DISCUSSIONS on “Rules for online games in HoMM3 and formulas for calculating player rating.”

Everything is interconnected; it's a multi-parameter balance, after all... It's easy to say – ban (nerf Shakti, Galtran, Mullich...). But what guarantee is there that, for example, Kirah (Tazar, and who knows who else) won't become unacceptably overpowered? And then, after... twelve weeks, you'll nerf them too? Has anyone calculated the consequences of a nerf (ban), gathered enough game statistics, held a test tournament, and then analyzed the results of that tournament with discussion? What guarantee is there that the imbalance won't increase with the proposed measures? If you want to experiment, create your own alternative rules project. But no, you have to force everyone to play with little-tested solutions...
P.S. This doesn't apply to most of the changes proposed by Pasha (he mainly doesn't touch the gameplay, i.e., the balance) – although I would suggest (and I have many suggestions :) ) – to only play with blue Goshans (and/or the player using Goshans has only one resurrection, and/or the player against Goshans can recruit banned heroes and/or have some other advantages). I could claim that my version is better because it doesn't "cut into the living" according to the principle – "the best remedy for dandruff is the guillotine." But I won't claim it :) – because I haven't thought everything through properly, haven't listened to all the pros and cons, haven't held a test tournament, etc...
And specifically regarding Shakti: for example, Rampart and Dungeon are quite a good pair; when playing as Dungeon, you go for a wider range of good and bad outcomes (analyze the game statistics for specific cities in a season...). Yes, there are sometimes uber-builds for Shakti, like 2-3 consecutive structures nearby, but that's a bonus, relatively infrequent... So, you nerf Shakti, Goshans, who's next on the chopping block? :) This impoverishes the essence of the game, in my firm opinion.