Healers make a huge difference in PvP so it's surprising they lag in general...Shaman being the exception is no surprise: They are arguably the best class for PvP. Not only can they heal but they have decent armor and can do massive burst damage, purge buffs, interrupt casts, slow opponents, flag run, and recover from diabling effects. Further, most of their abilities are instant cast which makes them more difficult to stop in PvP.

It would be interesting to track wins/losses horde vs alliance now vs a month after Burning Crusade. Not sure if there is a way to gather that information. Honor earned in a week might be one available proxy. Inspection requires very close range though and the bots would have to pick it up at the battlemasters as players came up to get in queue. Another proxy for wins could be mark turn-in events at the same locations.

Don't know if your research methods can address this but:
1) It is widely believed that the Horde dominates the battlegrounds. What is the Horde advantage ( = what % of the time does Horde win)?
2) Where does the Horde advantage come from? Higher level characters? The Shaman (the unique Horde character)? More experienced players (e.g., the Horde might have more experienced players if people are most likely to pick Alliance for their first characters)?
3) Are Horde BG players engaging in more social cooperation (more teams, more coordinated tactics)?
I am struck by how poor the social coordination in battle is. Creating an army isn't as easy as it looks :-)

Your data look very interesting, but have you run any stats on them? There are no error bars shown. Is there a statistically significant difference between the priests and shamans on this graph?

With such a large N, everything becomes significant. The resulting ANOVA has an F = 22.6, p

- woops, the "less than" symbol cut me off :(

(less than) .001. In the post-hoc Tukey, Shamans are significantly higher than all other classes (let alone just Priests). This is part of the reason why we usually don't list the significance tests in similar cases, because they would in fact be somewhat deceptive with such a large sample.

That's there that we can officially say : Shaman are so a cheated (overpowered if you want) class !
And now we (alliance) know why we always lose versus horde in PVP...

I think this statistic is not usefull, because a lot of players don't play a lot PvP (if there is 3000 Priests and only 300 play PvP, the others will lower they PvP global score) or are too low level (a lot of people have some chars lvl 20 or less who are not playing PvP, and some classes are more selected than other => there is only a small population of shaman and warrior, and they love do PvP, you know what I mean?).

I think you need to choose only Rank 14 PvP to make a good statistic.

Sorry for my english writing, I'm not english.