30 March 2008

Titled Players, Sizing the Data

Continuing with Titled Players, Soviet and Otherwise, my first task was to come to grips with the historical rating data. The data that I'm using spans 38 years and has gone through many hands, including my own. There are inconsistencies from year to year that can cause problems with any comparison across time.

The following table compares data from four different sources.

  • 1974: An early rating list that included titles.
  • 1978: Chapter 9.4 of Elo's 'The Rating of Chess Players Past and Present' (RCPP).
  • 1999: The first of the more recent FIDE lists that included titles.
  • 2008: The January 2008 list published by FIDE.

Most of the table's columns are self-explanatory. For example, the first line says that in 1974, there were 91 GMs with an average rating of 2535.

The last column 'Src' (Source) is an attempt to document some of the data's quirks. Elo's RCPP listed two ratings for the titled players: the FIDE rating on January 1978 and a 'Best 5-year Average' rating calculated by Elo. All but four of the players on his list had one rating or the other, and many players had both.

The values for 1999-M and 1999-W document a small problem in the 1999 data. Players were listed with two flags: one indicating whether they were titled and another indicating whether they were women. Where a woman was titled, the value in the title field was her women's title (WGM etc.). Where the woman also had an unrestricted title, it was flagged against her name. For example, the three Polgar sisters were listed like this:

g; Polgar, Judit (GM) ; HUN; 2677; w
g; Polgar, Sofia (IM) ; HUN; 2505; w
g; Polgar, Zsuzsa (GM); HUN; 2565; w

For no particular reason, I list men and women separately. Finally, in 2000 or earlier FIDE started flagging certain players as 'Inactive'. This shows in the data for 2008.

The bottom line is 91 GMs in 1974, 1109 in 2008. Somewhere between those two years, the GM club lost much of its exclusivity.

No comments: