Ratings

[liam stephens]

The argument between which is the best system "Elo or Glicko" is largely a red herring.
The one grew out of the other. After several years in use the Elo system was found to contain imperfections,
in particular the drift factor. ( some even claimed it was basically unsound) The Glicko and other systems were introduced to correct these errors.

Here is a simple and logical explanation:

The problem:
Under the points per game system if you are playing someone with a significantly lower rating than yours and they get the draw with the strong side of a 3 mover (even though they lost the other side) then they will gain rating points from the much higher rated player. The Masters section of the US 3 Move National Tournaments is,
I understand, the only remaining case where scoring by rounds or ballot applies.
The points per game system is almost universally more popular and is unlikely to be changed.

The solution is clear:
Where the average tournament rating of the entrants in a particular Tournament is below that of Master level, then master level players or above should not be allowed to enter it and they should save their energies for a higher strength event. Any proper set of Tournament regulations and rating system would enforce such a ruling.

After all, why should a master player get credit for beating the minnows ?
They should not be swimming in the same pond !

In an Olympic field or track event you are not allowed to enter just because you want to appear in the company of the big boys. You have to reach a minimum qualifying standard and earn your place. That is another purpose of the ratings.

[tommyc]

Ive been preaching that for years now Liam...............players should play in their own field according to their standard,its plain and simple.Or endure a handicap.

[Alex_Moiseyev]

Eric, I think Liam and Tommy are absolutely right and pointed to weaknesses of GLICKO system: punishing strong players who doesn't play in their own class events ... true, agree.

In reality this simply means that I shouldn't go to IN and many other (perhaps most) events.

This is correct concept in situation when we have a plenty choice of events for any level. Unfortunately in checkers today, when overall membership is droppiing, events are disappearing or getting lower overturn for years, punishing strong players to participate doesn't make any sense.

The problem could be resolve, if player with rating 200-300 points above average event rating, doesn't earn or lose any rating points. This would resolve the problem mostly and minimize "punish factor".

Perhaps ACF Exec Commitee should make a life important decision - what they want to award and what they want (or don't want) to punish. It is clear now that choice of rating system is not only mathematical or statistic question, but also political.

Igor, I want to remind you, that any statistic model works better only on big numbers. When numbers are relatively small, than we got paradoxes and chances to have those paradoxes are growing.

Regards,

Alex

[liam stephens]

Tommy Canning wrote:

Or endure a handicap.

An interesting idea. I think it happened once. In the good old days of Drummond, Wyllie and Matrins it was a common occurrence in stake matches for the champions to offer a 2 or 3 game start, or the draws as wins,
to their opponent.

Perhaps we should introduce a handicapping system like they have in Horse Racing events, with the weights adjusted. How about a minus 6 game points starting handicap for Alex and Ron in future tournaments. That at least would even up the odds for us stragglers.

[Alex_Moiseyev]

Thats not a bad idea !!! I have improvement - you can "buy handicap" ... lets say $50 for one win.

So, by paying $300 you will get handicap +6 games. We have to play 30-40 games at least to compensate this handicap.

Alex

[tommyc]

Or lets say in a handicap situ...........the ratings dont apply,then ev1 is happy ...its simple.!!

Well ive done it many times in Tys.

[liam stephens]

I would refer you to an earlier posting I made:


The following, you may find, are among the chief distorting elements.

The drift factor ( A sort of Parkinson’s Law – all ratings drift inevitably upwards and those of the top rated player drift up absolutely, the chief Elo fault, refer to Glickman which corrects this trend)

The compression syndrome (scoring by points per game instead of points per round will greatly compress the bandwith of the ratings)

The apples and oranges syndrome ( The dubious practice of including such a wide range of strengths of tournaments from the strongest master tournaments right down to the non serious friendly or fun day events, all in one huge higgledy piggeldy amorphous mass ) I would now add handicap events to this class.

[tommyc]

I would refer you to an earlier posting I made:


The following, you may find, are among the chief distorting elements.

I would now add handicap events to this class.[/color]

They point i was making was,if "handicap events "were excluded from the ratings ,would that solve the probem of high rated players losing on there rating position??..........even when they win every /most of their games.

[liam stephens]

They point i was making was,if "handicap events "were excluded from the ratings ,would that solve the probem of high rated players losing on there rating position??..........even when they win every /most of their games.

Yes, + the fun day and very weak events.

[shinzen]

Ach sooo......eternal problem...
why always about to all want tell-men, who the but at all I don't understand?
Liam told true, no problem ELO or GLicko but problem with big difference between players. Because exist and use coeficent!!!
wherewith higher coefficient, major difference and wherewith height built (high rating) player, thereby lower coefficient. IT MAKES NO DIFFERENCE in ELO and GLICKO. In GLICKO is in addition only next coefficient. Alex, to those would was best play only with field with ratings 2480 and height
Regards,
Igor
p.s. Sample ELO rating k-coefficient (standart, in chess too)
up 2450 10
2350-2450 15
2280-1350 20
2100-2280 25
2000-2100 30
etc etc...

[Alex_Moiseyev]

Poor Elbert Lowder He was most active ACF player ever and for more than 50 years checkers career played in more than 350 events. Shame on him !

My record today is - around 110 events in 13 years. I am dreaming one day to beat Elbert record - this is more respectable goal than rating. In order to do this I have to be very active for another 30 years Hence, I started to play Anglo-American Checkers at age 37 and he started at 15 - isn't a handicap ?

I don't like all this "elite" and "aki-primadonna" behavior and I don't see for myself much difference between high calibre events and small local fun events. Both groups have one similar thing - no money involve !

Practice is practice and nothing could replace it. In IN State tournament I didn't have much competition, but from single games which I played, I find some inaccuracies and weaknesses in my play and this ty helped me to pay more attention to few things and fix it. At least on two occasions I learned new moves in openings - worhty for variety.

For those who wants to be on top among world players and compete for title, I would never recommend to stay away from any opportunity to play in real checkers events.

[william]

Hi everyone

For those who wants to be on top among world players and compete for title, I would never recommend to stay away from any opportunity to play in real checkers events

AMEN!!!!!!

The only way to improve is to play , and I am jealous of anyone who has the possibility to play in more than 1 ty a year ( which to my great sadness is all that I can make due to familly , and geographical situation ) BUT who does not proffit from their luck and play more tys.

However , as john stated , a more subtle math model should be integrated so that most situations are catered for and that he who plays 20 tys a year , or he who plays 1 a year finds his balance correct , and noone gets punished in the end

[Michael Holmes]

Any progress on this?

[Eric Strange]

The Glicko system is the system that is unfair to very active people. I have decided against this completely. I am rolling with the standard Elo system for now and then I will work my way towards systems for higher rated players vs low rated... ways to handle draws that sort of thing.

I have a coder (not my traditional coder) creating a web based Elo calculator. From what he said he can make it completely automated. Load the excel document with a cross table into it, it calculates the ratings for each player and then updates them in the ratings table.

I will make sure its all 100% before I use it for the ratings.

Thank you guys for your arguments and suggestions on ratings.

[Alex_Moiseyev]

I am rolling with the standard Elo system for now and then I will work my way towards systems for higher rated players vs low rated... ways to handle draws that sort of thing.

Eric, I think such thing can be fixed by "direct order", not trying to change formula.

My proposal is simple - if someone has rating 300 (number can be vary) or more points above or below average event rating - he/she must be excluded from calculating rating in this event.

This would open the gates in both ways without impact for anyone: strong players can play in weak events and weak players can play in strong events.

Think about ...

Alex