Glicko-2 converges within 1000 iterations or something is wrong
As described in http://glicko.net/glicko/glicko2.pdf with 15 games per period: The main iteration of the Illinois algorithm to narrow the bracket around ln(σ02) is reasonably quick. From simulation analyses, the median number of iterations to narrow the bracket to a width of less than ε = 0.000001 was 5, with a mean of 5.6, and a maximum of 19 (in 10000 simulations).pull/8279/head
parent
34d775c73d
commit
4136df6bcc
|
@ -25,7 +25,7 @@ public class RatingCalculator {
|
|||
private final static double DEFAULT_TAU = 0.75;
|
||||
private final static double MULTIPLIER = 173.7178;
|
||||
private final static double CONVERGENCE_TOLERANCE = 0.000001;
|
||||
private final static int ITERATION_MAX = 3000;
|
||||
private final static int ITERATION_MAX = 1000;
|
||||
private final static double DAYS_PER_MILLI = 1.0 / (1000 * 60 * 60 * 24);
|
||||
|
||||
private final double tau; // constrains volatility over time
|
||||
|
|
Loading…
Reference in New Issue