Alex Table Tennis - MyTableTennis.NET Homepage
  New Posts New Posts RSS Feed - Rating System is Broken at Lower Levels
  FAQ FAQ  Forum Search   Events   Register Register  Login Login

Rating System is Broken at Lower Levels

 Post Reply Post Reply Page  <12
Author
stiltt View Drop Down
Assistant Admin
Assistant Admin
Avatar

Joined: 07/15/2007
Location: Paradise Inc.
Status: Offline
Points: 467
Post Options Post Options   Thanks (0) Thanks(0)   Quote stiltt Quote  Post ReplyReply Direct Link To This Post Posted: 12/25/2022 at 2:38am
Hi Sean,

Thank you so much for enlightening us. 

Is there a number at which a newbie's rating be at first no matter their level of play? 1500 is an example I get from FICS in chess, I did not play in ages on ICC but I remember it's about the same. They use the Glicko system like the enhanced version on Rating Central. Does RC do the same? Where do we start? 

I have a question about the Reliability Deviation factor that will correct the newbies quickly up or down while their very new appearance will not affect established players as much.

We need communication about the concept of Reliability Deviation, often mentioned as Rating Deviation. If we all learn about it, we will understand better the value of Rating Central v. other rating systems in table tennis.

Richard Feynman used to say that if we can't explain a complex idea with simple words, it means we didn't get it. How do we explain in simple words that the RD way and Rating Central can solve the issue that mjamja is raising? 





Back to Top
Sponsored Links


Back to Top
mjamja View Drop Down
Platinum Member
Platinum Member


Joined: 05/30/2009
Status: Offline
Points: 2846
Post Options Post Options   Thanks (0) Thanks(0)   Quote mjamja Quote  Post ReplyReply Direct Link To This Post Posted: 12/27/2022 at 4:24pm
Originally posted by Tony T Tony T wrote:

Originally posted by soneill soneill wrote:

In a standard USATT event, often I will win 25 and my opponent will lose 25. .

I thought this was only in USATT leagues and not tournaments.

For tournaments, I was told USATT uses some kind of multi pass system where players are adjusted if they gain or lose more than 50 points & then they reprocess with couple more passes.  Is this still true ? 
If so, it seems fair to me.

  

It is true that they adjust if your gain is over 50 points.  The problem is that the adjustment system used does not adjust players to near their actual playing level if that player is entering the tournament with a very low rating (below 1000) and is playing 400 points or more above his entering level.  In the past this kind of individual was very rare so it was not a problem.  As we got more and larger kids training programs we started seeing a lot of these underated kid players.  As a result kids and adult rating events were separated at Nationals and the Open.  This worked pretty good for a few years.  We are now seeing a significant number of adults with this same underating (1000 level beating 2-3 1600 rated players).   The current adjustment system causes excessive drops in the rating of players who have these kind of losses and does not bring the rating in line with the actual playing level until multiple tournaments are played (ie multiple sets of players having excessive rating drops).  We can not separate adults from adults so if this gets worse,  players with stable ratings may not find it that enjoyable to go to these big national events.

Mark
Back to Top
David Marcus View Drop Down
Super Member
Super Member
Avatar

Joined: 10/28/2010
Location: Somerville, MA
Status: Offline
Points: 145
Post Options Post Options   Thanks (0) Thanks(0)   Quote David Marcus Quote  Post ReplyReply Direct Link To This Post Posted: 01/20/2023 at 2:55pm
Originally posted by Lightspin Lightspin wrote:

The simple solution is to offer no prize money outside of the women's and men's open.


Rating-limited events should not have prize money. Unfortunately, it just encourages players to manipulate their rating. Few players would bother to manipulate their rating if all they could win were trophies.
Back to Top
David Marcus View Drop Down
Super Member
Super Member
Avatar

Joined: 10/28/2010
Location: Somerville, MA
Status: Offline
Points: 145
Post Options Post Options   Thanks (0) Thanks(0)   Quote David Marcus Quote  Post ReplyReply Direct Link To This Post Posted: 01/20/2023 at 2:57pm
Originally posted by stiltt stiltt wrote:

I am sure the situation is ripe for Dr Markus and Sean to sit and negotiate with irreductible USATT leaders to find a solution where nobody loses.


Sean and I are not hard to find, if anyone wants to communicate with us about ratings. No one from USATT has done so.
Back to Top
David Marcus View Drop Down
Super Member
Super Member
Avatar

Joined: 10/28/2010
Location: Somerville, MA
Status: Offline
Points: 145
Post Options Post Options   Thanks (0) Thanks(0)   Quote David Marcus Quote  Post ReplyReply Direct Link To This Post Posted: 01/20/2023 at 3:00pm
Originally posted by jfolsen jfolsen wrote:

How would Ratings Central have fixed underrated juniors who get unleashed on a national tournament? I don't know enough about Ratings Central to know this answer.


