Main changes
- * A table of expectancies can be obtained from a switch, to illustrate what the rating points mean.
* The players who won all, or lost all, have a special treatment after the general rating calculation. These players are reported in the ranking as '>' or '<' a given rating. For instance, for a winner of all games, the rating assigned correspond to the one at which the probability to win all the games is 50%. Since this rating is at the midpoint, and we know that the event of winning all the games occurred, the best estimation of the rating should be above this level. This is a safe and conservative way to determine a rating in these situations without assumptions. That is why the rating is reported as "bigger than".
* Connections or lack thereof between different groups are reported, so the user could make an informed decision.
* Simulations use an empirical formula that fits real draw data, improving the accuracy of the error margins.
* The confidence level for error margins now could be set with a switch (-F). Before it was fixed to 68% (~one standard deviation), now the default is 95% (~two standard deviations) since most people are accustomed to using that.
* The names in the .csv files have been removed from spaces (bug fixed).
Code: Select all
-T display winning expectancy table
-g <file> output file with groups connected
-F <value> confidence (%) to estimate error margins. Default is 95.0Miguel