Rework Performance Ranks into a Point System
Lightning Test
The Problem: Currently the ranking system encourages you to max out everything at your rank. You're "Poor" because of a single Light? Might as well add 16 Skinned Meshes, 16 Meshes, and 32 Materials. Same rank of "Poor" either way. You're "Medium" because of 3 Skinned Meshes? Might as well add 5 more, and 8 normal meshes too. Same rank of "Medium" either way. The actual difference is an approximate 533% increase in draw calls, but you get the same rank either way! That needs improving!
NOTE: We still need a system to properly account for hidden objects (or one that ignores them)! This rework only targets problems with the Ranking system for active components.
Here's a (rough draft) alternative:
- Every avatar gets 100 Points
- Points are taken off for components
- Safety Setting can block by precise score (0-100)
Excellent = 90 Points (96 on Quest)
Good = 80 Points (92 on Quest)
Medium = 60 Points (88 on Quest)
Poor = 40 Points and up (84 on Quest)
39 and lower is "Very Poor"
Animator = 1 Point (counting base animator -- everyone loses at least 1 Point)
Trail/Line Renderer = 1 Point
Mesh = 1 Point
Skinned Mesh = 2 Points
Material Slot = 1 Point
Collider = 4 Points
Particle System = 2 Points
Audio Component = 2 Points
Dynamic Bone Component = 2 Points
Dynamic Bone Transform = .5 Points (1 per 2)
Particle Trails On = 20 Points
Particle Collision On = 20 Points
Particles Active = .01 Points (1 per 100)
Light = 40 Points
Cloth = 5 Points
Rigidbodies = 5 Points
Cloth Vertices = .1 Points (1 per 10)
Every 75 Bones = 1 Point
Every 5000 Poly = 2 Points
Bounds Size = Shown, but not included in point calculation. Sometimes everyone is actually on-screen, it's good to plan for the worst.
---Benefits for optimization---
This more accurately assesses the impact of components on players. If someone only has 3 Skinned Meshes and 3 Materials, they will likely be "Good" or "Excellent" (depending on poly count). This reflects the actual impact they have. Someone with 8 Skinned Meshes (16) and 8 Material Slots (8) is pushed to "Medium" or lower where they should be. It still needs refining, but I could give a dozen examples where this better reflects performance.
---Benefits for creators---
We won't be pushed to avoid certain components for no reason. On Quest, you could have 1 Skinned Mesh, 1 Trail Renderer, and 2 Material Slots (1 on each of those components) without getting marked as "Poor" or hidden by default. On PC, you could have an "Excellent" avatar with limited Dynamic Bone components now, as long as you use very few, and your poly count and other stats are low enough not to push you to "Good". People who want to have even higher than 70K Poly counts can do so now too, as long as materials and other components are minimal.
And again, this does not necessarily apply to hidden objects. We still need a system accounting for activity of components if possible.
Log In
Silhouette
Every year, suggestions like this about the Avatar Stats go ignored by VRC's devs and staff members and the Avatar Ranking Weights get more lopsided and more convoluted.
Every particle counts as a separate material, now, even if its just the default particle. You can have one material, but still somehow count as 10 because a few disabled objects have slots using that same one material on them.
I want to believe suggestions like these will be received, but we've been asking for it since the Avatar Stats were added. I've given up hope that they'll suddenly start caring about the game, now.
Observer.
What you have effectively outlined here is a weighted performance rank which is something we have asked from the very beginning of the initial system implementation.