> How do ratings compare between servers?
>
<snip>
Thanks for putting this together Daniel, quite interesting.
Does anyone know if anyone has done a study of the performance of the
servers. Things like, Time to generate next dice roll after player makes a
move, average time to complete a game against a bot, how these numbers
vary with reference to the number of logged in users. etc etc?
I didn't find anything like this on deja news (though I could have asked
the wrong question).
--
JP White
Mailto:jp.w...@nashville.com
Below, 3 tables compare ratings on three popular servers: FIBS,
GamesGrid and Netgammon. The tables were compiled from server-provided
ratings lists and include all players and only those players who meet
a server's requirements for inclusion on its list:
FIBS, all users with 50 or more experience points;
GamesGrid, all rated accounts;
Netgammon, members with 10 or more experience points.
The FIBS list, therefore, includes 6148 players but excludes 7736
registered users who have less than 50 experience points. The
Netgammon list includes 4575 players but excludes perhaps 1500 members
with less than 10 experience points and an unknown number of
nonmembers (Netgammon claims 16,000 players in all). The rather
impossible task of identifying and excluding all nonplaying,
duplicate, staff, monitor, and/or spurious accounts has not been
attempted with the exception of 3 obviously nonplaying accounts at the
very bottom of the GamesGrid rating list.
Data is current as of 9/19/99 (FIBS) and 9/22/99 (GamesGrid and
Netgammon).
TABLE 1 -- % of Players By Rating
This table presents the number and percentage of players at or
particular ratings:
----FIBS---- --GamesGrid- --Netgammon- Rating
# % # % # %
1 00.02% 0 00.00% 2 00.04% 2200.00
1 00.02% 0 00.00% 23 00.50% 2100.00
9 00.15% 10 00.45% 81 01.77% 2000.00
24 00.39% 32 01.42% 158 03.45% 1950.00
75 01.22% 88 03.92% 259 05.66% 1900.00
173 02.81% 169 07.53% 404 08.83% 1850.00
368 05.99% 296 13.20% 615 13.44% 1800.00
604 09.82% 415 18.50% 892 19.50% 1750.00
973 15.82% 596 26.57% 1264 27.63% 1700.00
1424 23.16% 772 34.41% 1689 36.92% 1650.00
2035 33.10% 1011 45.07% 2212 48.35% 1600.00
2708 44.05% 1245 55.50% 2751 60.13% 1550.00
3570 58.07% 1550 69.10% 3311 72.37% 1500.00
4323 70.32% 1817 81.00% 3756 82.10% 1450.00
4991 81.18% 2001 89.21% 4081 89.20% 1400.00
5744 93.43% 2178 97.10% 4397 96.11% 1300.00
6021 97.93% 2227 99.28% 4510 98.58% 1200.00
6148 100.00% 2243 100.00% 4575 100.00% 0.00
TABLE 2 -- Ratings and Rankings by percentile
Table 2 is perhaps more useful than Table 1 for comparing the ratings
curve on different servers.
Percentile -----FIBS----- --GamesGrid-- ---Netgammon---
Rating Rank Rating Rank Rating Rank
Top 100% 2239.76 1 2080.47 1 2276.53 1
99% 1907.52 61 1964.74 22 2040.57 46 <<Top 1%
95% 1813.72 307 1881.13 112 1915.58 229 <<Top 5%
90% 1748.10 615 1828.51 224 1833.06 458 <<Top 10%
85% 1706.85 922 1783.65 337 1786.61 686 <<Top 15%
80% 1669.00 1230 1737.85 449 1740.29 951 <<Top 20%
70% 1614.43 1844 1674.34 673 1685.29 1373 <<Top 30%
60% 1566.99 2459 1626.10 897 1634.59 1830 <<Top 40%
Median 50% 1526.96 3074 1577.14 1122 1593.47 2288 <<Top half
40% 1491.91 3689 1533.07 1346 1550.34 2745
30% 1451.34 4304 1496.25 1570 1509.91 3206
20% 1405.72 4918 1455.37 1794 1460.97 3669
10% 1337.23 5533 1391.68 2018 1392.58 4118
Bottom 0% 533.26 6148 961.56 2243 766.68 4575
TABLE 3 -- Ratings creep
Table 3 presents the percentage of currently rated players on FIBS,
GamesGrid and Netgammon rated above 1500.00, the starting rating and
nominal starting point for new players:
FIBS GamesGrid Netgammon
% # % # % #
58.06% 3570 69.10% 1550 72.37% 3311
Over time, ratings systems suffer from "ratings creep" as lower ranked
players drop out of the system with greater frequency than higher
rated players. This is normal and unavoidable. Median rating and the
ratings curve in general will also rise if rated players earn points
from players who are excluded from the ratings list because of minimum
experience or membership requirements and, assumedly, have generally
lower ratings than the players who are on the ratings list.
