John, in 3-moves ACF Nationals Master Division historically opponents play 4 games. You may get 4 points if you win overall micro-match, 2 points if match drawn or zero points. If you drew 3 games and lost 1 game - you will get 0 points and your opponent get 4 points. Even if you lose all 4 games your opponent still gets 4 points.
All Eric needs to do - just enter points as they are printed in cross table. The program should be able to find "a median" point and go from there.
In this case it will be neutral for Rating Processor and up to Federations and organisers - how to give points. I gaveup on my idea to count ballots - it make it more complicated and tricky for Eric and raise some other questions.
I think in 3-moves National Major and Minors ACF started count each game.
Regards,
Alex
ACF ratings restoration
-
Alex_Moiseyev
- Posts: 4358
- Joined: Sat Nov 12, 2005 5:03 pm
- What do you like about checkers?: .....
Re: ACF ratings restoration
I am playing checkers, not chess.
-
Eric Strange
- Posts: 438
- Joined: Thu Mar 12, 2009 12:10 pm
- What do you like about checkers?: What's not to like?
- Location: Colorado
- Contact:
Re: ACF ratings restoration
After studying elo, glicko, and David Butlers way I am still undecided. To be honest I do not like any of them for checkers. Glicko (which is used by the WCDF) uses a factor which gives an advantage to someone who doesn't play in many tournaments. So basically I could have 2 2100 rated players and if the guy who plays in a tournament once a month wins he doesn't get as much as if the other person won because that guy only plays every 4 years. The logic behind that is the guy who plays less tournament doesn't play very often so if he wins it means he is definitely higher than a 2100 rating. The reason I don't like this (which I should because it advantages me) is because with internet today many players play checkers daily on the internet and get plenty of practice so there is no reason to get an advantage. It is the same reason i don't agree with being able to compete at a lower tournament level if you haven't played a tournament in 3 years. To use Albert Tucker as an advantage... he is in kurnik practicing checkers atleast 3 times a week. So he was no he was not out of practice to be able to drop to majors.
Glicko derived from the Elo system which gives 2 players of the same rating a 65% chance to win which is part of the formula for calculating the rating. They choose this magical 65% because it is the only thing that seems to actually work in their formula. The output would be incorrect if they used 50% or some other percentage. At least that's what was said in what I read.
David Butlers rating system is too difficult on new tournament players. The rating fluctuates around intensely it seems until you have reached 128 tournament games played. After the 128 games played it takes the average and that becomes your stable ratings. If you calculate the tournament games played by an average working american that would take forever.
I am going to read up more on all of these and see if maybe I am thinking too simply about them. When I decide what I am going to do I will let everyone know.
Glicko derived from the Elo system which gives 2 players of the same rating a 65% chance to win which is part of the formula for calculating the rating. They choose this magical 65% because it is the only thing that seems to actually work in their formula. The output would be incorrect if they used 50% or some other percentage. At least that's what was said in what I read.
David Butlers rating system is too difficult on new tournament players. The rating fluctuates around intensely it seems until you have reached 128 tournament games played. After the 128 games played it takes the average and that becomes your stable ratings. If you calculate the tournament games played by an average working american that would take forever.
I am going to read up more on all of these and see if maybe I am thinking too simply about them. When I decide what I am going to do I will let everyone know.
-
Alex_Moiseyev
- Posts: 4358
- Joined: Sat Nov 12, 2005 5:03 pm
- What do you like about checkers?: .....
Re: ACF ratings restoration
Eric, below is my proposal how to handle ratings for new players. If you like it, you can use this along wiht basic system which you choose.
======================================================================================
1. Setup "provisional period.
A) For new players it can be 4 events but no less than 50 games. Pro-active agressive players can do this in less than one year. I usually play in 7-8 events annually, most players have 3-5.
B) For players on list "B" (who came back) provisional period is double short: 2 events but no less than 25 games.
2. In each event during provisional period rating is counted independently from scratch, like it's your first tournament (or match).
