I also doubt any of the developers were aiming for that too.

I've been looking through the decompiled code for the purposes of modding for the last few days, wrapping my head around their ECS/DOTS code, and the game has a much better foundation for a scalable simulation than CS1 ever had, even after years of optimizations.

What probably happened is that certain dev teams built ridiculous things on top of the ECS/DOTS code to see how much they could push the detail. So you wound up with things like the teeth. Which was probably some internal demo-driven development that everyone 'wow'd at the fact that it could be done. Times 100x or something like that. With nobody owning overall performance or whoever owned the overall performance didn't have any management support before launch because it was deprioritized. Nobody gets promoted for ripping out someone else's perf-expensive features and being the bad guy, you get promoted for building new clever features on top of the shiny new engine.

> So you wound up with things like the teeth.

Author here. The teeth are completely unrelated to the simulation and not even really related to pushing maximum graphical fidelity. They are just using completely unoptimized - possibly stock - character models, which include a mouth with teeth even thought the characters never open their mouths.

The teeth certainly aren't helping anything, and something has to make a decision not to render them somewhere, and they're wasting disk and RAM if nothing else.

Also think that's a distraction from my point, so just pretend I wrote "100,000 vertex logpiles" instead of teeth.

The teeth are just the most obviously useless unoptimized thing they shipped that nobody had the time/will to cleanup before launch.

But they are also using LOD... as per your article [0]. So it's quite possible (you have the source, go look) that they call this lib with the vertex buffer of the mesh in question and get back an LOD3 mesh that's decimated and simplified but with the projected UV's. This is kinda how Unity wants folks to do "AAA" in Unity. Use a high resolution model and decimate LOD's in real-time like Unreal Engine does. Only, it doesn't work half the time. It requires that the original high poly mesh be properly UV wrapped and not vertex shaded. So to further throw gas on the fire. It's entirely possible instaLOD is the source of a lot of these issues. Combined with what you found in the rendering stages.

[0] https://instalod.com/

InstaLOD is only used in the asset pipeline. I haven't seen any references to real-time use. Haven't found any evidence of real-time decimation either.

To be clear there are LODs, but only for some of the meshes.

That’s cool. Didn’t realise modders still do that sort of low level stuff

Considering Paradox/CO are dragging their feet to release the modding docs/tools, we don't have much choice if we wanna start building our mods today :)

I bet whoever worked on the actual simulation worked really hard for years to make it run efficiently.

Then the meme developers in the front end went “lmao JS and CSS for the interface, 100k for a log” and ruined it all

Not sure how the "frontend" (UI) developers have anything to do with the assets, but it's a easy scapegoat I guess...

No, we can have both things. Graphics are incredibly important to a game, especially for a simulator like this, it really helps with the immersion. It’s as important as gameplay.

> It’s as important as gameplay.

Important certainly but colour me unconvinced on equally important. If the game play isn’t fun on a simulator then it isn’t much of a game.

People still play chess despite terrible graphics

I wouldn't even care if the graphics were exactly the same as in the first game.

It's all about the gameplay.