Other reasons for ratings creep include ratings manipulation by
players (unauthorized ratings "boosting" or authorized resetting
(starting over at 1500.00) and by servers (assigning new players
ratings not earned through play on the server, such as starting a new
player with a rating earned on another server, or some other awarding
of points not earned through play against other players on the
server).
________________________________________________
Daniel Murphy www.cityraccoon.com/
Humlebæk Backgammon Klub www.hbgk.dk/
Raccoon on FIBS www.fibs.com/
Raccoon on GamesGrid too
It's amusing that the server that has been online the longest, and
that has the largest rated player base, also has the smallest ratings
creep.
-Patti
--
Patti Beadles | He's a brilliant lunatic and you can't tell
pat...@netcom.com | which way he'll jump... you can't dissect him,
http://www.gammon.com/ | predict him-- which of course means he's not a
or just yell, "Hey, Patti!" | lunatic at all. [From the musical Chess.]
This quite interesting Daniel, thank you for your work. I am a little
baffled about a few things. Why is it that servers assign new players
ratings not earned through play on the server or start a new player with a
rating earned on another server and especially "award points not earned"
against other players on the server.
Does this mean if I have a rating on one server I can ask another server to
sign me up with my rating from the first server?
Another interesting paradox is how can we explain how Netgammon has 81
players rated over 2000, while FIBS has only 9 and GamesGrid only 10. Is
there a rocket scientist in the house?
Regards,
Michael
^____^
HI.
I can be wrong but i had computed the both rating formula of Fibs and
Netgammon respectivly.
I made that to get my rating against JF setting it at 2050.
I played more than 400 5 point matches (lv 7) starting at 1500 exp 0.
using Netgammon formula i reach 1870.
using Fibs formula i'm just 1839.
May be there is a bug in my computation.
But ,may be ,this is a bit of explanation.
BTW when i left Netgammon ,10 months ago ,my rating was 1858 exp +-700.
Actually on Fibs :1768 exp +-650.
Ive been dropped just a little more on Fibs (at beginning ,just by low
rated
opponent).
MD.
Blot on Fibs.
Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.
Michael Strato wrote:
>
> Daniel Murphy wrote in message <37c08e5e...@news.inet.tele.dk>...
> >
> >Other reasons for ratings creep include ratings manipulation by
> >players (unauthorized ratings "boosting" or authorized resetting
> >(starting over at 1500.00) and by servers (assigning new players
> >ratings not earned through play on the server, such as starting a new
> >player with a rating earned on another server, or some other awarding
> >of points not earned through play against other players on the
> >server).
>
> This quite interesting Daniel, thank you for your work. I am a little
> baffled about a few things. Why is it that servers assign new players
> ratings not earned through play on the server or start a new player with a
> rating earned on another server and especially "award points not earned"
> against other players on the server.
>
> Does this mean if I have a rating on one server I can ask another server to
> sign me up with my rating from the first server?
>
> Another interesting paradox is how can we explain how Netgammon has 81
> players rated over 2000, while FIBS has only 9 and GamesGrid only 10. Is
> there a rocket scientist in the house?
that's result of rating creep, but Daniel's creep measurement isn't so
good,
course there is only count how many players is higher rated against how
many lower,
however better and most precise method would be average player rating.
>
> Regards,
>
> Michael
> ^____^
--
_________________________
remove first name from address,
if you want answer with e-mail
>Michael Strato wrote:
>>
>> Another interesting paradox is how can we explain how Netgammon has 81
>> players rated over 2000, while FIBS has only 9 and GamesGrid only 10. Is
>> there a rocket scientist in the house?
>
>that's result of rating creep,
I doubt it. I'd venture that ratings boosting, ratings formula and
opponent quality have much more to do with it.
> but Daniel's creep measurement isn't so
>good,
>course there is only count how many players is higher rated against how
>many lower,
>however better and most precise method would be average player rating.
Could be. In any case, one can never have too much trivial
information, can one.
FIBS GamesGrid Netgammon
# of rated players 6148 2243 4575
% over 1500.00 58.06% 69.10% 72.37%
Median rating 1526.96 1577.14 1593.47
Average rating 1534.78 _______ _______
(some other trivial datahound is welcome to fill in the blanks)
--
Gregg Cattanach
Zox at GamesGrid, VOG
gcattana...@prodigy.net
http://gateway.to/backgammon
Patti Beadles <pat...@netcom.com> wrote in message
news:7pqsah$k...@dfw-ixnews17.ix.netcom.com...