If the ratings are accurate, and players whose ratings are uncertain are treated as unrated, then rating events are fairer. So, you need an accurate rating system that keeps track of how much data it has on each player.
Back to Top
David Marcus View Drop Down
Super Member
Super Member
Avatar

Joined: 10/28/2010
Location: Somerville, MA
Status: Offline
Points: 145
Post Options Post Options   Thanks (0) Thanks(0)   Quote David Marcus Quote  Post ReplyReply Direct Link To This Post Posted: 01/20/2023 at 3:02pm
Originally posted by stiltt stiltt wrote:

with features of the glicko system like the Australian federation is using, rating/reliability deviation anyone?


Australia is using Ratings Central. Glicko has some similar features, but only looks one level deep and has a different temporal update.
Back to Top
David Marcus View Drop Down
Super Member
Super Member
Avatar

Joined: 10/28/2010
Location: Somerville, MA
Status: Offline
Points: 145
Post Options Post Options   Thanks (0) Thanks(0)   Quote David Marcus Quote  Post ReplyReply Direct Link To This Post Posted: 01/20/2023 at 3:08pm
Originally posted by stiltt stiltt wrote:

I forgot one thing: In chess and backgammon ratings systems, I see players starting at 1500. Maybe that's too high but making people starting around there in addition to rating central will avoid those issues. At first the newbies maybe overrated will drop a lot while their established opponents will not win as many points as the newbies have lost due to their RD being lower, avoiding bad inflation.


You need to be accurate. Giving inaccurate ratings to players only makes things worse. Of course, if the rating is a single number, rather than a distribution, you are forced into ad hoc approaches like provisional ratings. If you have a distribution, then it can be spread out to cover all the possibilities.
Back to Top
David Marcus View Drop Down
Super Member
Super Member
Avatar

Joined: 10/28/2010
Location: Somerville, MA
Status: Offline
Points: 145
Post Options Post Options   Thanks (0) Thanks(0)   Quote David Marcus Quote  Post ReplyReply Direct Link To This Post Posted: 01/20/2023 at 9:39pm
Originally posted by soneill soneill wrote:

USATT attempts a head-to-head exchange, RC looks at the entire field's matches and gives individual points to reflect all the matches in the tournament.


Ratings Central looks at your opponents and your opponents' opponents. It doesn't look any further than that. Looking further would be computationally too difficult, and there isn't much information there.
Back to Top
David Marcus View Drop Down
Super Member
Super Member
Avatar

Joined: 10/28/2010
Location: Somerville, MA
Status: Offline
Points: 145
Post Options Post Options   Thanks (0) Thanks(0)   Quote David Marcus Quote  Post ReplyReply Direct Link To This Post Posted: 01/20/2023 at 9:46pm
Originally posted by stiltt stiltt wrote:

Is there a number at which a newbie's rating be at first no matter their level of play? 1500 is an example I get from FICS in chess, I did not play in ages on ICC but I remember it's about the same. They use the Glicko system like the enhanced version on Rating Central. Does RC do the same? Where do we start? 

I have a question about the Reliability Deviation factor that will correct the newbies quickly up or down while their very new appearance will not affect established players as much.

We need communication about the concept of Reliability Deviation, often mentioned as Rating Deviation. If we all learn about it, we will understand better the value of Rating Central v. other rating systems in table tennis.

Richard Feynman used to say that if we can't explain a complex idea with simple words, it means we didn't get it. How do we explain in simple words that the RD way and Rating Central can solve the issue that mjamja is raising?


In Ratings Central, the initial law for a player depends on the population that the event draws from.

Ratings Central calculates the standard deviation of each player's law. This measures how well the system knows the player's playing strength.

The reason to model a player's playing strength as a probability distribution (aka law) is so that the system can keep track of how well it knows the player's playing strength. If all you have for a player is a number, then you've forgotten how much data you used to get that number. That's like saying that you know the level of a player who just walked into your club just as well as a player who has been playing at your club for years.
Back to Top
 Post Reply Post Reply Page  <12
  Share Topic   

Forum Jump Forum Permissions View Drop Down

Forum Software by Web Wiz Forums® version 12.01
Copyright ©2001-2018 Web Wiz Ltd.

This page was generated in 0.125 seconds.

Become a Fan on Facebook Follow us on Twitter Web Wiz News
Forum Home | Go to the Forums | Forum Help | Disclaimer

MyTableTennis.NET is the trading name of Alex Table Tennis Ltd.

Copyright ©2003-2020 Alex Table Tennis Ltd. All rights reserved.