| From | Message |
mattafort
6/13/2003
18:07:37
[ report this post ] |
 Subject: Newcomer with 2610 in rating
Message:
Say a newcomer at GK, started out with 20 games with cyrano
And resigned every game after 3 moves.
For every game he can add 3010-400= 2610 to his score.
And average after 20 games would be the same.
So we now have a player at GK with 2610 in rating, and 20 losses in 20 games!
A new brunetti
Besides, it would give cyrano 20 more points.
|
kremator
6/13/2003
18:12:11
[ report this post ] | HEY
Message:
Don't give people wicked ideas.
|
very_bad_player
6/13/2003
18:47:00
[ report this post ] | What
Message:
do you mean with a new Brunetti? Alex is a fine player, don't think otherwise ...
Regards,
VBP
|
mattafort
6/13/2003
18:49:29
[ report this post ] |
Message:
what do you think I mean, very_bad_player ?
|
lordoftherings
6/14/2003
04:23:22
[ report this post ] | there is not ??
Message:
a rule who say that a newcomer cant play against the same player in his first X games for exemple.
but this new player will lose the other games if he is not so strong like a player of 2610 .no?
|
aggropolis
6/14/2003
06:44:39
[ report this post ] | I think
Message:
that there should be a limit for 1200 players so that they can't play players over say 1800 until they have finished 5 games.
|
bogg
6/14/2003
08:07:22
[ report this post ] | aggropolis
Message:
I would have never played a game at GK if that rule was in effect.
|
mattafort
6/14/2003
08:17:15
[ report this post ] | System is not perfect
Message:
My example, although extreme, shows that GK-rating system is not perfect.
But I would not say that GK-system is VERY bad.
It works in most.
Gives a hint about players strenght.
That is for me the most important.
I don't think the ratings need to be an instrument to tell us who is BEST.
As if we are having a COMPETITION for Best Rating
where the rules have to be 100% fair.
|
mattafort
6/14/2003
08:17:42
[ report this post ] | System is not perfect
Message:
My example, although extreme, shows that GK-rating system is not perfect.
But I would not say that GK-system is VERY bad.
It works in most.
Gives a hint about players strenght.
That is for me the most important.
I don't think the ratings need to be an instrument to tell us who is BEST.
As if we are having a COMPETITION for Best Rating
where the rules have to be 100% fair.
|
superblunder
6/14/2003
11:17:18
[ report this post ] | That is just my point...
Message:
in the other threads, That the rating system is very simple here on gK, and too primitive to be accurate, mattafort just points one of many examples of how to abuse this simple rating system, but we who want to see a more fair system seem to be in the minority, as many would have to lose rating points in order to adopt an ELO type of sytem. I myself would probbaly also lose rating points if it were ELO, so I do not say this in selfish or 'jealous' way, I just want to see this great chess site rank players more appropiately.
|
astinkyfart
6/14/2003
11:39:19
[ report this post ] | some people will be unhappy no matter what.
Message:
the rateing system is fine. why do i say this? because no matter what system you use its still man made and is only a gauge of strength compared to others useing the same system. the example mattafort uses may be possible but very unlikely. what makes elo so much better?? this is only a man made measurement. we know who the stronger players are and the weaker. i dont care what system we use its all the same to me. i dont think this system is unfair at all. so keep it, change it , get a new one it doesnt matter to me because unless god makes a rateing system that says this is the rock solid way to rate a chess player, then no matter what system you use its just a guess and estimate if you ask me.
|
tambori
6/14/2003
13:44:40
[ report this post ] | very right stinky...
Message:
i agree....just play chess you people....points are not everything in life...ok, any rating system can have certain disadvantages but so do the political systems...that of course does not mean that we have to change them all the time or find them useless...
|
aggropolis
6/14/2003
14:03:08
[ report this post ] | bogg
Message:
Then we can draw a fast line to the fact that you only care for rating.
That really doesn't reject my suggestion as it suggests that some players (read you and others) here just want to get a high rating and are not willing to play 100's of games to proove their skill and take the hard way like some of the top players (read cyrano and brunetti).
