[Discussion] Balance, Bias, Matchmaking and Battle Ratings

[Caernarvon02]

Cent 2 is my most used tank, It is basically invulnerable to Tiger 1 frontally in downtiers, can traverse maps quite easily and will pen near enough anything quite easy unless its a full uptier in which you become useless, 6.7 is its sweet spot. American Heavies are good at 6.7, would do just fine at 7.0.

I think you’re mistaking overtiering for BR compression. the tech difference from 6.7 to 8.7 is the most dramatic in the game making some extremely good lineups (like 7.7 britain) still be destroyed in uptiers because its tanks from the early 50s seeing stuff from the 60s and newer. Dont put them lower, spread them all out more

[Whiskey_077]

@Caernarvon02

Cent 2 is my most used tank, It is basically invulnerable to Tiger 1 frontally in downtiers, can traverse maps quite easily and will pen near enough anything quite easy unless its a full uptier in which you become useless, 6.7 is its sweet spot. American Heavies are good at 6.7, would do just fine at 7.0.

You didn’t really play with it. The issue is a mess to play with it if you’re not under 7.0 battle and of course, if decompression is getting fix, it would be ‘fine’ in 6.7

[Caernarvon02]

for now they are fine enough at that BR as any higher or lower they become OP or trash, with decompression of course the BR will change

[Yilmaz1001]

Lmao @ moving the Sturer emil (an already op TD) down to 4.0 this update.

anyway (all of the following is said about arcade):
Premiums are blatantly, by and large, OP

Explanation: If premiums were balanced, the lines for premium (gold) and Tech tree (blue) vehicles would be coincident, or at least statistically indifferentiable)

Russian Bias is immense and impossible to ignore (+ Gaijin hates Italy):

Explanation: If Russian bias weren’t real, the Russian line would be in line with all the others instead of nearly twice as high. For the second graph, The Russian line would have an average of 1, not start at nearly 2 and go off into space at BR>10

AAs (especially at lower BRs) and, to a lesser exent, LTs are systemically overranked in arcade:




Explanations: These graphs display SL per match, RP per match, Win%, and K/Ds by vehicle type. If vehicle types were acurrately assigned BRs, all these lines would be coincident

Seriously though, Russian bias and, to a lesser extent, American bias is insane and obvious:


Explanation: If the game were balanced, in the first graph, every nation except Israel would have a value around 2, with israel slightly below (because it starts at rank IV, where more skilled players play), and all nations would be around 1 on the second graph. Note france and Israel being below 2 on the first graph, but above 1 on the second, this indicates that they’re actually unbalanced, and their seeming deficiencies are an artifact of the fact that I have a lot of data for France from higher BRs and Israel starts at rank IV. Countries like The US, Russia, and to a lesser extent, Germany, being above average in both indicates poor balance.

and AAs are really obviously on the backfoot compared to ground vehicles:

Explanation: better blaance between AAs and other AFVs would be achieved if AAs (high Air KD vehicles) could avhieve KDs similar to ground vehicles, but by a factor of around 3, they’re SOL)

A couple egregiously undertiered vehicles are as follows:

FR	D2
RU	SU-57
JP	Ro-Go Exp.
SW	SAV 20.12.48
SW	Strv m/42 EH
RU	T-28E
IT	Sherman I Composito
JP	M24
GE	Tiger H1
GE	Pz.IV G
FR	AMX-30 S DCA
RU	KV-7
UK	Stuart I
CN	T-26
GE	Flakpanzer 38
UK	Crusader III
RU	T-35
US	T14
RU	T-28
FR	2C bis
CN	M42
GE	Sturer Emil
US	M3 GMC
SW	Strv m/41 S-I
GE	15cm sIG 33 B Sfl
FR	SANTAL
RU	T-26-4
RU	SU-122
FR	Panther Dauphiné
GE	Nb.Fz.
US	M3 Stuart
RU	T-34 (1940)
GE	Sd.Kfz.251/9
FR	2C
JP	Chi-Ha Short Gun
GE	Flakpanzer 1
RU	SU-76M
GE	Pz.IV H
US	M24
US	M2
RU	T-50
RU	T-34 (1942)
RU	T-III

