One Click Rating System in a Nutshell
Updated: Tuesday, August 29, 2006 - 5:00 p.m. Pacific Time
Rationale for Ratings
One Click matches players in time, space, and skill level. Most players enjoy close matches like
6-4, 6-7, 7-5
and despise lopsided matches like 6-0, 6-1. At One Click, we take pride in arranging close matches
sophisticated rating system that we developed using over 7,000 matches played in the Sunnyvale tennis
One Click rating system places players on the scale 1.5 through 7 used by National Tennis Rating Program
by United States Tennis Association (USTA) (Click
We adjust ratings after every match arranged by One Click for which scores were reported.
The system is proven to be statistically accurate in predicting whether a given pair of opponents will
a close or lopsided match. For example, pre-match rating difference of less than 0.1 leads almost
inevitably to a close
match, while difference of more than 0.4 leads to an easy win by a higher rated opponent.
Your rating posted on the website reflects your performance in four recently played matches (or less
if you just joined). The posted rating is a weighted average of raw ratings recorded after each of the
This is why in general the post-match rating transitions of both opponents may indicate the same
direction of change.
The raw rating transitions related to a specific match are always in opposite direction for each
opponent, but the multi-match
average may behave differently from the raw rating.
The calculation of raw ratings is, in a nutshell, based on the pre-match rating difference between
opponents and the game-count
match outcome to produce rating adjustments. The principal rating adjustments follow an empirically
established curve that
provides statistically expected raw rating difference for given match outcome.
Fig. 1: Example graph of function D=F(G)
The graph shows how difference D between ratings of match opponents statistically depends on parameter G
that we call
"games won ratio", the number of games won by the first opponent divided by the total number of games
played in a match.
The actual "rating difference"-versus-"games won ratio" curve that we use has been determined by
analyzing about 7,000
matches played by our community over past two years. We use separate curves for singles and doubles
matches to reflect
specific dynamics of each match type.
When two opponents play a match, we obtain the games won ratio Gm and calculate the
of the rating difference Dm. This match difference may be different from the initial
difference Di. The discrepancy Dm-Di indicates whether the ratings of
opponents should be subject to adjustment. It also suggests the direction and magnitude of potential
The following example shows the idea of determining the adjustments.
Player rated 3.37 scored 7-6, 6-4 in a match with player rated 3.12. The games won ratio Gm
for the match
is (7+6)/(7+6+6+4) = .565. Suppose that the curve shown in Figure 1 gives Dm = 0.19 for
Gm = .565.
Pre-match rating difference Di is 3.37 - 3.12 = 0.25. The rating difference discrepancy
Di is 0.19 - 0.25 = -0.06. The match outcome indicates that the opponents are closer to each
rating than the pre-match difference Di = 0.25 suggests. We want to reflect this observation
the rating of 3.37 by a fraction of the discrepancy 0.06 while increasing the rating 3.12 by a fraction
of the discrepancy
0.06. The actual fraction size depends on several factors mentioned below.
Please note that in this example the match winner's rating will go down while the match loser's rating
will go up.
This shows how the rating system works. You don't have to win matches to increase your rating. All you
need is to perform
better than the pre-match rating difference indicates based on our statistical model.
Empirically Established Factors
The principal adjustments derived from the "rating difference"-versus-"games won ratio" curve are
subject to refinement
by applying several empirically-established factors:
- Match score "strength" - the more games played in a match the more
reliable indication of the skill difference
between the opponents
- Randomness related to skill spread among players in doubles matches
- the wider range of skills among a foursome
the less predictable match outcome
- Patterns of individual performance fluctuations over time - this is
based on the direction and variability of
the rating adjustments recorded in the recently played matches
- Singles versus doubles match differentiation - the rating formula
parameters differ between these for singles
and doubles matches
- Compensation for the effect of closed community - the highest rated
players of our community do not face opponents
with rating higher than theirs; we slightly boost their rating adjustments to offset this
- Slam win provision - the hardest case for any model-based rating is
a match in which all games were won by one
of the opponents; we make proper exceptions for such matches
- Initiation matches - the initial rating of self-rated players may
be unreliable until they play
a few "initiation" matches; we make provisions to avoid undesired impact of these matches on
players with established ratings
Rating for New Players
Players who register with One Click receive the initial dynamic rating value based on their
USTA-published or self-declared
NTRP level. For male players, the initial dynamic rating value is the NTRP level minus 0.25 as we
place players in the middle
of a respective NTRP half-point bracket. For female players, we multiply the "NTRP minus 0.25" by
designed to put players of both genders on the same rating scale. The scaling factor value,
currently 5/6, is subject to change
in the future based on the observed statistical relationship between the One Click ratings of our
female members versus
their published NTRP levels.
The values derived from self-declared NTRP figures are subject to possible revision by the One Click
if the initial matches played by newly admitted players indicate a need for such revision.
Our multi-year experience in tennis rating indicates that most players become accurately rated after
playing just a
few matches in the system, regardless of the initial self-rating.
Learning More about Ratings
Most players just watch how their ratings move match-to-match as reported on the One Click website.
Majority of rating
transitions are consistent with our intuition about who was a better opponent in a given match.
However, if you do not mind exploring
"heavy math" of the One Click rating model, you are welcome to click here
to request a
white paper that describes the details of this unique system.