Which metrics give a better measure than global win rate?

Question

Inspired by this question I wanted to find out which metrics within World of Tanks are a better measure of performance than the global win rate when comparing tanks of different nations?

In my opinion using global win rate, or win rate at all, is a highly flawed measure of a tanks balance against other tanks. When comparing like for like, such as tier V heavy tanks, opinions seem to be split by which tank each person prefers rather than any objective measure or weighting system.

Measuring by win rate ignores many of the factors that go into making up the win rate such as the composition and quality of the players playing a particular tank, class, or nation. It also doesn't factor in the difference between high tier and low tier battles as some nations, like France appear to be balanced between high and low tiers. i.e. Because the French low tiers are so awful they appear to be balanced by making the high tiers much better.

Answer 1

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.

Answer 2

Global win rate is not perfect metric of performance, but after having enough battles, it alone will give a decent indication of that persons skill. Combined with other metrics (damage done, kills, cap/def points etc.) you get efficiency/WN6/WN7 -ratings, which again gives you information on a person.

On tank-level, like you asked, it gets bit trickier. This is problematic because the only tanks-level stat that is publicly available is winrate. There are tools like mentioned http://vbaddict.net, which use the dossier/battle-report files wot generates to give very detailed information on your performance. That tool also has bit scewd information since most players that use tools like that are better than average, thus the information doesnt represent the whole WoT playerbase. http://www.noobmeter.com/tankStats/eu has quite good information about individual tanks WR and also the average WR of the tankers that have driven them. The difference between those two numbers can tell how well a certain tank if performing against other tanks of similar tier. http://www.visualstats.net/cgi-bin/visual is another site that has very informative graphs of different statistics. They focus around player stats rather than tank stats tho. http://wot-news.com/stat/server/eu/norm/en is one more site that does quite a lot of stats and there you can also find different tanks to compare. You even see the fluctuations in tanks WR over time so it is easily noticeable how certain nerfs affect tanks performance.

Like said, the problem with tank comparison is that there is so little data available for "data-mining" and analysis that WR is unfortunately the number you have to rely on. There are fortunately different ways to use WR and get the wanted information from it.