Provisional BR recommendations, for those I excluded, my mathematical model broke down:
|D2|3.3|
|SU-57|3|
|Ro-Go Exp.|3|
|Strv m/42 EH|3.3|
|T-28E|3.7|
|Sherman I Composito|4.7|
|M24|3.7|
|Tiger H1|7.7|
|KV-7|5.7|
|Stuart I|3.3|
|T-26|3|
|Flakpanzer 38|3|
|Crusader III|3.3|
|T-35|3.7|
|T14|6.7|
|T-28|3.7|
|2C bis|3.7|
|Strv m/41 S-I|3.3|
|15cm sIG 33 B Sfl|3|
|SU-122|3.7|
|Panther Dauphiné|7.3|
|Nb.Fz.|3.3|
|M3 Stuart|3.3|
|T-34 (1940)|3.7|
|Sd.Kfz.251/9|3|
|2C|3.7|
|Chi-Ha Short Gun|3.3|
|Flakpanzer 1|3|
|SU-76M|3|
|M2|3.3|
|T-50|3.3|
|T-34 (1942)|4.3|
|T-III|3.3|

[Ion_Protogen]

@Yilmaz1001

Premiums are blatantly, by and large, OP

What is that graph showing?

And where are you getting this data from?

@Yilmaz1001

A couple egregiously undertiered vehicles are as follows:

I don’t see that many undertiered tanks in that list. The majority of those are low tier tanks that are incredibly good in the hands of a good player. Some of those are undertiered such as the M24s, but some are not good, such as AMX-30DCA or T-26-4.

@Yilmaz1001

Provisional BR recommendations, for those I excluded, my mathematical model broke down:

I think your model needs some work.

I think data analysis of BRs is a great idea, but it isn’t entirely clear what all the graphs are showing, in my opinion.

[Yilmaz1001]

What is that graph showing?

Assuming you’re talking about the graph regarding preems.

it is displaying the average K/D ratio of premium and tech tree vehicles at each br sampled. More skilled players tend to play top tier, so on average the lines slope down (seal club effect), the difference in height between them is the relative disparity in vehicle performance between premiums and TT vehicles.

And where are you getting this data from?

It is my own work. I don’t think sites like thunderskill have a robust enough sampling methodology and aren’t really adept at providing enough insight into the data to say anything meaningful. Therefore I manage my own spreadsheet and collect data about the game.

I don’t see that many undertiered tanks in that list. The majority of those are low tier tanks that are incredibly good in the hands of a good player.

That’s up for contention, but I’d like to see your model for how you differentiate between “a tank with lots of room to move up in br that experienced players can recognize as a tank with stats inappropriate for its br” and that.

I think your model needs some work.

That is correct! If Gaijin would make more data public, then this would be much easier, but until then I have to collect data about the game the hard way, so I made some compromises in the model that lets me work with less data but sacrifices accuracy to do so.

I think data analysis of BRs is a great idea, but it isn’t entirely clear what all the graphs are showing, in my opinion.

I’m flattered, and I’ve never been very good at labeling or otherwise presenting my data, so I’ll try to work on that.

[ZendikarHaven]

@Yilmaz1001

|Strv m/42 EH|3.3|
|T-28E|3.7|

…what

@Yilmaz1001

|Panther Dauphiné|7.3|

EXCUSE ME ACTUALLY WHAT

[ZendikarHaven]

@Ion_Protogen

I don’t see that many undertiered tanks in that list. The majority of those are low tier tanks that are incredibly good in the hands of a good player. Some of those are undertiered such as the M24s, but some are not good, such as AMX-30DCA or T-26-4.

The list is quite frankly nonsensical. They have the french panther, but not the german one, and their BR recommendations are even wilder.

Like, Strv m/42 EH to 3.3? What? Yeah, it can be very good when used right, but its cannon has anemic penetration and against any sort of 75mm guns its armor is downright useless.

Hell, they even recommended the Sherman I composite go to 4.7.