Ever thought of really enjoying the game? You know it isn't just points. I think gameknot will do just fine without players that only focus at rating.
Kizzes and hugz for da ignorant blemishes.
|
bogg
6/14/2003
14:09:50
[ report this post ] | asf
Message:
is correct in general terms. I can't think of a system that is perfect. The ELO system only claims to be accurate to within 50 points. I would estimate that the system here at GK is accurate to within about 400 points for established players and within 1000 points for newer players. That would indicate that if the rating difference between 2 players with established GK ratings was greater than 800 points the higher rated individual is nearly certain to be the stronger player. Yeah, I guess they are fairly comparable at that.
|
north
6/14/2003
15:20:02
[ report this post ] | This would only work
Message:
This would only work for the first 20 games... after that the *actual* rating system comes into play, and our "New Brunetti" would drop by 1 point each game. So the above cheat would only work if GK used every game was like those first 20...
-Derek
|
astinkyfart
6/15/2003
00:33:57
[ report this post ] | heres what i dont get, and im serious
Message:
bogg elo is accurate with in 50 points?? says who?? actually elo is 100 percent accurate by elo rateings because why?? thats elo rateing. gameknot is accurate 100 percent because that gameknot rateing. bogg says elo only claims to be accurate within 50 points. so what do they use to measure the possible margin of error. nothing!! because to do this they would have to have an existing perfect system which doesnt exist so how do you know elo isnt a completely absurd rateng system that wont let players reach 3000 rateing like they are suppossed to?? you may say because no one is that good,,,, but according to who? ok im done i feel like im talking in circles
|
gohan2535
6/15/2003
01:06:41
[ report this post ] | rating means...........
Message:
nothing take a look at olympio he has a low rating but play him and find out
|
bogg
6/15/2003
02:25:27
[ report this post ] | asf
Message:
The Elo rating system is based upon statistics, the fact that nearly everything dealing with humans, when measured fits into a bell curve. In an earlier post someone complained about the shoddy testing that Arpad Elo did when deriving his system. Bunk, the system is intuitively obvious, all he needed to do was to verify that chess, like nearly everything else dealing with us humans, fits into a bell curve. But I digress, when Elo says that the ratings are accurate to a 50 point range he is stating that a player with a current rating of 2500 has a 95% probability of having a playing strength that falls between 2450 and 2550. In my earlier post my 1000 point number for GK ratings is a guess from observed anomalies in the ratings, but not backed up by any serious mathematics, for this I apologize.
GK is not 100% accurate, nothing dealing with the human condition can be, but the GK rating system doesn't pretend to be accurate:
1) The GK ratings system assigns a different value to identical results, i.e. a 2 point match split 1 - 1 made up of either 2 draws or a win for each player.
2) You win the same number of points for beating a player 350 above your rating (3% probability) and for beating a player 350 points below your rating (97% probability). The previous probabilities are from memory but I guarantee they are within 2%.
A functional rating system is supposed to be predictive of the relative playing strengths of the players that make up its pool. It doesn't matter that the GK ratings don't compare to some other rating bodies ratings, what matters is that the players ratings within the GK pool don't compare to each other!
If you want to understand the basis for rating systems I suggest you read Arpad Elo's book The Rating of Chess Players Past and Present.
|
mattafort
6/15/2003
04:29:39
[ report this post ] | o lympio is a special case
Message:
olympio is a special case
if you go way back in his "past history"
you will find that he is a very strong player.
He once was above 1800 in rating!
Then suddenly he got lots of MYSTERIOUS losses, until he was downto 1100.
