Tri Count estimate

Hey there! As you've guessed, that's always a tricky question simply for the fact that it heavily relies on the project that it's going into!

Some factors to consider are:

• How close can the player see this asset?
• Does this asset have geometry deforming animations?
• Does the level/game have a poly-budget?

The nice thing is, most game engines and computer hardware have gotten better at crunching polygons, so being a couple polygons heavy isn't as significant as it has been in the past (don't' get me started on Unreal's Nanite system tho). You are more likely to get nailed on other aspects like texture streaming (such as texture sizes and amount of textures).

In terms of an ideal poly count for this asset, I'll answer that based on a couple generic scenarios:

• First-person camera, the wand is directly in the players face most of the time (5-10k triangles, depending on size and complexity)
• Third person over-the-should camera (1-5k triangles)
• Top down RTS camera (100? Not very many at all haha)

Like you've said, these are rough, generic ballpark numbers and don't take into consideration requirements of a specific project!

All assets are different in their priority as well! You may want to allocate more topology to small hero assets (weapons, cut-scene props) that has a lot more focus, versus a large background prop (barrel, box), since the topology is rather simple and you may need to have lots of those on screen at a given time, meaning you want to save on topology over a bespoke item that only has one instance in the scene. A "bang-for-your-buck" situation.

Something to consider as well, is that depending on the asset you may be able to rely on LODs (Levels of Detail). Assuming the asset isn't a weapon that is in the players face the whole time, you can get away with some higher poly counts since it will be a low resolution version of the model most times as well. A very handy technique, for sure, but also not a reason to submit un-optimized models haha. 

The studio you work at will often indicate a budget for you too, so this isn't always something you need to figure out yourself. Sometimes you'll have a poly count that you are well under, but need to add more polygons to your model for certain reasons, such as ensuring there is consistent topology density for vertex shader painting. 

As you can see, there's lots of different answers you could have while trying to prepare a model for a game. Some poly counts that work on one project may not work on others or even across studios. My rule of thumb is simply to reduce as much as I can while preserving a "silhouette read". If I've removed an edge loop that fundamentally changes the look of the model from the distance the player will be looking at it, I revert and keep that loop. It becomes a bit of a feeling thing you will develop over time, but as long as you aren't exporting with Subdivision Surface modifiers still applied, you should be in a good place. :)