3. So - at the end of provisional period player has 4 (or 5) provisional ratings. Permanent rating is counted as average of those 4 (or 5) numbers.
4. Provisional rating is based on player opponents average rating and his / her performance.
Lets say average rating of his / her opponents is 2100 and (s)he finished with 50% points. This means that his / her Provisional Rating (PR) #1 (or #2, or #3, or #4) is 2100. It doesn't have any influence on his/her PR in the next event. Each PR is independent.
5. For opponents of player with PR there will not be any changes - rating of PR player shall not be counted for his / her opponents until it becomes permanent.
Smart ? Smart !
Clever ? Clever !
Simple ? Simple !
Thinking several hours about this procedure, I didn't find anything visual wrong and any obvious weaknesses. This system has very high weighting and average numbers several times which should be a good protection for unusual cases ... like some genius started playing in ACF events etc. I also don't see any possibilities here for cheating and/or abuse system. Even if you play in weak events or high calibre events, at the end of provisional period your permanent rating will be close to reality.
Regards,
AM
======================================================================================
1. Setup "provisional period.
A) For new players it can be 4 events but no less than 50 games. Pro-active agressive players can do this in less than one year. I usually play in 7-8 events annually, most players have 3-5.
B) For players on list "B" (who came back) provisional period is double short: 2 events but no less than 25 games.
2. In each event during provisional period rating is counted independently from scratch, like it's your first tournament (or match).
3. So - at the end of provisional period player has 4 (or 5) provisional ratings. Permanent rating is counted as average of those 4 (or 5) numbers.
4. Provisional rating is based on player opponents average rating and his / her performance.
Lets say average rating of his / her opponents is 2100 and (s)he finished with 50% points. This means that his / her Provisional Rating (PR) #1 (or #2, or #3, or #4) is 2100. It doesn't have any influence on his/her PR in the next event. Each PR is independent.
5. For opponents of player with PR there will not be any changes - rating of PR player shall not be counted for his / her opponents until it becomes permanent.
Smart ? Smart !
Clever ? Clever !
Simple ? Simple !
Thinking several hours about this procedure, I didn't find anything visual wrong and any obvious weaknesses. This system has very high weighting and average numbers several times which should be a good protection for unusual cases ... like some genius started playing in ACF events etc. I also don't see any possibilities here for cheating and/or abuse system. Even if you play in weak events or high calibre events, at the end of provisional period your permanent rating will be close to reality.
Regards,
AM
I am playing checkers, not chess.
-
Eric Strange
- Posts: 438
- Joined: Thu Mar 12, 2009 12:10 pm
- What do you like about checkers?: What's not to like?
- Location: Colorado
- Contact:
Re: ACF ratings restoration
I do like some aspects of your provisional rating idea. I think once someone has a permanent rating but drops out of the picture if they come back there is no reason to start provisional again because they already have an established rating. Currently I am leaning more towards Elo rating system over Glicko. As far as calculating international tournament I believe I can calculate all ACF players who are rated in the WCDF Glicko and come out with an average point difference and create an algorithm based on that in order to better calculate the international games. The glicko system gives too much advantage to players who play in less tournaments. Players who play in a lot of tournament are almost punished for doing so which I don't agree with because people like me who practice online a lot don't need an advantage. Elo system is what is being used in yahoo, pogo, kurnik etc etc.
I am also thinking to incorporate some sort of calculation with draws. Where if say I played you in a tournament and got a draw you wouldn't be punished. I dont believe a master's points should be punished because both players played the line properly. But I believe someone with a substantially lower points should be rewarded for being able to draw someone on such a large rating difference. I am not 100% on this yet but if your 400 points above someone. a draw should not hurt you in my opinion. We will see how this could effect rating as a whole.
I am also thinking to incorporate some sort of calculation with draws. Where if say I played you in a tournament and got a draw you wouldn't be punished. I dont believe a master's points should be punished because both players played the line properly. But I believe someone with a substantially lower points should be rewarded for being able to draw someone on such a large rating difference. I am not 100% on this yet but if your 400 points above someone. a draw should not hurt you in my opinion. We will see how this could effect rating as a whole.
