Comments on: Statistics of all the tanks using the “new method”

By: EvilSoul

EvilSoul — Tue, 24 Dec 2013 15:45:53 +0000

Finally a reliable method to examine real performance of the tanks.

By: Eject91

Eject91 — Tue, 24 Dec 2013 13:01:46 +0000

What surprised me the most – Excelsior.

By: bjshnog

bjshnog — Tue, 24 Dec 2013 02:39:00 +0000

By normalize, I mean use some algorithm to calculate a rating to replace overall win rate, which treats each player’s performance between their tanks approximately equally, instead of treating 60% pure seal-clubbers as great players.

By: bjshnog

bjshnog — Tue, 24 Dec 2013 02:36:32 +0000

In order to come to a proper *accurate* conclusion about the relative winning ability of a tank, you need to normalize overall win rate first. It could be that the best players who play the WTE100, for example, have a strong tendency to avoid playing any tiers other than 10. Thus, their overall win rates would be offset negatively by a small amount, making it appear more OP than it actually is. This, if course, is just an example, but it shows how overall win rate skews or just generally distorts the measured player quality distribution.

By: MarcusTheAbsolute .

MarcusTheAbsolute . — Mon, 23 Dec 2013 23:07:14 +0000

Is it normal that I post a comment and it doesn’t appear in the comments section for me?

By: MarcusTheAbsolute .

MarcusTheAbsolute . — Mon, 23 Dec 2013 23:04:21 +0000

Though an interesting read, some of this stuff seems pretty weird. I’ve been going over the complete statistics and some of it just doesn’t make any sense whatsoever. Four examples:

1) The premium Sexton I seems to be much more friendly to people around 51% wr whereas the Sexton II (which is basically better in all respects) is only good for noobs by the stats.

2) The soviet IS is disturbingly underpowered according to the graph. IS2…more or less an identical tank is shown as grossly OP.

3) Chinese Type T-34, despite being more or less the Soviet T-34 without the top gun apparently completely rolls over the Soviet counterpart. Two identical vehicles…and the one with an extra gun is way worse?

4) VK4502A vs Tiger II. VK has more mobility. Tiger has more armor and incomparably better gun. Tiger II comes off underpowered in all WR regions compared to VK’s stats. How?

Something smells fishy.

By: Homer_J

Homer_J — Mon, 23 Dec 2013 21:40:03 +0000

Only intelligent life forms are represented. ;)

By: Homer_J

Homer_J — Mon, 23 Dec 2013 21:37:14 +0000

Because bad players become good when they own a tank for a long time, but only with that one tank.

I don’t buy that theory at all.

By: Homer_J

Homer_J — Mon, 23 Dec 2013 21:34:28 +0000

It would be just as bad at tier 8 with historical armour. It’s just a fail design. Far too tall (and long) to work with a rear turret.

The hull option for 4502A would make the problem go away because nobody would use it.

By: hillin

hillin — Mon, 23 Dec 2013 21:09:03 +0000

A few thoughts:
1) axises should be made Gaussian rather than linear to represent the real win rate distribution. These graphs are only partially useful, because it’s very rare to see a player with a WR>55%, however they occupied 1/3 more of the axises.
2) sample size is very important. Something like a PzII ausf J is definitely OP, however the sample size is small enough to make its graph very inaccurate. So ignore those rare vehicles.
3) Gosh I would like to have these data in Tank Inspector. Very impressive.