Storm: IS-7 drivers are not nubs, unlike stats theorists say

[Garbad 10,517]

For the Retards poasted this today:

 

- Storm states that the theory that Maus stats are higher than they should be because more experienced players play it than for example IS-7 was checked by WG and proven false

 

I would like someone very fluent in russian to check this.  I can see two possible interpretations 1) Strawman -- maus owners have more games played than is7 drivers, or 2) actual theory -- russians tend to pick russian tanks as their first line, and thus, tend to be more nubish and get bad stats.  This explains why great russian tanks get bad stats on the RU server but not other servers.  By contrast, obscure lines like brits get played mostly by vets, which is why these lines have great average stats despite sucking dick.  <-- If they claim to disprove this, I'd like to know how as the evidence we have suggests this is true.

 

What do you think?

[Servios 467]

"Of all tier 10′s, [Maus is the] first in damage blocked by armor and second in winrate and average XP per battle."

[CraBeatOff 5,741]

RichardNixon has some nice data on recency effects and general trends of lines played and how it effects outcomes. Hopefully he'll have a write-up of what he's been collecting for a few months now, as it touches directly on the phenomena (and supports it) you've identified here. 

 

Its also just harder to fail in a Maus at the skill level of the bulk of the population. You don't exactly succeed either, but those same players don't really succeed in any tank but our measures. 

[MaxL_1023 1,295]

The "statisticians" at WG are probably forgetting to chronologically check the IS-7(Example) battles against the account sampled. If someone went up the IS-7 line, the IS-4 line and then the E-100 line, for most players their stats would be better for each successive line played. However, if you just checked the 5000 battle player with these 3 tier 10s by their stats page, it would look like the Russian tanks are underpowered due to performance. 

 

Measures like "class-tier WR" are useful for tanks which tend to be fairly randomly played throughout an account's lifetime - they assume that the skill of the player has not changed for the duration of the study. Taking a sample of greater then ~10% of an account's games poses a severe risk of introducing this type of bias. 

 

The best way to mitigate this is to take a large sample of accounts with more then 10,000 games played (therefore multiple tier 10s for the most part) and then check only the last ~10% of their battles. (A player with 40K games is probably not getting better or worse in the last 4K, while a 10K likely would). They can then measure the play frequency and relative performance of every tank played with a reasonable assumption that the player skill is the same for all vehicles within any one account. 

 

This would allow measures such as class-tier WR to be effective, and to draw statistical studies on play frequency versus account WR, play frequency vs 1000 battle WR and to determine other advanced performance metrics using factorial analysis. OFAT is shit - WG probably uses it to balance tanks.

 

If it turns out that not many people are using the IS-7 in their last 1000 battles, but a lot more are using the E-100, a simple comparison to account wide play frequency will easily reveal the type of chronological bias which we suspect exists. 

 

P.S: The IS-7 is not underpowered. The IS-4 is probably below average, as is the KV-4 and maybe KV-3. The ST-I and KV-IS are above average (or broken for the IS) and the IS/IS-3/IS-8 and T-150 are all fairly well balanced. They are just not that noob friendly.

[CraBeatOff 5,741]

Nothing I've seen coming out of WG looks like they have anything remotely like a statistician on retainer, let alone in house. WG Personal Rating was prob contracted out, and it actually worked out pretty well for them but the guy(s) who did that are not on the regular "to call" list for Serb & co. Never forget 1.3 sigma...

[Ollie Tabooger 614]

The Maus is probably only that high in winrate because the armor and HP pool give it a high skill floor. It's kind of hard not to do at least 3-4 shots of damage when you can soak so many hits I would imagine.

[SoliDeoGloria 417]

My IS-7 was my first 10, and it is by far my worst performing, but now I have been improving in it, whenI actually take it out.

[SFC_Storm 58]

For the Retards poasted this today:

 

- Storm states that the theory that Maus stats are higher than they should be because more experienced players play it than for example IS-7 was checked by WG and proven false

 

I would like someone very fluent in russian to check this.  I can see two possible interpretations 1) Strawman -- maus owners have more games played than is7 drivers, or 2) actual theory -- russians tend to pick russian tanks as their first line, and thus, tend to be more nubish and get bad stats.  This explains why great russian tanks get bad stats on the RU server but not other servers.  By contrast, obscure lines like brits get played mostly by vets, which is why these lines have great average stats despite sucking dick.  <-- If they claim to disprove this, I'd like to know how as the evidence we have suggests this is true.

 

What do you think?

We have known this forever.

