 |
|
GigaRatings is a points based rating system aimed at regional/national sports
associations based on sound probability theory. Scales, constants and variables
are set by the association. The algorithm used to calculate a player's new
rating after playing a match is based on Professor Arpad Elo's algorithm used
by the National Chess Federation.
It is based on the probabilities of winning or losing the match. If a player
plays a much higher rated player, then his/her probability of winning that
match is small. Thus, if he/she wins that match, then his/her rating is
adjusted upward by a lot. On the other hand, if they lose that match, their
rating is lowered only very slightly, if at all. Conversely, the higher rated
player would have their rating lowered by a lot if they lost to the much lower
rated player, but only raised by a little, if at all, if they win that match.
The math behind the algorithm is as follows:
WOR = Winner's old rating
WNR = Winner's new rating
LOR = Loser's old rating
LNR = Loser's new rating
K = constant
D = denominator
PW = Probability that the winner would win
PL = Probability that the loser would win
WNR = (WOR+K*(1-PW))
LNR = (LOR+K*(0-PL))
PW = (1/(POWER(10,(-(WOR-LOR)/D))+1))
PL = (1/(POWER(10,(-(LOR-WOR)/D))+1))
Like all GigaSports products, GigaRatings is an online based system, but it is
not a real time system. When matches are input, new ratings are not immediately
available online - first the GameMaster (administrator) must ensure that all
matches were input, that they are in the correct order, that new players are
allocated initial ratings etc. When all this is done, the GameMaster calculates
a new Baseline using the previous Baseline and all the matches that have been
recorded since. A Baseline is a stored record of all rated players as of a
specified date. After the GameMaster has reviewed the new Baseline and made any
necessary adjustments, he/she can release the results to the general public
with one click.
|
|
|
|
|