[Yilmaz1001]

The Panther Dauphiné is a very good tank. I´m a pretty mediocre player and have been able to get an average of a 5.0 K/D with it. T-28E is also a bit of a beast, and the m/42 EH is pretty nice if you know what you’re doing with it.

TSR %	VG1 %	VG2 %	Reload %	HA1 %	HA2 %	HA3 %	TA1 %	TA2 %	TA3 %	Pen %	Crew %	Mass %	HP %	speed %	Visibility %	HG1 %	HG2 %	PWR %
0.8371808385	0.9873241318	0.6267837529	0.8785222704	1.514420953	1.499093418	1.433889275	1.976430593	1.247989797	1.428651029	0.8444857966	1.191389079	1.547707651	1.317581685	0.8224174996	1.302316114	1.158795183	1.15850878	0.8354970103

As you can see from its stats (this measure is a percent figure, the stats of other AFVs are compiled, a line of best fit is drawn, and these percentages express (the panthers stats / the value of the line of best fit at br=6.0)), where it excels, it excels quite a lot, where it suffers, it suffers quite modestly. 7.3 is the calculated value where the sum of the residuals for each stat is minimized as each stat is weighted by the elasticity between it and K/D, win%, SL/M, and RP/M.

[Yilmaz1001]

I haven’t spaded any German Panther other than the Panther D, so they’re not included in my dataset for final recommendations, only as part of the trendline. It will spit out an equivalent recommendation for an equivalent Panther in the German tree. Had you made fewer assumptions instead of getting indignant, you could’ve asked that and found that out. Relative to the Dauphiné, the Panther D suffers with turret slew rate, which makes my model spit out only a 6.3 for it. Idk why on earth you started acting like an overemotional child and assumed I would be treating French vehicles capriciously.

[MotorolaCRO]

Will Gaijin ever reconsider changing 1.0 BR MM spread to something lower ?
As it stands, 1.0 BR spread is a legacy system from a decade ago when WT had much less active players, so they needed to make sure queue times aren’t long.

But now, nothing really stops them to at least try 0.7 BR spread to see what happens.

[ZendikarHaven]

@Yilmaz1001

T-28E is also a bit of a beast, and the m/42 EH is pretty nice if you know what you’re doing with it.

I do know that. Hell, the EH is one of my favorite tanks! However, that does not mean they need to go up. The EH’s gun has extremely anemic penetration, especially against shermans/T-34s/KV-1s.

The T-28E may have a half-decent gun, but it is a massive target and its armor is quite frankly paper.

7.3 for the panthers is quite frankly insane. That is .3 above the Tiger IIH, a vehicle which the Panther is absolutely inferior to.

[RedRockingHorse]

@Yilmaz1001

|Tiger H1|7.7|

what

[CleanupOnIL-2]

I have to assume this is on some sort of 20.0 br scale with full decompression bc otherwise these placements makes no sense

[BlueTrain]

My Hot Take:
Leave top tier BR the same except for the Pantsir.
Make the Pantsir the first BR 13.0 land Vehicle in GRB
It may help both matchmaking and put the Pantsir at a more understandable BR for its overall power.

[MotorolaCRO]

Literally makes zero sense.

[BlueTrain]

It would make sure Pantsir players only go against top tier. I think it makes lots of sense. In what world is the Pantsir a BR 12.0 when compared against other BR 12.0 SPAA.

Edit: It’s a hot take. I know many people won’t agree with it. It wouldn’t be a hot take if they did.

[MotorolaCRO]

@SoultraneMonk

In what world is the Pantsir a BR 12.0 when compared against other BR 12.0 SPAA.

In what world is the 2A7 a BR 12.0 when compared to Arietes/Merkavas ?
Leaving everything at 12.0 and only moving Pantsir up to 13.0 is ludicrous.

[BlueTrain]

If they fixed the bug reports 12.0 for those would make sense. They don’t need to re-rank them. Just fix them.

[MotorolaCRO]

Even when fixed those wouldn’t be 12.0 worthy, especially Ariete.