 

When you see a tank like the T54 [esp before E50 was good] be barely 50% W/R you knew it was the RU server. On the US server it was more like 53% and still every newb and his bro had one.

 

This is also why the T30 was nerfed so hard when it was a heavy. I remember there reason was its W/R was like 56% and the T30 was way OP.

 

In reality the T30 needed much skill to be used and had T7 armor, but great gun handling and a huge gun, this made the pros on RU get it ASAP and made all the baddies not want it because it was an obscure US tank not famous like the Patton/Pershing etc. I remember great CW teams running 7 T30`s routinely.

 

Anyways, this is also why tanks like the T57 stick out, in pro hands its insanely good and on RU many baddies don`t have it still. How about the M48 or E4, both were used heavily on RU by a very small number that made them OP.

 

IMO this is why tanks need to be balanced across all servers and not just RU. This really is the def of Russian Bias.

[Painbringer71 3]

Quicker game exposer?

 

The Maus when it gets there... lots of times the battle has been going on a while.  Just like a tog cleaning up scraps at mid match... going slow is its asset in that situation. 

[Methos 7]

My Is-7 was my first tier 10 as well. All my noobish friends who have a tier 10 even have one. I never got back to playing it after I sold it when I left my first clan, but I intend to buy it back in the future.

[Crossfader 1,142]

I still dont think that the IS-7 should have a .4 accuracy and 3.4 second aim time (which i do believe is the worst in tier, not sure).

 

Maus should have great DPM, its not like it can run around to abuse it, if you sit in front of it, you should get punished.

[Alphabloom 95]

By contrast, obscure lines like brits get played mostly by vets, which is why these lines have great average stats despite sucking dick.  <-- If they claim to disprove this, I'd like to know how as the evidence we have suggests this is true.

 

What do you think?

Im not sure I classify myself as a vet, but those British tier 10's are just comfortable to play. In response to your main question, Im not sure, i have never really grinded the Russian lines, once or twice I played on the test server and the tank itself was pretty comfortable, considering im getting 400+ ping   

[GehakteMolen 2,063]

Nothing I've seen coming out of WG looks like they have anything remotely like a statistician on retainer, let alone in house. WG Personal Rating was prob contracted out, and it actually worked out pretty well for them but the guy(s) who did that are not on the regular "to call" list for Serb & co. Never forget 1.3 sigma...

 

Serb is a nuclear physicst or engineer or so, so im quite sure he knows more math stuff as most of the players

 

that said, Serb is job is ``now`` mostly ``njet`` stuff i think, the time Serb was active involved in game development was long gone (that Zlobny guy or whatever hes name was came after Serb (or he worked there and got promoted))

 

Zlobny or so was a big fan of OP bs tanks like autoloaders, arty with big alpha and other extreme bs, Serb is more the historical accuracy guy

 

That said, most answer were and are trolling, Storm just makes it less obvious as Serb, but the answer remain often the same (even if maus would be total underpowered they wont tell)

 

ps: an old ``forum wisdom`` said:

- On Ru server ussr tanks are UP

- On EU server Ger tanks are UP

- On US server US tanks are UP

 

people grind there own nation first, fail in that, and ruin the stats of it

 

to compare tank stats, the first few k games should be disgarded...

 

Snib did this long, long ago already:

http://forum.worldoftanks.eu/index.php?/topic/65256-unofficial-eu-server-statistics/

 

last spoiler:

Here, we are only looking at players playing more than one tank of the same class in the same tier. In other words, e.g. a tier 8 arty player's stats will only be considered here if they own at least one other tier 8 arty and played it during the analyzed time-frame. 

 

This already removes a big part of the ``stupid``

 

ps ps: and WG does that for balancing, or atleast did, the way they balanced was something like:

- pick 20k people (random)

- remove all with <47% win

- remove all with > 55% win (can also be higher 57 or so, dont remember, was ages ago, when 60% win was really good)

 

And: remove all with < x k battles (3k i think, also cant remember, but im not even sure if this was included)

 

This filters out most bads, only the popularity is hard to check, but for that avg dmg, avg income and avg exp are good measurments

 

Avg exp and avg income beiing 2 balance parameters!!!, and unlike what all sorts of people have said over the time, for a certain lvl of skill, they should be almost identical, expensive ammo gets compensated by higher credit coefficient

 

ps ps ps: wild guess, avg exp is the main balance parameter, winratio is only used as back up and to filter strange odities, winratio gets afterall influenced by popularity (see KV-1s, if there are 5 KV-1s every battle the avg will be 49% no matter how OP it is, avg exp will however show it...)