>> GoldToken
>> BrainKing
>> Dailygammon
>> FIBS
>> GamesGrid
I should in my initial posting have included my own stats, FWIW. I'm
a player with about two months' experience, though have read several books
and play often. My rating on GoldToken is about 1550 and rising, while
my rating on Dailygammon is about 1430 and dropping, and will likely drop
quite a bit more before it stabilizes.
One of the ops at GG told me that FIBS - 300 will equal GG.
That seems pretty big a difference but maybe he has some data on it.
But that does tie with the idea the FIBS is slightly higher than Daily.
--
Michael Sullivan
> I should in my initial posting have included my own stats, FWIW. I'm
> a player with about two months' experience, though have read several books
> and play often. My rating on GoldToken is about 1550 and rising, while
> my rating on Dailygammon is about 1430 and dropping, and will likely drop
> quite a bit more before it stabilizes.
> One of the ops at GG told me that FIBS - 300 will equal GG.
> That seems pretty big a difference but maybe he has some data on it.
> But that does tie with the idea the FIBS is slightly higher than Daily.
Are you sure he wasn't talking about DG? FIBS may be working upward
relative to GG, but I don't get the sense they are anywhere near that
far apart. Mgnu-expert on FIBS is equivalent to GG-raccoon (aimed at GG
1900) and it doesn't stay over 2000 long, usually sitting in the high
1900s. I haven't played on GG and only a little on FIBS, but my FIBS
rating hasn't risen all that fast from it's initial 1500 (nothing like
on DG, where I hopped up to the 1700s very quickly, am now at 1900 and
only just beginning to stabilize). I played a few 1700 level FIBS
players and subjectively they felt about as strong as mid-1900s dg
players. I'd be very suprised if FIBS is more than ~100 points below
GG.
Michael
Bob Newell
>Bob Newell <bne...@linux.chungkuo.org> wrote:
>
>> That seems pretty big a difference but maybe he has some data on it.
>> But that does tie with the idea the FIBS is slightly higher than Daily.
>
>Are you sure he wasn't talking about DG? FIBS may be working upward
>relative to GG, but I don't get the sense they are anywhere near that
>far apart. Mgnu-expert on FIBS is equivalent to GG-raccoon (aimed at GG
>players. I'd be very suprised if FIBS is more than ~100 points below
>GG.
No, for sure he was talking about FIBS, because I quoted him my FIBS rating
(1550 based on exactly one match) and he said I'd be in the 1200s on GG.
Of course I don't know if he has stats to back it up; he was pushing a
membership on me and trying to tell me how strong play at GG is.
--
Douglas Zare
Bob Newell wrote:
> No, for sure he was talking about FIBS, because I quoted him my FIBS rating
> (1550 based on exactly one match) and he said I'd be in the 1200s on GG.
> Of course I don't know if he has stats to back it up; he was pushing a
> membership on me and trying to tell me how strong play at GG is.
The suggestion of a difference of 300 sounds quite wrong.
For a while, GamesGrid and FIBS were about equal. I'm
pretty sure that FIBS has weakened while the GamesGrid
ratings have dropped, at least at the top, but the gap isn't
close to 300 points. Keep in mind that there are players on
both sites who manipulate their ratings, there are separate
populations of players within the servers whose ratings
are not comparable, and that individual ratings may
have standard deviations of over 50 points through normal
swings of the dice.
Here are a few examples of ratings for players who have
recently played on both sites:
1850 (GG), 1897 (FIBS), +47
1893 (GG), 1979 (FIBS), +87
1911 (GG), 1921 (FIBS), +10
1905 (GG), 1917 (FIBS), +12
It could be that things are different at lower ratings, but
I doubt it. There are fewer than 50 out of 3000 players
on GamesGrid rated under 1300.
Douglas Zare
RedTop
essentially immune to manipualtion and cheating.
GammonZone (www.gammonzone.net) runs tournaments in Tournament Room 16
on the MSN Gaming Zone, and some others off our website. All
tournament matches are rated, and you cannot select your opponent,
match length (except in that if you enter a 3pt tournament, you'll
play 3pt matches) or manipulate in any other way.
In general, I think our ratings are about 50 points higher than
GamesGrid. A lot of our players are not very strong, but some are
pretty good. What matters most, though, is that the ratings are valid
and accurate.
On 05 Apr 2004 19:43:05 EDT, bne...@linux.chungkuo.org (Bob Newell)
wrote:
happyjuggler0
First, The "Zone"'s rating system is hopelessly flawed for several
reasons but the main ones are that most of the other flaws draw from
is that it rounds off to the nearest integer. The other one is that
players can "play" matches vs themselves via using 2 pc's and the Zone
does not seem to give a fig. Thus anyone within +/- 400 points (I
believe the cutoff is actually around 440 but it has been a long time
since I played in their rated rooms) who plays a 1pt match both gains
and loses the same amount of points (2). Subtract anywhere from 0pts
to about 4000pts (not a typo) to get an accurate comparison of Zone
ratings with the following 3 systems.
The next 3 rating systems are all based (more or less...with an
emphasis on more) on the same rating system.
GammonZone (GZ). Free. This is probably the rating System I play in
the most at the moment. This is entirely tournament based. There are
daily tournaments of 3pts or 7pts (3pts are more common unfortunately
imo...but give the people what they want). Each month they have 2
playoff tournaments (tpts each...one is double elimination) based on
results from the daily play. They also have a round robin league with
several bracket and divisions with 21pt playoffs. They also run a
monthly tournament called Medallion which is mostly 11pts with 13 or
15 for the final rounds. Progressive Consolation for Medallion is also
11pts. Second chance Medallion is 9pts but anyone can enter with
preference given to players who lost in Medallion that month(I believe
the cutoff for 2nd Chance is 128 players). They also run a couple of
tough tournaments a year (lasting 3-4 months) that are basically best
3 of 5 11pt series in a single elimination format (13 and 15 pts for
the later rounds) and the consolation has a "crazy" match length of
something like 17pts early rounds to 29pts later rounds.
GZ's rating system is sound and rating updates are posted weekly. The
strength of the tournaments varies wildly with the longer ones tending
to draw much stronger players.
GamesGrid (GG). Pay site. GG is a site that runs some small
tournaments (mostly 1pters) but is mostly pick your opponent. A lot of
the world's best players play there (e.g. Woolsey, Robertie). They
have guest memberships that are free but are not available for ratings
or money accounts. Those with a full membership play their choice of
rated or unrated matches of money sessions. You can play for money if
you have a money account but you don't need a money account to play.
There are bots available to play (for members only). GG has a "saved
match" system that is adjudicated by Snowie after a couple of weeks if
the players do not resume play and GG keeps track of a players
completion percentage on the player info page.
GG's rating system is mostly sound but many players are deliberately
underrated so as to better entice others to play with them for money.
FIBS. Free. FIBS is almost entirely a pickup match system with no
daily tournament I am aware of and a very small number of website
based tournaments that take a few months or so to play. Getting
started the first time is a pain in the butt but once you get there it
is worth it for the most part. There are many different interfaces to
use but I have only tried 3DFIBS and love it. There are bots available
to play. They have a "saved match" system designed to allow you to
replay unfinished matches due to booting but some players abuse it and
quit instead of allowing the rating system to knock off points for
losses. Fortunately you can view the players' "saved" games numbers
and simply not play those with lousy ratios to matches played.
FIBS seems mostly sound but I am not sure I have figured out all the
"angles" that players may use to abuse artifically inflate their
ratings.
When I played on GG my ratin was about 1830 when I stopped and had a
high rating of about 1910. This is with about 5000 or so experience
points if memory serves. My GZ rating is officailly about 1830 but I
had a good week so far and it will probably be about 1860 on Sunday
when the list is updated assuming I don't play more and have it go up
or down more. My high GZ rating is 1945. This is with about 5000
experience points. I am currently in a "down" period for my rating and
I think (perhaps very wrongly) that is "belongs" somewhere around
1900. My FIBS rating is about 1800 with a high about 1830 and
experience points of about 800. I do not think my rating is high
enough for my relative strength there yet but this is obviously
subjective assessment.
The problem with cross-platform rating comparisons is that you need a
huge body of players who play in each of the different systems to be
able to call the systems numbers compatible but I believe my numbers
are roughly normal for a player on the relatively high end and Douglas
Zare's examples ring true based on what I have personally experienced.