Jump to content


Balancing Soviet TD`s

SU-85 SU-85i SU-100M1 SU-101 SU-122-54 SU-152 Obj. 268

  • Please log in to reply
102 replies to this topic

Karukute #41 Posted 09 December 2014 - 08:25 PM

    Warrant Officer

  • Player
  • 6145 battles
  • 558
  • Member since:
    07-12-2012

Scratch that, WN8 is heavily influenced by damage done, and is therefore skewed.


Edited by Karukute, 09 December 2014 - 11:35 PM.


BorisZTZ99 #42 Posted 10 December 2014 - 12:39 AM

    Private

  • Beta Tester
  • 14156 battles
  • 14
  • Member since:
    11-17-2010
Plz refer to this 704 video.

Karukute #43 Posted 10 December 2014 - 12:47 AM

    Warrant Officer

  • Player
  • 6145 battles
  • 558
  • Member since:
    07-12-2012

View PostBorisZTZ99, on 10 December 2014 - 12:39 AM, said:

Plz refer to this 704 video.

 

So you posted a video of the 704 performing well, when the OP has nothing to do with the 704.



String_Emil #44 Posted 10 December 2014 - 12:05 PM

    Corporal

  • Player
  • 29355 battles
  • 169
  • Member since:
    12-23-2012

Below some charts showing the correlation based statistics of some Tier 5,7,8,9 TDs. For those who are not familiar with that metric, it shows the winrate of a vehicle in function of players winrate. A vehicle is considered to be balanced, if the winrate of the vehicle is approximately equal to the winrate of the players.

 

Tier 5 TDs: T67, S-35CA, SU-85, SU-85I

Conclusions:

  • T-67 and S-35CA can be considered OP
  • SU-85I is balanced
  • SU-85 is not balanced and is not OP

 

 

Tier 7 TDs: AT7, SU-122-44, SU-152, SU-100m1

Conclusions:

  • AT7 is OP
  • SU-122-44 is slightly OP
  • SU-100m1 is balanced
  • SU-152 is not balanced and is not OP

 

 

Tier 8 TDs: AT15, Rhm-Borsig, ISU-152, SU-101

Conclusions:

  • AT15 and Rhm-Borsig are balanced
  • ISU-152 slightly underperforms
  • SU-101 definitely underperforms

 

 

Tier 9 TDs: Tortoise, WT auf Pz. IV, Object 704, SU-122-54

Conclusions:

  • Tortoise, WT auf Pz. IV, and Object 704 are balanced
  • SU-122-54 definitely underperforms

 

 

Source: http://wot-news.com/.../tankinfo/en/eu (images taken a few days ago)

 

 

Edit: my personal opinions on those TDs i played so far.

Tier 5: i liked both T-67 and S-35CA. SU-85 i cannot really remember. As for SU-85I i definitely don't feel it like being OP.

Tier 7: i found SU-100m1 playable, but nothing special, definitely not OP. SU-152 was also ok to play, i preferred SU-100m1 over SU-152.

Tier 8: i liked the ISU-152, but SU-101 was a nightmare. I found Rhm OP, i preferred it over ISU. 

Tier 9: i liked both the WT auf Pz. IV and the Object 704, i preferred the former one. I also liked T30. SU-122-54 is definitely more difficult to handle, i prefer it over the Foch however.

 


Edited by String_Emil, 10 December 2014 - 12:23 PM.


CaravanOfCourage #45 Posted 10 December 2014 - 12:21 PM

    Warrant Officer

  • Player
  • 47188 battles
  • 881
  • Member since:
    07-03-2011
There is nothing wrong SU-152 lol. You can Derp every 14-15seconds or use the 122mm that has about 4000dpm lol. Frontal armor will bounce some shots but shouldn't be relied upon. Didn't bother buying the ISU because SU was too much fun.

