Right now we must argue from the rounds we have played, if the games is balanced or not. But because there are many factors which will influence this it is difficult to have an objective result. In my opinion it's time to add some real numbers to the balance process. Therefore I suggest to have a .log file or small database, which offers this data. Examples: Tank vs. Tank emp_canyon NF_LT;8xRegen;BioDiesel;StdMG;StdCan;StdML; killed BE_AFV;12xReflective;Standard;StdMG;StdCan;StdML; Tank vs. Building emp_money BE_HEAVY;24xCompo;Coolant;RailGun;RailGun; killed NF_BARRACKS; Infantry vs. Tank emp_mvalley NF_GREN;MINE; killed BE_APC;12xStandard;Standard;StdMG; Infantry vs. Infantry (not so important, because there are already other statistic tools for this) emp_slaughtered BE_RIFLEMAN;HMG; killed NF_ENGI;SMG1; Final result emp_crossroads NF_WON It would be perfect if every server writes down these data and users can easily download it. After analyzing many rounds from different servers by hand or with simple tools we should get more accurate results for balance decisions. And when enough rounds are logged some fail rounds will only have a small impact of the result. From this data we will get for example the following informations: -Which tanks with which configuration kills other tanks more or less often -Which armor, weapons and engines are researched on which map -On which map NF or BE wins more often -...
Data doesn't define what is and isn't balanced. In TF2 the scout has very low damage output, kills, assists and points compared to other classes. He's also balanced nigh perfectly, his role isn't to rack up points, it's a variety of other things, and he's highly dependant on skill. You can't measure how good the scout is, and the same goes for empires. Except in very large scale things, like as you said, Win/Loss ratio for each team in general and on specific maps.
True, but statistics can be useful. If as an example, the scout statistics would show it has the highest kill count of all, yet as you say it's not his role, then statistics could help a bit, even though they won't give an accurate representation of balance.