So in this case we have
A Wolf In Sheep's Clothing = olympio
|
aggropolis
6/15/2003
04:31:29
[ report this post ] | bravo
Message:
Olympio has already written this in another thread.
|
mattafort
6/15/2003
04:37:53
[ report this post ] | yes
Message:
at 30 april 2003 he had 1840
then he has started a lot of games just to lose them
90 straight losses - most of them within 3-6 moves
then he got 909(!) in rating
|
mattafort
6/15/2003
04:46:14
[ report this post ] | olympio was banned 1 month, yes
Message:
posted by olympio 5/07/2003 19:53:20
Subject: coyotefan
Message:
well.. you really REALLY made yourself look like an idiot there
a. i AM a paying member.. you must be blind or just stupid to have not noticed that
b. my winning percentage has nothing at all to do with my ability to give opinions on this matter..
c. i won my first 30 or so games against opponents rated from 1200 all the way up to 2400.. then i decided to resign all my games and get my rating as low as i could for fun. so if chess skill (as you imply) really implies competence in offering suggestions on the options a chess SITE has.. then i would be more than qualified.
but it doesn't.. only the biggest of morons would think that.
Note from webmaster: olympio has lost his forum posting privileges for 1 month as a result of this post (personal insults).
|
aggropolis
6/15/2003
04:56:37
[ report this post ] | svenne
Message:
And what will you achieve by posting this?
|
bogg
6/15/2003
11:18:07
[ report this post ] | An open reply to a private message
Message:
In a private message a person politely told me that they were entitled to their opinion and I was entitled to mine regarding the topic of ratings here at GK. Without naming names, as I feel that is rude and I am only commenting on one statement, I am going to reply to that message publicly as it seems that I am not stating my point clearly.
I am not stating an opinion. If I was saying that I didn't like the rating system here at GK that would be an opinion. I am saying that GK does not have a rating system. I have given the definition of the term rating system in several posts. This is not my definition, it is the definition. GK has a point race like the American Contract Bridge League, where nearly anyone that can keep breathing long enough can achieve the life master title. This fact is blurred because everyone starts out with a provisional rating that is calculated like an actual rating and by the fact that a player can lose points by losing a game. As the points won or lost after a contest is completed have no relationship to the results probability the system at GK can't be called a rating system.
If it wasn't called a rating system I would quit pointing out that it isn't a rating system.
|
astinkyfart
6/15/2003
13:25:12
[ report this post ] | bogg
Message:
i want to make my comment public. you did not mention my name so i dont know if you even meant me but to let you and everyone else know (just in case you did mean me) i want to post my message here. being rude wasnt my intention and i hope you werent talking about me :) comment>>Private message from astinkyfart:
Hello, bogg!
i mean no disrespect to you in my post. i do understand your point but i also have my views. your a great player and i wish you would stay on gameknot. this really is a good site. either way good luck to you. :)
|
astinkyfart
6/15/2003
13:52:20
[ report this post ] | oops
Message:
ok sorry. it was my statement. he wasnt calling me rude. he stated nameing names would be rude so i completely misunderstood the post sorry.
|
sjoroos
6/15/2003
14:27:07
[ report this post ] | Have you thought about 1 thing ?
Message:
Bogg is actually right ! This site does not have a
rating system, or mostly a completely worth-less one !
I've said this before, and i will tell you to
stop and think about it once more:
Definition: A rating system is a system that meassures
an individuals playing strength in realtion to others.
Question: Can you REALLY stand up and say that
the system used at G.K gives the slightest hint of
any players strenght ??
Well, in some cases it might do, but in many cases
the points differs from the actual strength by a
HUGE margin! And why is that ? It's because the
system take's as in consider who can play the MOST
games as much as it considers WHO you play!
It also has WAY to big margins between the borders
that decides how many points is earned\lost.
Personally i think it's ridiculous that you get any
point at all for "slaughtering" opponents more than
800 points below yourself!! But as many claim they
are ONLY doing it of kindness against their opponents
i guess we will have to beleive them ;-)
I still (pleeeease) pray for the implementation to
play unrated game so we could watch all those
tutors play those games as unrated instead (or would they?).
Conclusion: If the system doesn't live up to the
definition, why have one at all ?? I see 3 things
that i could accept:
1. Change to a more accurate system.
2. Skip the whole system
3. Change the description of it, so it clearly shows
that it's not a "rating" system, but a points race.