Karukute #46 Posted 10 December 2014 - 05:37 PM

    Warrant Officer

  • Player
  • 6145 battles
  • 558
  • Member since:
    07-12-2012
String_Emil I thank you for finding the data that I couldn't.

w_o_t_q #47 Posted 11 December 2014 - 02:57 PM

    Warrant Officer

  • Player
  • 22213 battles
  • 680
  • [BRT_6] BRT_6
  • Member since:
    11-16-2013
Problems with this stats is not correct data collection. Small WOT "feature" if playing till first victory any tank % will be higher ... So this ruin all this statistics. Total crap no sense at all.

String_Emil #48 Posted 11 December 2014 - 03:22 PM

    Corporal

  • Player
  • 29355 battles
  • 169
  • Member since:
    12-23-2012

View Postw_o_t_q, on 11 December 2014 - 02:57 PM, said:

Problems with this stats is not correct data collection. Small WOT "feature" if playing till first victory any tank % will be higher ... So this ruin all this statistics. Total crap no sense at all.

 

First, statistics are not perfect thats true, but still better than your persional opinions imho (which is rather subjective).

Second, since its computed from large number of samples i think such irregularities are compensated adequately (law of large numbers). Tho i would be interested in the confidence intervals of the statistics.

 

 



_Grim_ #49 Posted 11 December 2014 - 03:55 PM

    Second Lieutenant

  • Player
  • 22866 battles
  • 1,431
  • [-DFA-] -DFA-
  • Member since:
    04-13-2011

View Postw_o_t_q, on 11 December 2014 - 03:57 PM, said:

Problems with this stats is not correct data collection. Small WOT "feature" if playing till first victory any tank % will be higher ... So this ruin all this statistics. Total crap no sense at all.

 

So I guess you made up BS is supposed to make much more sense :facepalm:,

Karukute #50 Posted 11 December 2014 - 03:58 PM

    Warrant Officer

  • Player
  • 6145 battles
  • 558
  • Member since:
    07-12-2012

Much like Sting_Emil said, the data is based on a large data set, and is most likely weighted based on the number of battles in the tank (which would be the correct way to calculate this data). Look up a weighted average w_o_t_q, learn from it.

 

Especially since the data set will be massive, which means the data is a pretty accurate representation



BigFellas_Dad #51 Posted 11 December 2014 - 04:22 PM

    Lieutenant

  • Player
  • 41455 battles
  • 1,610
  • Member since:
    06-19-2013

View Postdavethejackal, on 08 December 2014 - 01:03 PM, said:

Wait the ISU needs a buff? Seriously? Why are you not using the BL10 with HE? One shots many tanks and leaves all heavily wounded right up to tier 10.

 

BL-10 with HE?  No thanks, just use AP.

w_o_t_q #52 Posted 11 December 2014 - 05:24 PM

    Warrant Officer

  • Player
  • 22213 battles
  • 680
  • [BRT_6] BRT_6
  • Member since:
    11-16-2013

View PostString_Emil, on 11 December 2014 - 03:22 PM, said:

 

First, statistics are not perfect thats true, but still better than your persional opinions imho (which is rather subjective).

Second, since its computed from large number of samples i think such irregularities are compensated adequately (law of large numbers). Tho i would be interested in the confidence intervals of the statistics.

 

 

 

It's feature is used not be me only .... Working ... Not personal opinion sadly.  Not correct statistics = means not stat at all ... 

Karukute #53 Posted 11 December 2014 - 06:03 PM

    Warrant Officer

  • Player
  • 6145 battles
  • 558
  • Member since:
    07-12-2012

View Postw_o_t_q, on 11 December 2014 - 05:24 PM, said:

 

It's feature is used not be me only .... Working ... Not personal opinion sadly.  Not correct statistics = means not stat at all ... 

 

You saying that "Russian tanks are OP" is personal opinion, that's all you ever say, not actually playing the tank in a realistic environment also means your views are completely subjective.

 

The data sets don't lie, especially when the data set is pretty much pulled from the API of players, so that means ALL of them. And then comparing the players winrate with their winrate in the tank shows you how well they perform in the tank relative to their overall "skill". Hence why the data is much more reliable than say, you.



