FIDE rating calculations
Posted: Mon Jun 13, 2011 11:01 pm
I have a set of questions about the calculation of FIDE ratings. The questions are about the sequence in which games are rated and how different tournaments and leagues decide which grading list to use when rating their events. The post is a bit long, but there is a summary at the end.
The questions arose when I realised that the FIDE rated games I've been playing this year have been rated out of sequence to the order in which they were played, and some of the grading changes due to play in league chess (4NCL) have been delayed by some considerable period of time, which doesn't make too much sense (IMO), when FIDE issue rating lists every 2 months. But maybe this is an unavoidable consequence of the system.
This seems to have only happened recently because I've been playing in FIDE rated events that take place within the period in which the 4NCL takes place. So the games from those tournaments are rated in an earlier list to the ones in which the 4NCL ones get rated (at the end of the season). But I can't figure out how the system can cope with this.
Anyway, games I've played have been rated in four of the last 5 rating periods:
http://ratings.fide.com/hist.phtml?event=470589
Five games from the September 2010 Sunningdale Major were rated using the 1910 rating I had then. New rating of 1933 in the November 2010 rating list. That's easy enough to understand. Similarly for the January 2011 list, where a game from the November 2010 Brighton Major got rated, giving rise to the new rating of 1937. Nothing in the March 2010 list. Then 3 games from the April 2011 Surrey Major, bringing the rating down to 1929 in the May 2011 list.
That's all fine so far, but then the four 4NCL games I played on the weekends of 15/16 January 2011 and 26/27 March 2011 appear in the rating summary for the July 2011 list:
http://ratings.fide.com/individual_calc ... 2011-07-01
From what I can tell, those games have been rated using my rating from the November 2010 list (1933), but the ratings given on the 4NCL website for the matches in question appear to be the rating from the list in force at the time the games were played (1937, from the January 2011 and March 2011 lists) - two examples linked below:
http://www.4ncl.co.uk/resdiv2_r3a.htm#C ... iversity_2
http://www.4ncl.co.uk/resdiv2_r7a.htm#C ... iversity_2
1) My first question is what is happening here? Are the 4NCL games rated using the grading from the list published at the start of the season? And is the rating on the 4NCL website just for "board order" purposes? (The ratings of my opponents can be different on the FIDE rating page when compared to the rating published on the 4NCL website - e.g. Jameson is 2088 on the 4NCL page and 2054 on the FIDE page).
2) My second question is how can a change based on a previous rating (1933) be applied to the current rating (1929) when that rating is different? I suspect the answer is something fundamental to do with how the FIDE rating system calculations work, as I had no idea that the system could cope with that sort of retrospective rating kind of thing. How does this work?
Also in the rating summary for July 2011 is the rating change from the May 2011 Sunningdale Major, which is based on the 1929 rating from the list current at the start of that event (i.e. the May 2011 list). The next set of rating calculations should be from an event from June 2011 (held in Liverpool), so if that gets onto the July list it will appear on that page I've linked to.
3) That leads to my third question. If an event you are playing in is using ratings in their published documentation from a previous grading list (March 2011 - this is quite common if an event receives entries over the course of several grading lists and they don't update their records), is it possible that the rating will be done using that earlier rating or should it be done with the one from the list in force at the time of the event (which would be the May 2011 list in this case)?
4) The fourth (and final) question is about chronological order. When games are submitted at the end of a league season (as for the 4NCL above), it seems the system is blind to the chronological position of these games compared to others from shorter events (the system doesn't actually know when they were played, and for the rating purposes it seems they are counted as being played at the time of the list from which the rating used for the calculations is taken). But can this introduce errors into the system? To take the examples from my FIDE rating history, the games from the Surrey Major (April 2011) were played after the games from the 4NCL (January 2011 and March 2011), but were rated before those games were rated. This seems a little bit confusing.
Summary: 4NCL games rated with rating from November 2010 list (1933), played during the period when the January 2011 and March 2011 lists were in force (rating of 1937), but rated for the July 2011 list (contributing to a provisional change from 1929 to 1924). So my actual rating of 1937 at the time of the 4NCL games doesn't seem to have been used at all in the calculations. Is that an accurate summary of how rating for league seasons works in an era of rating lists published every 2 months?