[Preferebly #1]
|
astinkyfart
6/15/2003
14:33:10
[ report this post ] | as i stated before
Message:
a different system would be fine with me. i do see how this system can be manipulated. but.. it still gives an accurate measurement of players ability to a certain extent. IF... an opponent doesnt inflate their rateing by playing under rated opponents you can get an idea of a players strength. if mike changes it this is fine with me. i think my problem is that im ignorant of other rateing systems.
|
bogg
6/15/2003
18:44:17
[ report this post ] | sjoroos
Message:
A drawn game doesn't affect either player's 'rating' if both players have established 'ratings', hmm.
|
sjoroos
6/16/2003
13:32:12
[ report this post ] | Yep bogg..
Message:
.. i know :-) There's MUCH to wish in the actual
"rating" system used here, for an example the fact
that you don't get A SINGLE point for drawing a
399 higher rated opponent :-)
And as "astinkyfart" states: We will NEVER get to
the cheaters even which system is used, but at
least the most of us would have a proper rating
showing approximately our right strength.
|
kaf1
6/16/2003
13:58:36
[ report this post ] | I agree with sjoroos
Message:
sjoroos writes:
-----
"Conclusion: If the system doesn't live up to the
definition, why have one at all ?? I see 3 things
that i could accept:
1. Change to a more accurate system.
2. Skip the whole system
3. Change the description of it, so it clearly shows
that it's not a "rating" system, but a points race. "
-----
This conclusion is right on. I agree 100%. I also prefer option 1.
|
mattafort
6/16/2003
14:37:01
[ report this post ] | Another extreme.
Message:
There are some ways to manipulate this points-counting-system (rating seems to be wrong word)
to acheive an improper high point.
It is even easier to lower your points.
You can do like olympio "get my rating as low as i could for fun",
but you can also do it more systematically
to set some sort of "World Record" in low rating.
eli_jah has now a rating of 196.
This is achieved by starting games and resigning after 3 moves.
In 660 played, eli_jah has lost 616
with a "Longest losing streak" of 319 games.
Again this example is EXTREME.
98% of Gameknot players is NOT mis-using the points-system.
/mattafort
|
lexherman
6/16/2003
15:12:32
[ report this post ] | Have you thought about 1 thing?
Message:
sjoroos your statement about the rating system was very clear. I am behind it 100 %.
|
squijum
6/16/2003
19:08:16
[ report this post ] | astinkyfart wrote:
Message:
"i do see how this system can be manipulated. but.. it still gives an accurate measurement of players ability to a certain extent."
It DOESN'T give an accurate measurement of players' ability to ANY extent BECAUSE it can be so easily manipulated. Since there will always be people who manipulate the system, anyone whose rating is effected by playing a game against these people will be either under-rated or over-rated through no fault of their own...as will be the people who play them, ad infinitum. You are gravely mistaken if you believe that because you do not abuse the system your rating is accurate. In truth, the ratings of every player within the system are corrupted by the few who DO abuse the system. There is no way around this fact, even if you don't, have never, and will never cheat.
I agree that the "rating system" on Gameknot is absolutely ridiculous, and must be improved drastically.
|
astinkyfart
6/16/2003
19:54:22
[ report this post ] | i disagree
Message:
not because it can be but if its manipulated. you cant tell me that most players rated 2000 or better on gameknot arent good players. they are!!! to some extent you can get an idea of a players ability this is a fact! by squijums definition a gameknot rated player of 800 is equal to a gameknot rated player of 2000 because gameknot rateing doesnt give a measurement to ANY extent. even if you dont like gameknots system you cant believe this. why do you think bogg said if there was a limitation to play certain rateings or below he would never have played a single game on gameknot. even he knows that this system does give some idea of a players ability.
|
squijum
6/16/2003
20:18:02
[ report this post ] | I wrote:
Message:
"It DOESN'T give an accurate measurement of players' ability to ANY extent..."
Note the word "accurate". Sure, it gives some vague, unreliable idea, but it certaintly isn't accurate.
|
dorisia
6/16/2003
21:03:09
[ report this post ] | I write:
Message:
There does not exist a practical rating system which is accurate.