Tonyb1968 #54 Posted 11 December 2014 - 06:07 PM

    Second Lieutenant

  • Player
  • 27420 battles
  • 1,422
  • [TEC] TEC
  • Member since:
    08-13-2011

View Postw_o_t_q, on 11 December 2014 - 05:24 PM, said:

 

It's feature is used not be me only .... Working ... Not personal opinion sadly.  Not correct statistics = means not stat at all ... 

 

Ok so you want personal opinions on these tanks then?

Considering you stopped with the SU-100 and I didnt, lets see what we can say about them.

 

SU-85 - view range struggles, its 30m less than any of the other t5 td's, armour is ok but not excellent, it wont bounce most T5 guns, definately not T6 or T7 guns which go through it easily.

SU-152, long aim/reload with the ML-20 gun but it is the best gun for the tank as it is effective at all tiers. Its armour sucks, T3's can penetrate it frontally at close range, its sluggish but the camo rating isnt too bad.

SU-100M1, the gun is ok for its tier, struggles outside of that due to having to fight at range because of its armour/hp/lack of gun depression, its quite accurate and shell cost is good but its far from excellent and the armour isnt to be relied on because having to shove your tank over a ridge flattens it making it easy to penetrate. It does under perform because in T9 games its useless unless you spam premium ammo.

SU-101. Its not that bad but the gun... realistically it needs a variant of the T-62A's gun and not the T-54's, shell cost is expensive but repair cost isnt, again maps dont suit this line of TD because of the lack of gun depression and low HP, armour is ok at range but not effective close up against the high pen guns at this tier, another shoot n scoot td like the SU-100M1 & SU-122-54.

ISU, needs a mobility buff to compensate the recent physics changes.

SU-122-54, has good view range of 390m, guns not bad but not good on a td that has to fight from range, mobility is good, gun depression sucks, gun arc sucks, armour sucks, it get ammo racked if hit in the side and can easily be 1 shot my quite a few T9/10 tanks. Relying on mobility isnt enough in this game any more unfortunately, especially when the maps dont suit 90% of this td's game play.

Obj268, gun arc suffers, can live with the long flight time on the shells as you just need to compensate for it, gun isnt as good as it is on paper and the armour isnt good for a T10 TD considering its based on an IS-8 heavy tanks a tier below it.

 

Now putting this into perspective, I have played more games in my SU-85 than you have in your entire soviet td line, in total I have played around 8800 games on soviet TD's, the one thing you learn is that the armour isnt to be relied on, stealth is, range use to be your friend but the maps killed that, fragile tanks like virtually all of them except the Obj704 (which theoretically has better armour than the 268) and the Obj263 (which is actually nice to play but an arty magnet with no gun depression) are the only ones that dont need anything doing to them.

 

Playing a couple of games on the test server does not give you a good opinion on these td's and neither does just "facing them", you have to live with them in a team and see how they perform, normally not that well as most are not team carrying tanks.

 



_Grim_ #55 Posted 11 December 2014 - 06:51 PM

    Second Lieutenant

  • Player
  • 22866 battles
  • 1,431
  • [-DFA-] -DFA-
  • Member since:
    04-13-2011

View Postw_o_t_q, on 11 December 2014 - 06:24 PM, said:

 

It's feature is used not be me only .... Working ... Not personal opinion sadly.  Not correct statistics = means not stat at all ... 

 

Seriously-can you even read ,cause there is no other explanation for the BS you are throwing at us.

w_o_t_q #56 Posted 11 December 2014 - 06:58 PM

    Warrant Officer

  • Player
  • 22213 battles
  • 680
  • [BRT_6] BRT_6
  • Member since:
    11-16-2013
Right now priority is getting tier X 110E5, M48 and WT E-100. No time for low tier TD even such total OP SU-100 ...:(

RichardNixon #57 Posted 11 December 2014 - 07:29 PM

    Captain

  • Player
  • 4395 battles
  • 2,351
  • Member since:
    06-26-2013

View PostString_Emil, on 11 December 2014 - 03:22 PM, said:

First, statistics are not perfect thats true, but still better than your persional opinions imho (which is rather subjective).