-
Alex_Moiseyev
- Posts: 4358
- Joined: Sat Nov 12, 2005 5:03 pm
- What do you like about checkers?: .....
Re: ACF ratings restoration
We had Elo in USSR and I beleive they still use it today in Russia and sone other countries - former USSR Republic. It is very neutral to active or not very active player and strictly depends on numbers - ideal math model. I was able to calculate my "estimate" my new rating without paper and pen - just looking at the final standings in cross table. It's very easy and logical system.
Concerning "punishment" of strong players ... I propose to you for now leave this behind. We can talk about this later and it may require additional ACF review/approval. Rating cannot depend on what opening you played (!!!) - with limited or high scope, uneven, critical etc
Regards,
Ale
Concerning "punishment" of strong players ... I propose to you for now leave this behind. We can talk about this later and it may require additional ACF review/approval. Rating cannot depend on what opening you played (!!!) - with limited or high scope, uneven, critical etc
Regards,
Ale
I am playing checkers, not chess.
-
Ingo_Zachos
- Posts: 1286
- Joined: Sun Nov 13, 2005 7:41 am
- Location: Dortmund, Germany
- Contact:
Re: ACF ratings restoration
Dear Eric,
I found the following link to be very interesting:
http://blog.chess.com/kurtgodden/elo-to ... -explained
Note also that currently the Elo-System of the FIDE is under discussion:
http://www.chessbase.com/newsdetail.asp?newsid=5418
especially the "k" factor is under scrutiny.
I can't givbe you advice which system is suitable, as there is no "perfect" system.
Glicko has some advanteges, but these are paid by other disdvantages and it is the same with the old ACF-System.
Tough I would always prefer an Elo-based system, may it be Glicko or any other, to the ACF-approach, as the scientific approach of ELO/Glicko looks more profound, but it is harder to calculate.
Here are other factors that make ratings in checkers even more problematical then in chess:
1. You have two styles: GAYP or 3-move.
There are players specialising in one style or are better in one then in the other for several reasons, most notbaly in some countries like Barbados or Germany GAYP is dominant, in others like the USA 3 move is more common, and others like the UK have both in about the same number of events.
How do you compare a "3move player" to a "GAYP player"?
You could use two ratings, but then you would have much less events for both styles and tat makes the rating less reliable, though the validity then would increase. Same problem as a Tennis ranking: Nadal is better on clay, but Federer better on grass, so a wight must be used to determine the no. 1.
2. In 3-move some ballots produce a 4-0 to the stronger players, others 3-1 to the stronger player, and LUCK determines a good portion of a players points scored.
You can turn to matchpoints scores, but then maybe a hard fought and original draw that was an achievment would not help the player that achieved it.
As I always pointed out, ratings are not reflecting strengths, but rather success of the past and is a weighted average over your performances with the latest one more important then the older ones.
So dont argue with comparisons about comparing strenghts with a rating system.
It does not get you very far.
It just really only gives an assesment of the success of the past.
Try to calculate a few tournaments with each sytem and try to find a software that is most convenient to you, as THAT is a also a very iimportant factor:
You must be able to hanlde the "tools".
If you take a look at the various programs for each system, which one do you think is best?
What was easiest to use?
Does it have all the features you woud expect?
Maybe that "pragmatical" advantgeof one system to have a program that works stable and is easier to handle, is the best benchmark after all.
Greetinx from sunny Dortmund,
Ingo Zachos
I found the following link to be very interesting:
http://blog.chess.com/kurtgodden/elo-to ... -explained
Note also that currently the Elo-System of the FIDE is under discussion:
http://www.chessbase.com/newsdetail.asp?newsid=5418
especially the "k" factor is under scrutiny.
I can't givbe you advice which system is suitable, as there is no "perfect" system.
