BR Deviation Assessment of the T-72/90 and T-80 Families Based on Linear Modeling of Modifications

Model Objective

Build a more reasonable Modification-BR system withoutu fitting current BR values. Use two family baselines (T-72/90 and T-80), compute curve weights, then compare to current BR.

Model Method

  • Form: linear hyperplane (A/B contributions summed linearly).
  • Constraints: A_i >= B_i >= B_min.
  • Rank order (soft): Thermal >= Engine >= Reactive Armor >= Dual Control >= Ammunition >= Active Protection.
  • Optimization: least squares + L2 regularization + soft ranking penalty.
  • Grouping: Chinese vehicles in T-72/90, Ukrainian vehicles in T-80.

Modification Tier Definitions

  • A: Gen2+ thermal imaging, 1200 hp engine with reverse speed >= 10 km/h, 3BM60 or equivalent ammo, full frontal Relikt + double-layer side Relikt, Arena APS.
  • B: Gen1 thermal imaging, 1000 hp engine, 3BM46 or equivalent ammo, side Relikt, Drozd APS.
  • C: No thermal imaging, 850 hp engine, 3BM42 or equivalent ammo, no Relikt, no APS.

Math Formulation

y_hat_k = alpha_T72*f_{k,T72} + alpha_T80*f_{k,T80}
        + sum_i (a_{k,i}*A_i + b_{k,i}*B_i)

min  sum_k (y_k - y_hat_k)^2
   + lambda * sum_i[(A_i-A_i0)^2 + (B_i-B_i0)^2]
   + mu * sum_i max(0, A_{i+1}-A_i)^2

constraints: A_i >= B_i >= B_min
rank: Thermal >= Engine >= Reactive >= Dual Rotation >= Ammo >= APS

Priors (default used here)

  • A prior: Thermal 0.60 > Engine 0.50 > Reactive 0.40 > Dual 0.35 > Ammo 0.30 > Active 0.25
  • B prior: B0 = 0.6 * A0
  • Floor: B_min = 0.05
  • L2 strength: lambda = 5
  • Rank penalty: mu = 100
  • T-72A penalty: = 0.3 (only apply to T-72AV)

Tuning Options

  • lambda (L2 strength): higher = closer to priors, lower = closer to data.
  • mu (rank penalty): higher = stricter rank order, lower = more flexible.
  • B_min: higher = larger minimum B.
  • A/B priors: replace with your own expert values.

Verdict Thresholds

  • |diff| <= 0.15: Reasonable
  • 0.15-0.30: Slight downrate / Slight uprate
  • 0.30-0.70: Downrate / Uprate
  • 0.70: Heavy downrate / Heavy uprate

Baselines

  • T-72/90 baseline: 9.99
  • T-80 baseline: 10.48

Upgrade Weights

Upgrade Item A Weight B Weight
Thermal 0.58 0.32
Engine 0.58 0.21
Reactive Armor 0.46 0.12
Turret Rotation Speed (Horizontal & Vertical) 0.35 0.21
Ammunition 0.33 0.21
Active Protection 0.27 0.2

Results

Vehicle Family Curve Weight Current BR Diff (Current-Curve) Verdict
T-72B T-72/90 9.99 10.3 0.31 Downrate
T-72B 1989 T-72/90 9.99 10.3 0.31 Downrate
T-72AV T-72/90 10.27 10.3 0.03 Reasonable
T-90A T-72/90 11.12 11.0 -0.12 Reasonable
T-72B3Ae T-72/90 11.53 11.0 -0.53 Uprate
T-72B3M T-72/90 11.96 11.3 -0.66 Uprate
T-72B3A T-72/90 12.24 12.3 0.06 Reasonable
T-90M T-72/90 12.3 12.7 0.4 Downrate
ZTZ96A T-72/90 10.57 10.3 -0.27 Slight uprate
ZTZ992/3 T-72/90 11.49 11.0 -0.49 Uprate
MBT2000 T-72/90 11.49 11.7 0.21 Slight downrate
VT4 T-72/90 11.84 12.3 0.46 Downrate
VT4A1 T-72/90 12.11 12.7 0.59 Downrate
T-80B T-80 11.01 10.7 -0.31 Uprate
T-80U T-80 11.59 11.7 0.11 Reasonable
T-80UM2 T-80 11.47 11.7 0.23 Slight downrate
T-80UE1 T-80 12.06 12.0 -0.06 Reasonable
T-80BVM T-80 12.79 12.7 -0.09 Reasonable
T-80UDBE T-80 11.06 11.0 -0.06 Reasonable
T-84 T-80 11.97 12.3 0.33 Downrate

Modifaction Matrix

Vehicle Family Thermal Engine Ammunition Dual Rotation Reactive Armor Active Protection
T-72B T-72/90 C C C C C C
T-72B 1989 T-72/90 C C C C C C
T-72AV T-72/90 A C C C C C
T-90A T-72/90 A B A C C C
T-72B3Ae T-72/90 A C A A C A
T-72B3M T-72/90 A A A A B C
T-72B3A T-72/90 A A A A B A
T-90M T-72/90 A A A A A C
ZTZ96A T-72/90 A C C C C C
ZTZ992/3 T-72/90 A A A C C C
MBT2000 T-72/90 A A A C C C
VT4 T-72/90 A A A A C C
VT4A1 T-72/90 A A A A C A
T-80B T-80 B B C C C C
T-80U T-80 B A B C C C
T-80UM2 T-80 C A B C C B
T-80UE1 T-80 A A B B C C
T-80BVM T-80 A A A A A C
T-80UDBE T-80 A C C C C C
T-84 T-80 A A A C C C

This was a pretty spontaneous idea, and I honestly just thought it would be fun to play around with.

I wanted to try fitting a hyperplane using the existing BR and modifications in the game, just to see what happens. Vehicles above the plane can be viewed as “OP”, while those below it are more on the challenging side and require more effort to make them work.

I want to be clear that this model is purely based on my personal experience and for fun. It’s not meant to suggest or influence any BR changes in the game.

The choice of priors and tuning parameters has a direct impact on the results, so discussion on those is very welcome. That said, I’m personally quite satisfied with how the current conclusions turned out.

If you’re into statistics or modeling, you’ll probably get what I’m trying to do here — and hopefully enjoy it as much as I did :D

Base BR 10.0

lets apply this to other nation tanks

Leopard 2A7V Base Leopard 2A6

Thermal: Gen 1 same as 2A6
Engine: same as 2A6
ERA: None same as 2A6
Ammunition: Same as 2A6
Active protection: None, Same as 2A6
Turret rotation: same as 2A6

result: BR should be the same as 2A6

Leopard 2a7v is gen 3 thermals

I thought you just used ChatGPT, tbh. Can’t be too sure these days.

ok then modifyer 0.58 it should be 1 stop higher

It does not make any sense as weights of modification are to be decided by the person, so BRs become personal preference