Second, since its computed from large number of samples i think such irregularities are compensated adequately (law of large numbers). Tho i would be interested in the confidence intervals of the statistics.

 

The confidence interval isn't a problem (at least for the middle chunk of the curve - the rest is mostly garbage) as their data sample is huge. However, there are several systematic errors:

 

1. Recency bias. Tanks typically played later in a career will be played better.

2. Winrate bias. Overall winrate is heavily dependent on tanks played, especially average tier.

3. Stock bias: Tanks more likely to be played stock or with poor crew will be underrated.

 

The one-month version fixes the recency bias (which is the dominant feature of the whole-account version), but not the winrate bias or stock bias. I've been working on a method that attempts to fix the remaining biases for WN9's expected values. Here's what I have so far:

 

https://docs.google.com/spreadsheets/d/1YddqVI-zs-OUsl4mDEYVAHmfFyoncLPJTiHh7wyv9Aw/edit?usp=sharing

 

The "exp win" column there is the winrate that an ~average tier 10 player will achieve in that tank, when played elite. The "winslope" column describes how that increases with skill.

 

 



Jetser #58 Posted 11 December 2014 - 09:06 PM

    Staff Sergeant

  • Player
  • 15354 battles
  • 491
  • [NOFUN] NOFUN
  • Member since:
    08-26-2011

View PostString_Emil, on 10 December 2014 - 12:05 PM, said:

Below some charts showing the correlation based statistics of some Tier 5,7,8,9 TDs. For those who are not familiar with that metric, it shows the winrate of a vehicle in function of players winrate. A vehicle is considered to be balanced, if the winrate of the vehicle is approximately equal to the winrate of the players.

 

Do you got the tier 10 one?

String_Emil #59 Posted 11 December 2014 - 09:47 PM

    Corporal

  • Player
  • 29355 battles
  • 169
  • Member since:
    12-23-2012

View PostRichardNixon, on 11 December 2014 - 07:29 PM, said:

 

The confidence interval isn't a problem (at least for the middle chunk of the curve - the rest is mostly garbage) as their data sample is huge. However, there are several systematic errors:

 

1. Recency bias. Tanks typically played later in a career will be played better.

2. Winrate bias. Overall winrate is heavily dependent on tanks played, especially average tier.

3. Stock bias: Tanks more likely to be played stock or with poor crew will be underrated.

 

The one-month version fixes the recency bias (which is the dominant feature of the whole-account version), but not the winrate bias or stock bias. I've been working on a method that attempts to fix the remaining biases for WN9's expected values. Here's what I have so far:

 

https://docs.google.com/spreadsheets/d/1YddqVI-zs-OUsl4mDEYVAHmfFyoncLPJTiHh7wyv9Aw/edit?usp=sharing

 

The "exp win" column there is the winrate that an ~average tier 10 player will achieve in that tank, when played elite. The "winslope" column describes how that increases with skill.

 

 

 

Cant you just simply assume that some people play a given tank earlier, while some play it later? That would solve recency bias. The same can be assumed in the case of stock bias, or not? Tho I dont know if such assumptions are valid.

The effect of Winrate bias might be cancelled by normalizing the winrate of a player per vehicle using your expected winrate valueof the vehicle. Do you think that would solve the issue?

 



String_Emil #60 Posted 11 December 2014 - 09:50 PM

    Corporal

  • Player
  • 29355 battles
  • 169
  • Member since:
    12-23-2012

View PostJetser, on 11 December 2014 - 09:06 PM, said:

 

Do you got the tier 10 one?

 

you can make one at http://wot-news.com/...tankinfo/en/eu 

its really just a few clicks






1 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users