Tournament Rules - input
See separate file with track changes in comparison to current TR2015. Key:
insertion - red on yellow,
black (strikethrough) on blue,
input for 2017 rules - red on green.
ICCF 2015-019: Revised Norm Calculations
Austin Lockwood and Mariusz Wojnar
Revised Norm Calculations
Phase One, Method Development and Documentation (August 2015 to Spring 2016)
Appendix II of the ICCF tournament rules will be rewritten, based on solid statistical principles and in consultation with a professional statistician. This document will be presented to ICCF Congress in 2016 for ratification.
Phase Two, Implementation (Summer 2016 to 31.12.2016)
If the proposed revisions to the ICCF title regulations are accepted by Congress in 2016, the ICCF webserver will be updated, and the new regulations will apply to tournaments starting from 1.1.2017 onwards.
Historical Need for Simplification
The current system of assigning categories to title tournaments, based on average rating and values looked up in tables was essential in previous years. Without the availability of powerful computing resources, and the need for tournament directors to perform these calculations unassisted and without error, a simplified system to estimated tournament performance, however crudely, was absolutely essential.
However the lack of computational power is no longer an issue, all ICCF tournaments (including postal events) are now managed on the ICCF server; this means that complex calculations can be performed reliably, consistently and without error.
ICCF can now adopt a far more sophisticated system for measuring tournament performance based on solid statistical principles, therefore making the award of title norms fully commensurate with performance in title tournaments.
Problem with Current Norm Calculations
Currently, title tournaments are assigned a ‘Category’ based on the average rating of all players in the tournament. Tournament Performance, and required scores, is looked up in a table. This means that tournament performance for an individual player is based on the average rating of all players in the tournament, not just the player’s opponents, furthermore, performance is rounded down to a fixed point defined by the categories.
This is a rather crude way to estimate tournament performance.
(We propose to keep category definitions as a way to describe the average tournament rating, but simply for future marketing purposes, not to determine individual norm requirements.)
Errors in Norm Table
In 2014 Mariusz Wojnar attempted to reproduce the norm table, but found that although he could reproduce most of the values in the table, there appeared to be several errors. This was highlighted in the 2014 QC’s report to Congress.
Congress agreed that a corrected version of the norm table would apply to title tournaments starting from 1.1.2015 following verification.
Unfortunately further problems were subsequently discovered with the methodology. As the original calculations were not documented, it became clear that to reproduce the tables and correct any errors would be an impossible task.
The Executive Board, therefore, decided to continue with the previous norm calculations from 1.1.2015 rather than confuse the issue any further.
Obscure Nature of Lookup Tables
As well as the lack of documentation for the norm tables, there is also no available formula for reproducing the probability tables which underpin the rating system. This means that our calculations are obscure, and do not have a well-defined methodology.
Professor Mark Glickman
Mark Glickman is Research Professor of Health Policy and Management at the Boston University School of Public Health, and Senior Statistician at the Center for Healthcare Organization and Implementation Research. He is also visiting professor at the Harvard University Statistics Department where he has been teaching both introductory and advanced statistics courses for many years. He was elected Fellow of the American Statistical Association in 2014.
In order to develop a robust system for measuring tournament performance, it will be necessary to consult a professional statistician. Professor Mark Glickman is a world renowned expert in the field of chess ratings having developed the Glicko and Glicko-2 rating systems that are in widespread use, and is currently Chair of the United States Chess Federation Ratings Committee; Professor Glickman has estimated the following costs:
Phase One Costs (2015/16)
Method development: USD$1,200 - $2,000
Write-up of statistical details that would appear in the new Appendix II: USD$800 - $1,200
Correspondence and e-mail with ICCF (e.g., deciding on fundamental assumptions underlying revised title system): USD$1,200 - $2,000
Total for Phase One: USD$3,200 - $5,200 (approx. EUR€3,017 - €4,902)
Phase Two Costs (2016)
Write-up of any technical documentation that would be implemented by a programmer: USD$1,600 - $2,400
Correspondence with a programmer about implementation issues: USD$1,200 - $2,000
Total for Phase Two: USD$1,800 - $4,400 (approx. EUR€1,697 - €4,148)
In addition to the cost of consulting the statistical expert, there will be programming and implementation costs for Phase Two, these costs will be estimated following the specification developed in Phase One and presented in the Services Director’s report to Congress in 2016.
Consultation with Nol van 't Riet
Whilst researching this proposal we were in contact with one of the fathers of the current ICCF rating and title systems, Nol van 't Riet. We are very grateful to Nol for his support and his patience in explaining the background to the current system, and also for his positive comments on this proposal.
Nol explained to us that in the early days of the title system, Norm tables were absolutely necessary because powerful computers were not available in the early days, however he believes that a more accurate approach is now essential. In Nol's words:
Of course we must follow FIDE. Working with Performance Ratings is not only much better, it’s the best way to do. FIDE needed that way of thinking and approaching when the Swiss tournaments were rising all over the world. And imagine: round robin tournaments are only a particular kind of Swiss tournaments.
On the time that we only had round robins and no computers or calculators Elo made it easy for organizers to calculate title norms by giving them a (rough) table.
Nowadays we have computers and calculators so we can calculate much more exactly the norms which are required for every single player in any kind of tournament, depending on the strength of his opponents. And now that we can do that, we also should do it!
So my advice would be to immediately take over the FIDE title approach. And forget about the table.
That’s also the transparent solution which you are looking for!
Nol passes on his good wishes to all his ICCF friends at Congress.
ICCF Tournament Rules Appendix II will be completely revised.
A vote of Yes will mean that a robust system for calculating norms in title tournaments will be ratified by the 2016 ICCF Congress, and if accepted, will apply to all title tournaments starting from 1.1.2017.
A vote of No will mean ICCF will continue to use a crude and obscure approximation to measuring tournament performance to assign norms.
A vote of ABSTAIN is not a vote but means the vote holder has no opinion and does not wish to represent the correspondence chess players of his or her federation in this matter.