Yorkshire Chess
Association |
Last
update: |
|
The Meaning of Grades |
|
Chess grades are an attempt to
measure the relative playing strength of players. They
are based on past results and are thus a measure of average performance over those games,
which may span a period of one, two or even three years, rather than a measure of current
"form".
Qualitatively, the higher the
grade the greater is the supposed strength. Thus
the higher graded of two players is presumed to be the stronger in average performance
against the same opponents.
Quantitatively, the difference
between a player's grade and that of an opponent is the predicted difference from 50% in
the player's score, over a number of games, against players of the opponent's grade. Thus if a player graded 135 played an number of
games against opponents graded 115, then that player would be expected on the basis of the
grades to score 70%. This is fairly clearly
related to the way a player normally scores his opponent's grade plus 50 grading points
for winning a game, and scores his opponent's grade in grading points for a draw. A difference of more than 50 points does of course
mean the same as a difference of 50 points. (See
Problems Caused by Simplification)
These predictions are not related to
the individual games results of win, draw or loss. In
the above example the player graded 135 could achieve his 70% with +7=0-3 (i.e.7 wins, no
draws and 3 losses), or with +6=2-2, or with +5=4-0.
A player who liked to press, albeit speculatively, for wins would be more likely to
score +7=0-3, whereas the cautious player whose first priority was to avoid losing would
be more likely to score +5=4-0. The grading
system does not seek to address such differences.
Importantly, grades offer quite a
poor basis for the prediction of the result of a single game, if the players are quite
closely matched. In a single game you can
score only 100%, 50% or 0%. So what does a
predicted score of 70% mean in a single game? You'd
be looking for an answer like "an x% chance of a win for the higher graded player, a
y% of a draw, and a z% of a loss". Such
a prediction regarding a single game is beyond the scope of the grading system.
There is nothing in the grading
system which says how much higher graded than your opponent you need to be in order to be
expected to win. Superficially, a difference
of more than 25 points would suggest that a win for the higher graded player was more
probable than a draw, as the chances would be nearer 100% than 50%, but that is rather
simplistic. An analysis of actual results in
comparison with grade difference could throw light on that question. There is also to be considered the disparity
between White and Black which is assumed to average out over a number of games, but which
is generally accepted as slightly tilting the odds in a single game between otherwise
equal opponents.
Predicting team match results
is, on the other hand, quite feasible. You
total each team's grades, divide the difference between the two totals by the number of
boards, and the result is the difference from 50% in the expected game point totals. Limiting the difference in grades used on any
given board to at most 50 will improve the prediction.