So, and this is just honestly some excessively complicated backtesting, if we had this new index and it was stable under different random seedings (proof that the index reflects the actual market), we could compare it with an index and see a divergence if the index was veering off because it became an echo-house.

So I'm curious now what an index created to avoid these problems might look like. I imagine we'd want an un-reproducible index. Something that cryptographically re-chooses its members constantly with our desired constraints. Sure someone does this already in some way. Even if it's just being a trusted authority instead of a cryptographic solution.

And of course it provides clean access to modern image manipulation tools. This texture synthesis alg is actually a bit too old. I've written my own, but it works here. After this I import the image into GIMP and use a cloning tool on the section that still look wrong.

Mathematica is built from the ground up for kind of messy work between formal programming and informal hand exploration. In this case, I've created a UI on the fly to adjust a parameter of an image process algorithm to detect the black stain.

I'm using Mathematica for most of this work. And what the code below illustrates for me is how important the UI is. Colors and Images are first class objects that we can manipulate both programmatically and in a manual style. It's possible to mix and develop both kinds of workflows together. In this case I'm getting the major colors to get a rough mask of the black stain.