This page gives background information on the Glicko DBM ratings system as applied in Ireland. To see the rules regarding application of the system, see the RINIWA Ratings Policy.

Explanation of the RINIWA DBM Glicko Ratings, at 26th Aug 2002

0. A quick explaination

Glicko ratings assume that the average score per game is a good estimate of player ability. Thus, Glicko ratings reflect a player's average score in every competition game of DBM recorded. If players have on average played against "better" (ie higher-rated) players then their own rating will be higher than their average score suggests. If they have on average played against "weaker" (ie lower-rated) players then their own rating will be lower than their average score suggests. Finally when calculating the rating, more recent results are more important than older results. This is the basis of a player's rating level or just "rating". All players start at a rating of 2000, so a player with that rating is considered to be of average playing ability.

It is important that enough data is collected to make a player's average score meaningful. For example if a player only plays one game and wins 10-0 then their average score is 10, if we weren't careful we would say that they are an excellent player as their average is so high. In reality we don't know as one win doesn't necessarily mean this. Hence we say that this player's rating has a low reliablity. For every player we calculate a reliability measure as well as the rating to make sure that we have enough data to decide that a player is (probably) strong or (probably) weak. To keep things simple the reliabilty measure has been split into five different categories from A to E, with A being the most reliable and E being almost totally unreliable.

That's the basic explaination of what's going on in the rating calculations. See the sections below for more detail.

1. Ratings system concepts

The chief design goal of the rating system is to provide an objective measure of the relative DBM playing strength in a competitive environment of each rated player. The ratings should be as fair as possible to all players regardless of their ability to attend competitions or level of playing success.

This system of calculating RINIWA DBM ratings is based on the Glicko system used in internet chess which is itself a statistically sophisticated refinement of the Elo system used in international tabletop chess.

Under this system two values are calculated for each player.

The most important one is the rating level of the player. This is a numeric value representing the estimated relative playing strength of each player. The higher the value, the stronger the player. If two players are equally rated, then based on the information to date, there is expected to be a 50% chance of either player ultimately defeating the other in a competitive game of DBM. The greater the difference in rating levels, the greater the expectation that the higher rated player will win a competitive game of DBM. In statistical terms, the rating level is a central or unbiased estimate of each player's DBM playing strength, as likely to be an overestimate as an underestimate of their "true" or long term playing ability.

The second value calculated is a reliability measure. This is an indicator of how accurate an estimate the rating level is likely to be, or in other words, how much trust can be placed in it. It is similar to a standard deviation. As a player plays more games in rated competitions the ratings level becomes more reliable however the reliability measure is subject to decay with time. That is, the longer a player goes without competing in a rated competition, the less reliable his estimated rating level becomes.

In the reported ratings, all rating levels have been rounded to the nearest multiple of 5 (any greater accuracy would be spurious) and the reliability of each rating has been categorised as one of A, B, C, D or E. Those rating with reliability of A are the most reliable and stable while those with a reliability of E are little more than early indicators and could be subject to large fluctuations as more results are received. Typically a player who has played 20 rated games over the last 2 years would be in category C and one who has played more than 30 would be in category B.

Each player's rating level is calculated based on results from each valid game played in rated public competitions. The previous rating level is adjusted depending upon whether the player has performed better or worse than expected based on their past results and the quality of the opponents they faced in each competition.

A player does not have to win every game to improve his rating. A mix of wins and losses may improve the rating of an average or weaker player. Similarly a stronger player does not necessarily have to win every game to maintain their high rating although they would normally be expected to win a higher proportion than a weaker player.

A player with a long established track record will have a stable rating level. One bad tournament will not cause that rating to plunge drastically. Similarly one good tournament will not cause it to soar to exceptional heights.

Byes and forfeits are ignored. The DBM 10 point scoring system is always used for rating purposes although competition organisers are free to use whatever system they like to determine their placings and prizes. The calculations do not use competition placings, just the results of individual games. The rating system should not interfere with or change the way any competition is run.

2. Published rating information

The published information will consist of 3 lists ("Full Irish Ratings", "Top Rated Players" and "Player of the Year League") and supporting statistics ("More Statistics"). The "Full Irish Ratings" is an alphabetical list of all rated players with their current rating, change in rating since last publication and score in the last competition they attended. The "Top Rated Players" list is where you will find the current RINIWA number 1, 2 etc.. It is hard to even qualify for this list as you must have a minimum reliability of category C, and be in the top half of RINIWA players with category C reliability. This means that players who perform magnificently in their first rated competition will not be included unless they also perform well in further competitions. The "Player of the Year League" is purely based on results over a calander year ending each 1st of January and unlike the main career ratings, does not penalise poor performances (other than that they represent a missed opportunity). There is also no reliability restrictions on being published here so a greater range of players are present. Only above average performances are rewarded, that is those that have an REP (Ratings Equivilent Performance) greater than 2,000.

Only players who have taken part in a rated competition in the last 25 months will have their rating information published.

Any player who does not wish to have their DBM rating information published may write to me at the e-mail address below.

3. Opting out

Players may elect to opt out of ratings for a particular rated competition. The usual reason for doing so would be to allow them to take a "fun" or "patheticon" army without endangering their ratings. Players are limited in the number of competitions which they may opt out.

To opt out, a player must contact me (see e-mail address below) prior to the commencement of the competition. I will acknowledge this as soon as possible. It is the player's responsibility that an election to opt out is successfully sent and received. Alternatively a sheet will be made available prior to the tournament commencing by the umpire for players to sign to say they will opt-out. It is the responsibility of all umpires of rated competitions to have such a sheet available. The player must sign this prior to the tournament commencing.

4. New players

New players are rated after their first rated competition. In Ireland, new players are given a initial rating of 2,000 and a reliability measure that indicates that this has virtually no credence. At the end of their first rated competition, their rating level will be published based on their results in that competition and will normally still be in reliability category E (most unreliable).

The early ratings of new players are very unreliable and thus can be subject to big fluctuations over their first few competitions. For this reason new players should not be at all discouraged if their first published ratings are quite low if the player takes a little while to adjust to tournament play. Some reasonable results later should soon see their ratings return to more acceptable levels.

5. Further information

Anybody who wants further information on this topic or has comments or suggestions, should feel free to contact the Ratings Administrator, Rob Brennan, by e-mail at brennanr@iworg.com.  If you want to read RINIWA's rules and regulations on the Administration of the ratings, see the Policy section of the RINIWA website. The Ratings Administrator is appointed by the RINIWA Committe and the RINIWA Committee has ultimate control of the Ratings. If there is something that you cannot resolve directly with the Ratings Administrator, please contact a Committee member.