In particular the Elo rating system is not only not accurate, it is also not based on sound mathematical principles, it has never been published in a renowned mathematical journal and has never been peer-reviewed by a mathematician. Instead it was published in the Journal of Gerontology.
About everything mathematicians have said about the Elo system is the following note in a book:
"His statistical testing is unsatisfactory to the point of being meaningless; he calculates standard deviations without allowing for draws, he does not always appear to allow for the extent to which his test results have contributed to the ratings which they purport to be testing, and he fails to make the important distinction between proving a proposition true and merely failing to prove it false."
- J.D. Beasely, The Mathematics of Games.
The Elo system is nothing more than a misplaced exercise in number crunching.
Therefore I don't understand why the GameKnot rating system is blamed of being inaccurate, while the Elo system is inaccurate as well.
At present the most accurate system is the Sonas rating system (which was derived empirically), but it isn't 100% accurate.
|
squijum
6/16/2003
21:46:44
[ report this post ] | Well...
Message:
I don't really care about the Elo rating system; I merely observed that the Gameknot rating system is absolutely terrible. The fact that the Elo rating system is flawed doesn't magically absolve the Gameknot rating system of its own (glaring) flaws.
|
dorisia
6/16/2003
22:59:30
[ report this post ] | Flaws
Message:
in the GameKnot system become apparent when a player plays many games, such as more than 1000. The same flaw exists in the Elo system:
Let two players, each with an established Elo rating of 2000, play a match of 1000 games against each other. Now player A turns out to be slightly stronger than player B; in fact she wins 600 games, instead of the expected 500. Her new Elo rating will be 3000, when a conservative constant K of 10 is used (this constant is used for Grandmasters), and it will be 4000 if a more "realistic" constant K of 20 is used. Some Federations, such as the USCF, have become aware of this glaring flaw in the Elo system and have recently made amendments to their adaptation of the Elo system, in order to prevent such "accidents" from happening; but these amendments are not part of the Elo system itself, they are just patches. Even when much smaller matches are played, this flaw is still at work, only to a much lesser extent so that it goes unnoticed.
Another flaw in the Elo system is that it uses a logistic curve, where statistical evidence suggests a linear function; this simply means that ratings are calculated inaccurately for some players and accurately for others; the accuracy depends on what performance a player happened to have.
Another flaw is that Kasparov has to perform 15 points above his rating in order to maintain his rating at its present level. This means that every year the Elo system steals 15 rating points from Kasparov; the same applies only to some of Kasparov's colleagues, so that it can be said that the Elo system puts Kasparov at a disadvantage.
And there are more flaws in the Elo system.
|
bogg
6/17/2003
00:12:15
[ report this post ] | dorisia
Message:
I would like to see your math!
In the first example using the ELO rating system the 2 players will have ratings of about 2040 vs. 1960 depending on the order of the games. Of course this can be manipulated by creating long winning streaks but long winning streaks between players of comparable skill is an unrealistic premise.
Your second point isn't very specific so all I will say is that the risk/reward ratios are constant for each rating difference irrespective of the performance a player has.
I can't see the logic in your 3rd point, if anything it is backwards. Because many of his opponents will fall below the point at which he should get 1 point for beating them he gets free fractional points frequently and never gets bit at the other end of the scale (see point 2 below).
There are only 2 flaws with the ELO system that I am aware of and I have heard of no good way around either of them.
1) it is treated as a closed system. Change of player strength is not handled. If a player improves the ratings are deflated. If a players weakens the ratings are inflated. Some systems assign bonus points for a great performance to try to handle this problem.
2) at the extremes players receive 1 point for winning a game that their risk/reward ratio doesn't justify and conversely at the other extreme they lose this same fractional point. Some would call this rounding error. I suppose this could be fixed by making the rating a 5 or 6 digit number, but as it is only accurate to about 2.5 places I don't see the point.
|
bogg
6/17/2003
00:26:04
[ report this post ] | dorisia
Message:
I now understand what you are talking about in your first point. A match longer than 2*K games causes a problem or more specifically multiple games against the same opponent in a contest that contains more than 2*K games. That is why the USCF at least limits match and tournament length for rating purposes. If this 1000 game match is broken up into 50 20 game matches the problem dissapears.
The Elo system is really a way to derive a players performance rating, within limits, over the last 2*K games. If to many games are rated in a single batch it defeats the system. It will also defeat any system that tries to adjust a players rating one game at a time.
|
dorisia
6/17/2003
03:20:02
[ report this post ] | bogg
Message:
In my example of the 1000 games match it is immaterial in which order the points were scored, since the Elo rating is calculated after the match ends. The well-known formula is used:
new_rating = old_rating + K(N - N_expected)
where N is the number of points scored,
N_expected is the expected value of N,
and K is the the attenuation factor.
With K=10 (conservative factor), the winner's new rating will be 3000.
But with a performance of 60%, the Elo system lets correspond a rating difference of about 70 points, which would mean that both players should have ratings such as 2035 and 1965 respectively.
If the above match is split in batches of 20 games, the new rating will be calculated in agreement with the performance:
1.batch: 2020, 1980
2.batch: 2028, 1972
3.batch: 2032, 1968
4.batch: 2034, 1966
5.batch: 2036, 1964
6. and subsequent batches: no rating change.
But with a match of 40 games, the above-described effect already takes place.
Match of 40 games, not split in batches:
2040, 1960
which is already too high/too low
This indicates that when a player plays 40 or more games during a time interval between two subsequent rating calculations, his/her rating will not be calculated accurately.
In many countries, ratings are calculated only once or twice a year, which makes the batches of most players too big.
I found a description of the 1000-game match paradox in the book "Shachmaty i matematika" by Evgeni Gik (Moscow 1983).
My second point I illustrate with the following example. With a rating difference of -300, the Elo system foresees a performance of 15%; but statistical analysis based on data from 266,000 games shows that *White* scores much higher than 15%, more like 25%. On the other hand, when White's rating is 300 points above Black's rating, the Elo system foresees a performance of 85%, and this time it is in agreement with observed data.
See the article by Jeff Sonas at the following link:
www.chessbase.com/newsdetail.asp?newsid=562
My third point (about Kasparov) is also explained in the above-mentioned article. It shows that Elo's predictions are biased against the higher-rated player, and that this effect cost Kasparov 15 rating points in the year 2000.
I don't think many of Kasparov's opponents fall below the point at which he should get 1 point for beating them. This would mean that Kasparov plays many rated games against players rated below 2450.
|
bogg
6/17/2003
03:21:33
[ report this post ] | sonas
Message:
About the sonas rating system.
I have read his documentation and, as I have reported to Jeff, I feel that he makes a few invalid assumptions. I myself would not give to much credence to his thesis. His primary argument seems to be that the plotted data differs from Elo's predicted curve. It should! The plotted data is only meaningful if all of the games are played under the same circumstances, say round one of a tournament. After that, tournament standing starts to play a part and the top players frequently find themselves not needing an optimal result. Hence the bias against higher rated players demonstrated by the plotted data. If his system predicts the results better than Elo's oddly enough this demonstrates its error. Many of the latter games of a tournament are half thrown, agreed drawn, and as his system assumes these are valid contests it is in error. On this point Elo's system also fails but only because the rating body includes these games.
The idea of a dynamic K is a good one, but not one that he originated. I know I was suggesting a variation of it to the USCF 25 years ago and I doubt I was the first.
After my previous comments you may find this hard to believe but I think the Sonas system may be a better practical system than the Elo system! A better system yet would be the Elo system, with variable K, not rating drawn games played towards the end of a tournament.
|
bogg
6/17/2003
03:33:42
[ report this post ] | As my e-mail to Jeff was returned
Message:
I will post it here. If anyone wants to they can forward to his current address.
Jeff:
I think you make a few errors.