Glicko has some advanteges, but these are paid by other disdvantages and it is the same with the old ACF-System.
Tough I would always prefer an Elo-based system, may it be Glicko or any other, to the ACF-approach, as the scientific approach of ELO/Glicko looks more profound, but it is harder to calculate.
Here are other factors that make ratings in checkers even more problematical then in chess:
1. You have two styles: GAYP or 3-move.
There are players specialising in one style or are better in one then in the other for several reasons, most notbaly in some countries like Barbados or Germany GAYP is dominant, in others like the USA 3 move is more common, and others like the UK have both in about the same number of events.
How do you compare a "3move player" to a "GAYP player"?
You could use two ratings, but then you would have much less events for both styles and tat makes the rating less reliable, though the validity then would increase. Same problem as a Tennis ranking: Nadal is better on clay, but Federer better on grass, so a wight must be used to determine the no. 1.
2. In 3-move some ballots produce a 4-0 to the stronger players, others 3-1 to the stronger player, and LUCK determines a good portion of a players points scored.
You can turn to matchpoints scores, but then maybe a hard fought and original draw that was an achievment would not help the player that achieved it.
As I always pointed out, ratings are not reflecting strengths, but rather success of the past and is a weighted average over your performances with the latest one more important then the older ones.
So dont argue with comparisons about comparing strenghts with a rating system.
It does not get you very far.
It just really only gives an assesment of the success of the past.
Try to calculate a few tournaments with each sytem and try to find a software that is most convenient to you, as THAT is a also a very iimportant factor:
You must be able to hanlde the "tools".
If you take a look at the various programs for each system, which one do you think is best?
What was easiest to use?
Does it have all the features you woud expect?
Maybe that "pragmatical" advantgeof one system to have a program that works stable and is easier to handle, is the best benchmark after all.
Greetinx from sunny Dortmund,
Ingo Zachos
You can rent this space for advertising, if you like!
-
Pedro Saavedra
- Posts: 205
- Joined: Fri Nov 18, 2005 10:54 am
Re: ACF ratings restoration
For what it's worth, I recently attended the Joint Statistical Meetings (I am a professional statistician) and in the program there was one round table on the chess rating system. I did not attend the session, but here is the abstract:
Number: 174
Type: Roundtables
Date/Time: Monday, August 3, 2009 : 12:30 PM to 1:50 PM
Sponsor: Section on Statistics in Sports
Abstract - #303719
Title: Rating the Competition: Lessons from the World of Tournament Chess
Author(s): Mark Glickman
Companies: Boston University School of Public Health
Address: EN Rogers Mem Hospital (152), Bedford, MA, 01730,
Keywords: Paired comparisons ; chess ; rating ; Elo ; playing strength
Abstract: Sports fans, handicappers, and fantasy league members constantly seek the ultimate quantitative approach to evaluate their favorite teams or players. Unfortunately, there are as many methods to rate competitors as there are quantitative sports analysts. So what's a sports fan to do? It turns out that this question has been considered in a rigorous manner as far back as the 1950s in the context of tournament chess. As Ratings Committee chairman of the U.S. Chess Federation for the last 17 years, I will describe the current approach to rating tournament chess players in the U.S., and some of the key insights learned over the years into making chess ratings reliable measures of playing strength. I will also discuss how ideas from chess rating methods can be translated to rating competitors in more traditional sports.
--------------------------------------------------------------------------------
Number: 174
Type: Roundtables
Date/Time: Monday, August 3, 2009 : 12:30 PM to 1:50 PM
Sponsor: Section on Statistics in Sports
Abstract - #303719
Title: Rating the Competition: Lessons from the World of Tournament Chess
Author(s): Mark Glickman
Companies: Boston University School of Public Health
Address: EN Rogers Mem Hospital (152), Bedford, MA, 01730,
Keywords: Paired comparisons ; chess ; rating ; Elo ; playing strength
Abstract: Sports fans, handicappers, and fantasy league members constantly seek the ultimate quantitative approach to evaluate their favorite teams or players. Unfortunately, there are as many methods to rate competitors as there are quantitative sports analysts. So what's a sports fan to do? It turns out that this question has been considered in a rigorous manner as far back as the 1950s in the context of tournament chess. As Ratings Committee chairman of the U.S. Chess Federation for the last 17 years, I will describe the current approach to rating tournament chess players in the U.S., and some of the key insights learned over the years into making chess ratings reliable measures of playing strength. I will also discuss how ideas from chess rating methods can be translated to rating competitors in more traditional sports.