The questions arose when I realised that the FIDE rated games I've been playing this year have been rated out of sequence to the order in which they were played, and some of the grading changes due to play in league chess (4NCL) have been delayed by some considerable period of time, which doesn't make too much sense (IMO), when FIDE issue rating lists every 2 months. But maybe this is an unavoidable consequence of the system.
This seems to have only happened recently because I've been playing in FIDE rated events that take place within the period in which the 4NCL takes place. So the games from those tournaments are rated in an earlier list to the ones in which the 4NCL ones get rated (at the end of the season). But I can't figure out how the system can cope with this.
Anyway, games I've played have been rated in four of the last 5 rating periods:
http://ratings.fide.com/hist.phtml?event=470589
Five games from the September 2010 Sunningdale Major were rated using the 1910 rating I had then. New rating of 1933 in the November 2010 rating list. That's easy enough to understand. Similarly for the January 2011 list, where a game from the November 2010 Brighton Major got rated, giving rise to the new rating of 1937. Nothing in the March 2010 list. Then 3 games from the April 2011 Surrey Major, bringing the rating down to 1929 in the May 2011 list.
That's all fine so far, but then the four 4NCL games I played on the weekends of 15/16 January 2011 and 26/27 March 2011 appear in the rating summary for the July 2011 list:
http://ratings.fide.com/individual_calc ... 2011-07-01
From what I can tell, those games have been rated using my rating from the November 2010 list (1933), but the ratings given on the 4NCL website for the matches in question appear to be the rating from the list in force at the time the games were played (1937, from the January 2011 and March 2011 lists) - two examples linked below:
http://www.4ncl.co.uk/resdiv2_r3a.htm#C ... iversity_2
http://www.4ncl.co.uk/resdiv2_r7a.htm#C ... iversity_2
1) My first question is what is happening here? Are the 4NCL games rated using the grading from the list published at the start of the season? And is the rating on the 4NCL website just for "board order" purposes? (The ratings of my opponents can be different on the FIDE rating page when compared to the rating published on the 4NCL website - e.g. Jameson is 2088 on the 4NCL page and 2054 on the FIDE page).
2) My second question is how can a change based on a previous rating (1933) be applied to the current rating (1929) when that rating is different? I suspect the answer is something fundamental to do with how the FIDE rating system calculations work, as I had no idea that the system could cope with that sort of retrospective rating kind of thing. How does this work?
Also in the rating summary for July 2011 is the rating change from the May 2011 Sunningdale Major, which is based on the 1929 rating from the list current at the start of that event (i.e. the May 2011 list). The next set of rating calculations should be from an event from June 2011 (held in Liverpool), so if that gets onto the July list it will appear on that page I've linked to.
3) That leads to my third question. If an event you are playing in is using ratings in their published documentation from a previous grading list (March 2011 - this is quite common if an event receives entries over the course of several grading lists and they don't update their records), is it possible that the rating will be done using that earlier rating or should it be done with the one from the list in force at the time of the event (which would be the May 2011 list in this case)?
4) The fourth (and final) question is about chronological order. When games are submitted at the end of a league season (as for the 4NCL above), it seems the system is blind to the chronological position of these games compared to others from shorter events (the system doesn't actually know when they were played, and for the rating purposes it seems they are counted as being played at the time of the list from which the rating used for the calculations is taken). But can this introduce errors into the system? To take the examples from my FIDE rating history, the games from the Surrey Major (April 2011) were played after the games from the 4NCL (January 2011 and March 2011), but were rated before those games were rated. This seems a little bit confusing.
Summary: 4NCL games rated with rating from November 2010 list (1933), played during the period when the January 2011 and March 2011 lists were in force (rating of 1937), but rated for the July 2011 list (contributing to a provisional change from 1929 to 1924). So my actual rating of 1937 at the time of the 4NCL games doesn't seem to have been used at all in the calculations. Is that an accurate summary of how rating for league seasons works in an era of rating lists published every 2 months?