1) the advantage of being White, while accurately placed at 35 points is mostly irrelevant. In a 5 round tournament it would affect players performance rating +/- 7 points in a 15 round tournament +/- 2 points. That might in some circumstances change a players rating by 1 point, maybe.
2) your assumption that a linear equation is better than the curve arrived at by Elo because the data points fall more closely to the straight line assumes that all of the games are played under the same circumstances. This is not the case. Tournament standings play a great part in the scoring of individual games.
3) The bias that you noted against high rated players, if actually present, is more than compensated by the points they receive for playing players that they are 100% certain to beat, tournament conditions aside.
4) The biggest bias against high rated players that I am aware of is the rating of games at the end of a tournament where a less than optimal score achieves their goals.
It is natural for the data to demonstrate a bias against high rated players as they more often find themselves not needing an optimal result.
I wonder if your system would do better than Elo's if you only evaluated games from the first half of tournaments? I doubt it. But congratulations on the hard work. I certainly am not willing to put forth that much effort when Elo's system seems intuitively obvious to me.
|
sjoroos
6/17/2003
07:09:48
[ report this post ] | One thing:
Message:
No one has stated that the ELO-system is perfect.
Actually it's not, but one thing is for sure: it is
ONE MILLION times better than G.K's !
We're not asking for Mike to invent the perfect
rating-system as nobody here probably has the
knowledge to do it ! All we're asking for is a RATING
system that is FAIRLY accurate, which i hereby
state that GK's system isn't even close to being.
The silly thing is that the changes needed aren't
big at all, as it would maybe be enough with:
It needs more "borders" where the difference in
rating decides how many points will be awarded.
29 is not needed, like in:
www.schachbund.ch/rating/fl/table_small_f.html,
but maybe one for each 100 points ?
That would also solve such strange things as the
thing that you don't get a single point for drawing
an oppenent 399 points higher rated !!
I also consider it more worth to win over a player
rated 399 points above than one 5 points above !
There must also be a "max" difference where you
don't get any points for beating a MUCH lower rated
oppnent, maybe at 800 or 1000 points.
This (i am SURE) will encourage many players to set
their games at a higher standard here at G.K.
And with the option to play "unrated games" which
is on Mike's to-do-list, the opporunity to play as a
"tutor" will be even better, and also won't fall under
the suspicion that many does it for "free" points.
|
cairo
6/17/2003
07:26:19
[ report this post ] | I
Message:
support what your writing here sjoroos. Especially am I looking forward to Mike's introduction of an option for unrated games.
Best wishes
Cairo
|
bogg
6/17/2003
09:49:13
[ report this post ] | sjoroos
Message:
Actually a very good system, IMO, for this format has been presented to Mike. It works well when the players get to pick their opponents and requires very little programming. It is to extend the provisional rating to 50 games and to then continue using the provisional formula as if the current game is the 50th game. The number of games is arbitrary and can be lowered if more volitile ratings are desired. It could also be a variable depending upon how many games an individual has played, but that would probably add a couple of lines of code. The only addition necessary is to install an edit check so that a player doesn't lose points for beating someone 400+ points below them nor win points for losing to someone 400+ points above them.
Believe it or not this is will arrive at the same answer as the straight line approximation of the Elo system if the same value is used for K where the conversion factor is 800 / K = games. To arrive at the same answer as the Sonas system change the 400 to 425. Additionally, as the players get to pick their colors the 35 point handicap for White should be implimented if you want to approach perfection.
|
sjoroos
6/17/2003
14:55:25
[ report this post ] | :-)
Message:
Nice work bogg !
You seem to have incredible knowledge in this kinda
mathematics and (as far as i understand) this looks
like a fairly accurate system !
Now let's hope that Mike listen to these very well
put arguments and puts a new rating system on the
"to-do-list" as soon as possible :-)
|
bluebabygirl
9/30/2003
22:05:21
[ report this post ] |
Message:
Yes ,I agree that old GK rating system and Elo both have flaws to be sure .Interesting reading the old topics--BBG
|
|
Post a reply to this message:
|
Please log in to your account to play chess online!