> In article <37c08e5e...@news.inet.tele.dk>,
> Daniel Murphy <rac...@cityraccoon.com> wrote:
> >Over time, ratings systems suffer from "ratings creep" as lower ranked
> >players drop out of the system with greater frequency than higher
> >rated players. This is normal and unavoidable.
>
--
RODRIGO
====================================================================
"Insurance is like marriage. You pay, and pay but you never get anything
back." -- Al Bundy
But the same thing happens on FIBS, if you stop playing. You
eventually get dropped from the ratings list.
It's possible that a low-rated player is less inclined to keep playing
on a pay server than he is on a free one, although I'm not at all
convinced that's what's really happening.
-Patti
--
Patti Beadles |
pat...@netcom.com/pat...@gammon.com |
http://www.gammon.com/ | The deep end isn't a place
or just yell, "Hey, Patti!" | for dipping a toe.
Could your '1000 pt penalty for cheaters' have something to do with
this ? ;-)
-Spud
I've got the precision dice, now all i need is a precision cube....
Patti Beadles wrote:
>
> In article <7ptvop$k18$1...@newssvr03-int.news.prodigy.com>,
> Gregg Cattanach <gcattana...@prodigy.net> wrote:
> >Probably the main reason for ratings creep at GamesGrid (and Netgammon?) is
> >that players that discontinue their membership drop from the ratings list.
> >This doesn't happen at FIBS as it is free. This isn't a defect of the pay
> >servers rating system, just a fact of life.
>
> But the same thing happens on FIBS, if you stop playing. You
> eventually get dropped from the ratings list.
>
> It's possible that a low-rated player is less inclined to keep playing
> on a pay server than he is on a free one,
he keep playing on gamesgrid but without ranking as guest
on the netgammon that is easy one -> reset rating and all go on
> although I'm not at all
> convinced that's what's really happening.
>
> -Patti
> --
> Patti Beadles |
> pat...@netcom.com/pat...@gammon.com |
> http://www.gammon.com/ | The deep end isn't a place
> or just yell, "Hey, Patti!" | for dipping a toe.
--
Highly unlikely. I've subtracted, oh, maybe 4000 ratings points in
the entire history of the server. Max. That would move the average
rating by less than half a point.
>TABLE 2 -- Ratings and Rankings by percentile
>
>Table 2 is perhaps more useful than Table 1 for comparing the ratings
>curve on different servers.
>
>Percentile -----FIBS----- --GamesGrid-- ---Netgammon--- G-F N-G
> Rating Rank Rating Rank Rating Rank
>Top 100% 2239.76 1 2080.47 1 2276.53 1 -41 196
> 99% 1907.52 61 1964.74 22 2040.57 46 57 74
> 95% 1813.72 307 1881.13 112 1915.58 229 67 34
> 90% 1748.10 615 1828.51 224 1833.06 458 80 5
> 85% 1706.85 922 1783.65 337 1786.61 686 76 3
> 80% 1669.00 1230 1737.85 449 1740.29 951 69 2
> 70% 1614.43 1844 1674.34 673 1685.29 1373 60 11
> 60% 1566.99 2459 1626.10 897 1634.59 1830 59 8
>Median 50% 1526.96 3074 1577.14 1122 1593.47 2288 50 16
> 40% 1491.91 3689 1533.07 1346 1550.34 2745 42 17
> 30% 1451.34 4304 1496.25 1570 1509.91 3206 45 14
> 20% 1405.72 4918 1455.37 1794 1460.97 3669 50 6
> 10% 1337.23 5533 1391.68 2018 1392.58 4118 54 1
>Bottom 0% 533.26 6148 961.56 2243 766.68 4575 428 -195
(snip)
I have added the last two columns to this table--the ratings
differences between GG and FIBS and between Netgammon and GG. Note
that except for the exteme bins (first and last for GG/FIBS), the
differences between GG and FIBS are between 40 and 80 and the
differences between Netgammon and GG are less than 20. The differences
at 50%-ile are 50 and 16. So if the ratings systems were "renormalized"
such that the medians were all the same rating, then the numbers would
not be different by very much. This seems to compensate for "ratings
creep". I'm not sure what the conclusion should be, but after this
renormalizing the ratings differences are considerably less
"eye-catching" than the raw numbers.
Chuck
bo...@bigbang.astro.indiana.edu
c_ray on FIBS