Here is a problem you run into if you do not balance all DGs for 1v1... Dgs will have different playability in 3v3 then 5v5 brackets.  Like some DGs may be too good in 3v3 but normal in 5v5. Some may stink in 3v3 but be acceptabnle in 5v5. There will still be DGs that stink in both compraed to other DGs!  And like you said you don't even want to discuss stacking DGs! 

This can all be observed in WoW Arena right now. A format where they had 2v2, 3v3 and 5v5. You quickly realize that because the game is not balanced in 1v1, some classes get pushed a side in different formats.  So while refusing to balance the game 1v1, blizzard esentially fails to balance the game at all.  What end up happening is that both sides have same line up, and people who play classes not in the "best" lineup are SOL!

From my experience cash reward tourneys always do bring out the imbalances in a game. No matter how lame, obscure, imba, glitched, or w/e, people playing for cash prizes will pull out all the stops and (ab)use whatever they percieve to be the strongest. Over the 3+ year lifespan of Guild Wars, the metagame has undergone literally hundreds of changes, and many of them were driven by the results of tournaments, the proliferation of tourney-winning tactics immedieately following the tourney.

Of course, even more of them were driven by patches following the tournaments which steadily nerfed whatever was popular in the preceeding tourney, a balancing policy I disagreed with.  

Which is exactly what I meant with my OP.

I just wanted to show that the good teams (premades) which account for a lot of wins are usually the same Demigods as you would see in a tournament: Erebus, Oak, UB and Sedna. The 4 Demigods with more wins than losses.

Most players adapt their playstyle after they see a more efficient one in their matches. We can clearly see that with the popular Minion Erebus build, which almost nobody played in the beginning. So imho the available data is quite representative for overall balance. Average favor points do not seem to be representative for balance.

I'm just waiting for the Spirit Oak hype...

Most of the data is noise generated by bad players, and without actually watching a game to see what happens statistics can only tell you so much. Tournaments demonstrate what happens when good players compete against each other, which is the level the game needs to be balanced at. There is so little micro involved in playing Demigod there's really no need to cater to poor play when balancing, the effort required to get good is minimal.

well ... from the few things I have seen, it seems there is a problem into items combo (I think).

In a local/offline tournament, I have played a game vs an immortal Regulus dodging almost 50/60% of our/enemy blows, and firing at an insane rate.

He flies into our base, and he stands against me and the other 3 demigods with almost no effort.

ps: I won the game by powering up troops, and succeded to destroy the enemy tower.

I agree that without closely monitoring games the stats will not really tell us much. We need to really see who does what and why people lose in order to make adjustments. Mee stats cannot provide this insight. Sure winning stats can give you some insight into what DGs are the most popular, but it cannot tell you how any DG should be buffed.

Monitoring the vast majority of games would be a bad idea I think. Monitoring the tournaments that mean something would be a good idea. Balancing those so that each and every DG can be made just as good as the next one in the right hands is the ultimate goal. It seems only fair that this game should be made about viable builds, startegy and good teamplay and not about four DGs that are owning all else.