The Ratings Committee (RC) this year was faced with a host of
small but important tasks.  In previous years, much of the RC
effort involved addressing major issues involving the new rating 
system, which has completed its fifth year of existence.  The tasks
this year have been in the spirit of refining and tackling
minor aspects of the rating system.

Based on some correspondence with scholastic tournament organizers,
we realized that the age-based initial rating formula had 
become inappropriate for the current group of scholastic
players entering the rating pool.  Based on some helpful
summaries put together by Mike Nolan, we determined a
revision of the age-based initial rating formula.  We
have recommended to the office the following substitutions
to the formulas:  If a player's age is less than 26 years,
then the imputed initial rating should be 50*Age ("Age"
measured in years).  If the player's age is 26 years or
older, the initial rating should be 1300.  Again, this change 
reflects the lower average ratings of young players entering the 
USCF pool.

Two changes to the rating system were made and approved in
early March, 2006.  The first was raising the "money" floors
from $1000 to $2000.  In the past, if a player won $1000 or
more as a class prize, the player's rating floor became
the rating corresponding to the tournament section upper rating 
limit.  Because it is now more common to win $1000 compared to when 
the rule was implemented years ago, the prize threshold was
increased to $2000.  The second change was the removal of the
half-K option for standard tournaments.  As communicated
to the RC by the office, very few tournament directors have been 
exercising the option to run their events as half-K, and the RC 
felt that there was no theoretically sound basis for rating
these events as half-K anyway.

Some members of the RC were involved in extensive
correspondence about the rating of matches between players
where one of them was at or near his rating floor.  The
concern was that the opponent of the player at his floor had
an opportunity to take advantage of the situation, 
possibly gaining a large number of rating points, whereas the
floored player had no risk of losing points.  The
suggestion put forth by Bill Goichberg, which appears to 
be the proposal on the table, is to rate matches
in such a way so that players at or near their rating floor
would have their floor dropped by 100 points for the match.
This would allow the floored player to have a post-match
rating that would be below his current floor.  Along with
this proposal, Bill suggested that 2-round events with 2
players be considered a match for rating purposes.

Last Fall, Mike Nolan provided the RC with ratings data on
players who had both FIDE and established USCF ratings, and
who were active tournament players in the past three years.
From this data, we determined a revised conversion formula
from the FIDE rating scale to the USCF scale.  The new
conversion is as follows:  For FIDE rating below 2200, the converted 
USCF rating is the same.  For FIDE rating between 2200 and 2600, 
the converted USCF rating is FIDE + (FIDE-2200)/8.  For FIDE ratings
above 2600, the converted USCF rating is FIDE+50.  We have
recommended to the USCF office to begin using these formulas
to impute initial USCF ratings for FIDE players who have not
yet entered USCF events.  The RC notes that, based on our
data analyses, the conversion is quite imprecise on an
individual basis.  Even though the formula prescribes that 
a 2200 FIDE player has a converted USCF rating of 2200, 
in reality the range of possible USCF ratings 
for FIDE=2200 can be as low as 1800 and as high as 2450.
This is the reason that the conversion formula needs to be 
applied with caution, recognizing the imprecision in the
correspondence between the FIDE and USCF scales.

Related to the issue of the FIDE-to-USCF conversion, there
has been extensive discussion on how to make use of
the results of FIDE events in updating USCF ratings
for USCF-rated players that compete in such events.
The RC last year worked out a set of formulas to convert tournament 
summary information that FIDE provides in order to update USCF
ratings - this method, which can be applied to all FIDE
events and is not terribly labor-intensive, requires 
that rating changes are updated using half-K due to the
simplifying approximations in the formulas, and due to the
imprecision of the rating conversion between FIDE and the
USCF scales.  Bill Goichberg presented an alternative option
where a small number of significant FIDE events would be rated
using the USCF programs, after converting all of the FIDE ratings 
to the USCF scale, and this method would use 0.8*K instead of
half-K.  The problem with this method is that it could
potentially require significant resources to carry out on an
ongoing basis.  The decision of what method to use is still
under consideration by the USCF office and the executive board.

The RC was asked to consider whether it was sensible to establish
USCF rating floors for titled FIDE players who had not competed in
USCF events, fearing that titled FIDE players might sandbag their 
performances in USCF events.  The particular method we were asked to 
consider involved converting the player's FIDE rating to the USCF scale, 
and then using the appropriate floor corresponding to the
converted rating.  Due to the imprecision of the FIDE conversion,
among other reasons, the RC instead proposed that titled FIDE
players have their initial converted rating based on 25 games
(instead of 10, as the current rating system dictates), which
would ensure that the FIDE player would have an established
USCF rating after his first event, thus creating a rating
floor based on the post-event rating.  This proposal is
currently under consideration.

One of the tasks the RC has not yet completed is a method to
connect the Quick Chess (QC) and regular rating systems.  A concern 
was expressed that many players' QC and regular ratings were
too far apart to believe that both were accurate representations 
of playing strength in the different systems.  This could
easily come about if a player competes actively in one system
(e.g., in tournaments with regular time controls) and not in the
other, so that the latter rating would be out of date.  This is a
tricky issue because it is not immediately clear within the
current rating system how to recognize the positive correlation
between QC and regular ratings, but also recognizing that
the correlation may not apply for every player.  We
continue to explore this issue this coming year.  

Two issues arose this year which were only tangentially
related to the rating system, but were considered by the RC.
First, the RC chair provided the office with a set of formulas that
could be used to determine statistical evidence of sandbagging.
These formulas could be applied periodically in order to 
identify players that may be abusing the rating system in
order to become unfairly competitive for class prizes.  The
RC does not recommend that punitive action be taken based on the
results of such statistical determination, but that players who are 
flagged as a result of these formulas can be monitored outside the
realm of statistical information.  A second issue that arose had to 
do with researchers requesting ratings data from the USCF.
The RC recommended a procedure whereby a researcher would
submit a proposal to the USCF, specifying the aims and
data requirements for the research, along with information
that would indicate that USCF members' privacy and
confidentiality would be protected.  We also suggested to the
office that the data would only be provided for a fee, since
compiling requested data in a usable form required USCF personnel
resources.  With a wealth of rich information on human 
cognitive performance, we expect that the USCF will be 
increasingly contacted about the use of ratings and
game outcomes data for academic research.

Finally, in response to questions about how to improve
the USCF rating system to track rapidly improving players, 
the RC began a discussion about the possible use of the Glicko-2
rating system as an eventual replacement for the
current system.  The Glicko-2 system is a methodology based on
peer-reviewed published research by the RC chair in 2001.
The system makes distinctions when performing ratings
calculations in the precision of ratings - a player
who has not competed in many tournaments, or a player that
competes infrequently will have a less stable rating than a
player who competes frequently, and the difference in
stability has implications on the ratings computations.  One of 
the main benefits of the Glicko-2 system is that it is able to recognize 
when a player is rapidly improving, much better than in the current
USCF system.  In support of the Glicko-2 system, we
discovered that the Australian Chess Federation now uses the
Glicko-2 system to rate all of its over-the-board tournaments.
While this discussion is very much in the early stages, 
committee members knowledgeable about the Glicko-2 system
were in support of its potential use for rating USCF players.