The Ratings of Chessplayers

 

Chess rating systems have been used for pairing purposes in chess tournaments as well as estimating playing strength.  A tournament director wants to avoid pairing the strongest players against each other in the early rounds of a tournament.  Ratings are also used for tournament sectioning and prize eligibility.  Sections or Classes in tournaments are popular and many players play other players with close to the same rating.

The introduction of chess rating systems may have done more to popularize tournament chess than any other single factor. 

In 1928, the first work to give serious attention to modeling chess ability was a paper by Ernest Zermelo.  He addressed the problem of estimating the strengths of chess players in chess tournaments.

In 1939, the first modern numerical chess rating system was used by the Correspondence Chess League of America (CCLA).  The CCLA was founded in 1909.  In 1940, it was using a point system designed by Kenneth Williams.

In 1942, Chess Review magazine began using the Harkness rating system, designed by Kenneth Harkness (1896-1972).  The ratings were first used for correspondence chess players.  At the time, there were only three classes: Class A – above 1050; Class B – 950 to 1050, and Class C – below 950.

In 1946, a rating system was proposed in the USSR by Andrey Khachatoruv (1917- ) and was published in Shakmaty v SSSR in 1946.  The USSR Chess Federation used a non-numerical system to classify players.

In 1948, the Ingo rating system was created.  It was designed by Anton Hoesslinger (1875-1959) and named after his home town, Ingolstadt in Bavaria.  It was used by the West German Chess Federation from 1948 until 1992.  It was then replaced by the Elo system, named after its creator Dr. Arpad Elo (1903-1992), a Hungarian-born American physics teacher and chess master.  This was a system where a player’s new rating was the average rating of his competition minus one point for each percentage point above 50 obtained in the tournament.

In 1949, Kenneth Harkness submitted his rating system to the United States Chess Federation (USCF) for use as a numerical rating system.  When a chess player completes a tournament, the average rating of his competition is first calculated.  Then, if a player scores 50%, he receives the average competition rating as his performance rating.  If he scores more than 50%, his new rating is his competition rating plus 10 points for each percentage point above 50.  If he scores less than 50%, his new rating is his competition rating minus 10 points for each percentage point below 50.

In January 1950, the USCF announce that a national rating system was being planned to cover local and national events.

In July 1950, the USCF adopted the Harkness system and invented by Kenneth Harkness.  It was used by the USCF from 1950 to 1960.   The first published rating list appeared in the November 20, 1950 issue of Chess Life magazine, which rated 2,306 USCF members from chess events from 1921 to July 31, 1950.  The highest rated players on the first USCF rating list were Reuben Fine (2817), Samuel Reshevsky (2770), A. Kevitz (2610), Arthur Dake (2598), Arnold Denker (2575), Isaac Kashdan (2574), I.A. Horowitz (2558), and Larry Evans (2484).  In 1950, the USCF classified a Grandmaster as anyone rated 2700 or higher.  A Senior Master was rated between 2500 and 2699.  A Master was rated between 2300 and 2499.  An Expert was rated between 2100 and 2299.  Class A was 1900 to 2099.  Class B was 1700 to 1899.  Class C was 1500 to 1699.  Class D was anyone below 1500.

In 1954, the first British Grading List was published, which listed 49 players.

In 1955, Irving J. Good (1916-2009) developed a system to rank chess players and published “On the Marking of Chess Players” in the Mathematical Gazette.

In 1958, the British Chess Federation (now the English Chess Federation or ECF) started using a rating system devised by Richard W.B. Clarke (1910-1975).  A players rating or grade is calculated by taking the opponent’s grade and adding 50 points for a win, subtracting 50 points for a loss, and taking the opponent’s current grade for a draw.   To convert the ECF rating to the USCF rating, USCF=ECF*8 + 600.  To convert the ECF rating to the FIDE rating, FICE=ECF*8 + 650.

In 1959, the USCF named Dr. Arpad Elo to head a committee (which also included Dr. Eric Marchand, Guthrie McClain and James Warren) to examine and review all rating system and make recommendations.  He devised a new system with a more statistical basis.  The flaw with the old Harkness system was that a player could lose every game in a tournament and still gain rating points.  He could also win every game against lower-rated players and end up with a very high, but unrealistic rating.  Elo was the Chairman of the Rating Committee of the USCF from 1959 to 1976.

In 1961, the USCF switched to the Elo rating system, invented by professor Arpad Elo of Milwaukee, Michigan.  The new Elo rating system was published in the 1961 June issue of Chess Life.  The Elo rating system assigned to every player a numerical rating based on performances in competitive chess.  Elo assumed that the chess player’s strength distribution was a normal distribution (bell curve).

The USCF has rating categories.  A senior master is rated 2400 and up.  A national master is rated between 2200 and 2399. An expert (or candidate master) is rated from 2000 to 2199.  Every 200 point range below expert is a class.  For example, a rating of 1900-1999 is Class A.  A rating of 1600 to 1799 is Class B,  A rating of 1400 to 1599 is Class C, etc.

The average rating for established USCF chess players is 1500.  About 96% of all USCF players have ratings less than 2200 (master).  For FIDE, about 23% of established FIDE players are less than 2200.

In 1970, the World Chess Federation (FIDE) adopted the Elo rating system.  The rating system was also the basis for the award of international titles.

