is a Rick Rating?
When I was a team captain for a senior men's
team I found it hard to rank players that I
didn't know and even worst I didn't rank players
that I did know correctly. I had one guy
who looked like a 4.0 player, but his results
were always like a low 3.0 player.
To help me deal with this problem and to give me
a competitive edge against other teams I wrote a
computer program to generate ratings based on
playing results using the USTA TennisLink data
My program generates USTA-like ratings that I
call "Rick ratings". I researched how the
USTA generates their ratings and I decided to
copy what they do and to do some things
differently. They don't like to see too
many players moving up and down in ratings
because that would be disruptive to the leagues
too much. I my case my sole goal is to
compare players to each other. So my
results will be different, but I think that my
program's output is fairly good for what it
tries to do: compare playing performances.
problems with my software, but even the USTA
rating system has problems. Like when two
players always play together you can't break
apart their playing levels.
My rating system is designed so that it spreads
out the rankings for a playing level.
For example: a person playing a 3.0
league with average results playing on court 2
gets a Rick Rating = 3.25 (The mid-point between
3.0 and 3.5). So it should work out that
50% of the players will be above and 50% below
My rating system works like the USTA system.
It doesn't care about who wins matches it only
looks at who wins the highest percentage of
games. Like the USTA system you could lose
all of your matches and still have your rating
go up if you win more games than you lose.
Like the USTA system it also factors in the
ratings of the players you are playing against.
If your rating is higher than your opponents by
a lot, then beating them by a little could cause
your rating to go down.
The software expected you to win by a
One problem is that the quality of the results
depend greatly on the amount of data.
The more matches a player is rated on the
better the rating.
So a weak player could play only 1 match,
but end up with a high rating if they had been
luckily enough to play with a strong player on
court 1 and the won that match.
Also the more players play with different
playing partners the better their rating will
reflect their playing level.
Two people who only play together will
end up with the same rating no matter how
different their playing skills are.
There are more than one way to run my software
so rating results can vary.
So it's not how accurate the ratings are
that should be looked at, but how accurate the
It is not so accurate that you can trust
who it ranks number 1 vs. number 2, or even a
bigger spread like number 5 vs. number 8.
The system is more correct as the
differences get larger.
The number 1 player is going to be a
stronger player than the number 10 player.
Really it's more accurate to say that the
number 1 player had better playing results in
the league than players ranked below.
The software outputs to an Excel file and I
color code the spreadsheet to help to highlight
To judge the quality of the Rick Ratings is easy
even without knowing anything about the players.
You add the ratings together of each double pair
playing a match and if the pair with the higher
ratings wins then the ratings predicted that win
correctly. In a test that I ran it
correctly predicted the winner 90% of the time.