--------------------------------------------------------------------------------
-
Eric Strange
- Posts: 438
- Joined: Thu Mar 12, 2009 12:10 pm
- What do you like about checkers?: What's not to like?
- Location: Colorado
- Contact:
Re: ACF ratings restoration
I am banging my head against a wall trying to decide which way I want to go with this and I appreciate everyones help with articles and whatnot to help make my decision.
Now what I want from everyone because rating effects all of you as well as me. I would like ALL of your solid opinions.
Glicko, or Elo and why. (I've decided David Butlers is out of the question)
If I can get a decent number of votes for one style over the other then it would help to resolve my issue and also keep all of the ACF members happy.
So please give me your answers!!! and if you don't feel comfortable answering here you can send me a private message.
THANK YOU!
I am leaning more towards Elo for my vote.
Now what I want from everyone because rating effects all of you as well as me. I would like ALL of your solid opinions.
Glicko, or Elo and why. (I've decided David Butlers is out of the question)
If I can get a decent number of votes for one style over the other then it would help to resolve my issue and also keep all of the ACF members happy.
So please give me your answers!!! and if you don't feel comfortable answering here you can send me a private message.
THANK YOU!
I am leaning more towards Elo for my vote.
-
Alex_Moiseyev
- Posts: 4358
- Joined: Sat Nov 12, 2005 5:03 pm
- What do you like about checkers?: .....
Re: ACF ratings restoration
Eric, count my vote: my choice is ELO with potential adjustments for new players and provisional period/ratings.
Reasons of my choice:
1) easy to understand
2) easy to maintain
3) neutral to players strength
4) neutral to players activity
Alex
Reasons of my choice:
1) easy to understand
2) easy to maintain
3) neutral to players strength
4) neutral to players activity
Alex
I am playing checkers, not chess.
-
Eric Strange
- Posts: 438
- Joined: Thu Mar 12, 2009 12:10 pm
- What do you like about checkers?: What's not to like?
- Location: Colorado
- Contact:
Re: ACF ratings restoration
I forgot to give my reasons why I vote ELO.
1. easy to understand
2. Many Many different software applications and web based apps which calculate it that I can have altered to store ACF database in it.
3. Makes more sense for checkers as a whole.
I would probably stick with one k-factor for EVERYONE ...probably 15 (I see K-factor changes for ratings over 2400) I need to read more on why before I decide one k-factor.
stable rating out of provisional status would be the average of so many tournament/so many games... based on the level of the provies opponents.
I need more votes on what system to use and also if someone has a good idea on how I should calculate international matches for ACF players who play against players who don't have ACF rating?
I'm thinking come up with an algorithm or system that will calculate what a glicko players rating would be (or atleast somewhat close) in the ELO system????? just a thought....
1. easy to understand
2. Many Many different software applications and web based apps which calculate it that I can have altered to store ACF database in it.
3. Makes more sense for checkers as a whole.
I would probably stick with one k-factor for EVERYONE ...probably 15 (I see K-factor changes for ratings over 2400) I need to read more on why before I decide one k-factor.
stable rating out of provisional status would be the average of so many tournament/so many games... based on the level of the provies opponents.
I need more votes on what system to use and also if someone has a good idea on how I should calculate international matches for ACF players who play against players who don't have ACF rating?
I'm thinking come up with an algorithm or system that will calculate what a glicko players rating would be (or atleast somewhat close) in the ELO system????? just a thought....