FIDE also classified tournaments into categories according to the average rating of the players.  Each category is 25 points wide.  Category 1 is the lowest category, rated from 2251 to 2275, category 2 is 2276 to 2300, etc.  A category 22 tournament, the highest so far, would be in the 2776-2800 range.

The first FIDE list was headed by Bobby Fischer (2720), followed by Boris Spassky (2690) and Viktor Korchnoi (2680).

On July 1, 1971, the first official FIDE rating list was published.  There were 83 grandmasters on the first official FIDE rating list.

When FIDE started using Elo ratings, the FIDE rating was rounded out to the nearest multiple of 5.  FIDE also only computed ratings for players over 2000.  A player only acquires a rating if it is calculated to be over 2000.  On average, initial FIDE ratings overestimate players’ abilities because players only receive ratings if their initial performances are strong.  Thus, the FIDE rating pool gets inflated over time.

In 1978, Elo published The Rating of Chessplayers Past & Present.  A second edition appeared in 1986.

In the 1990s, the USCF rating system added a rating for quick chess (USCF Quick).  Quick chess refers to games where the time control for a game is quicker than 30 minutes per person for the entire game.  A second rating system that parallels the USCF rating system for slower games was created for quick chess.

In the mid-1970s, the average rating of USCF players was declining.  There was a bigger discrepancy in USCF and FIDE ratings.  So the USCF introduced bonus points and feedback points.  When a player did very well in a tournament, his rating was increased by the addition of bonus points.  The justification for awarding bonus points was that the player was improving fast and the ordinary updating rating formulas did not track the player’s improvements quickly enough.  By the mid-1980s, these features were eliminated from the rating system.

In the late 1980s, the concept of a rating floor was established in the USCF system.  This prevented a player’s rating from decreasing below the 100-point multiple 200 points less than one’s highest rating.  More recently, the rating floor has been re-implemented so that the system now uses a 100 point margin.  Rating floors were to discourage players from purposely losing games to artificially lower their ratings so that they could play in lower-rated sections.

In 2001, the Glicko rating system was invented by Professor Mark Glickman, a mathematician at Boston University.  It is similar to the Elo rating system, but adds a “ratings reliability,” called RD, for ratings deviation.  A higher Glicko rating implies moiré skill playing chess.  This rating system is used on many chess game servers such as the Free Internet Server (FICS), chess.com, and others.  It has also been used by the Australian Chess Federation.

In 2004, another rating system, the Edo Historical Chess Ratings system, was created by Rod Edwards.  This was an attempt to retroactively rate chess players over time.  It uses an iterative method (Bradley-Terry).    It top peak Edo ratings were: Steinitz (2803),  Morphy (2796), Lasker (2752), Kolisch (2710), Tarrasch (2699), Zukertort (2678), von der Lasa (2676), Anderssen (2673), Neumann (2671), and maroczy (2665).

In 2005, the Chessmetrics rating system was created by statistician Jeff Sonas.  It is based on computer analysis of chess games that uses a weighted average of past performance and is intended to be more accurate than the Elo system.  The score considers a player’s win percentage against other players weighted by the ratings of the other players and the time elapsed since the match.  For one-year peaks, the top rated players were Bobby Fischer (2881), Garry Kasparov (2879), Mikhail Botvinnik (2871), Jose Capablanca (2863), Alxander Alekhine (2851), anatoly Karpov (2842), Vishy anand (2828), Vladimir Kramnik (2822), and Siegbert Tarrasch (2818).

In 2006. Matej Guid and Ivan Bratko published a research paper called "Computer Analysis of World Chess Champions," which rated world chess champions by comparing their moves to the computer program Crafty.

Since July 2009, FIDE has issued a rating list every two months.  Prior to that, a rating list appeared every 6 months.  There is also an unofficial “live ratings” which calculate the change in players’ ratings after every game.  These Live ratings are based on the previously published FIDE ratings, so a player’s Live rating corresponds to what FIDE would rate a person if they were to issues a new rating list that day.

In 2017, Jean-Marc Alloit published a research paper called "Who is the Master?", which rated chess champions by comparing their moves to Stockfish.

As of November 2020, the top 10 rated players are Carlsen (2862), Caruana (2823), Ding Liren (2791), Nepomniachtchi (2784), Vachier-Lagrave (2784), Aronia (2781), Grischuk (2777), Mamedyarov (2770), So (2770), and Rajjabov (2765).

The top woman is Yifan Hou (2658).  The top junior is Alireza Firouzja (2749).

Performance rating is a hypothetical rating that would result from the games of a single event only.  A performance rating for an event is calculated by taking the rating of each player that you defeated and adding 400 points, then taking the rating of each player that you lost to and subtract 400 points.  If you draw, you take your opponents rating.  You then sum these figures and divide by the number of games played.

Performance rating = [(Total of opponents' ratings + 400 * (Wins - Losses)) / Games].

 

To convert ratings from other systems to the USCF rating, the following adjustments are believed to be roughly appropriate.

 

For FIDE rated players, USCF = FIDE + 50

For English ratings, USCF = ECF*8 + 700

For Germany (Ingo), USCF = 2940 - (Ingo*8)

For Russia, USCF = Russian rating + 250

 

The USCF has a Player/Rating Lookup list located at

http://main.uschess.org/component/option,com_wrapper/Itemid,181/

 

FIDE ratings can be found at

http://ratings.fide.com/