EDIT: I quickly found this site after finishing this post; it contains averaged data for a large number of tanks across some 940,000 battles, including WN7 and PA Efficiency stats.
The various rating systems out there (WN7 is the one I prefer, and is the one Radium mentioned) can in fact be used. While Xenox accurately pointed out that such equations compare players to other players, data subsets could be substituted for full player stats to give ratings for that player in a given tank. (See footnote.) Averaging such data across a large number of players ought to give a moderately accurate metric by which tanks could be compared - a better tank would, on average, yield higher WN7 scores.
Unfortunately this would be by no means a perfect system - tank popularity could skew data, and it is difficult to account for player skill or tank crew levels - but it ought to give a decent impression. An additional difficulty lies in acquiring this data from a user level - there are a few tools that enhance the official statistics pages, but to my knowledge acquiring such data for individual tanks can only be done in-game, and the in-game service record lacks important data used in WN7 calculations.
As an example, I used the WN7 formula to score my BDR G1 B Tier V Heavy tank, and received a score of ~1400. I had to fudge a couple numbers (I couldn't find spot or defense point scores specific to my BDR tank) and there are likely rounding errors, but it can be done. With a little encouragement, the WoT community may even be convinced to write a tool to acquire and compile the relevant information and produce scores for specific tanks! I would be interested to browse such a tool myself...
Footnote: Minor modifications to the formula may need to be made to account for smaller numbers, particularly in the Number of Battles Played - alternatively, that segment could be left out entirely. It didn't come up during my calculations; using only stats from Tier V or higher tanks, the term will always result in 0. Comparing lower-tier tanks may be difficult and the term would likely need to account for some full-player data to avoid skew from experienced players going sealclubbing... but I'll leave such discussion to a larger and more math-experienced crowd.