Machine Learning!

[iasoon]

We could try using machine learning for scoring commits.

I think this could work out great, because the scoring problem is one of simple categorization.

• When you add 16k lines, you obviously didn't write them and should not get credit.
• When you add 500 files, you generated them. No cookies.
• When you delete something in one file and add it (or something very similar) in another file, you didn't write that either.
• Some languages (filetypes) are more verbose and could be worth less points.
• We finally have a way to ignore binary blobs ( Don't count binary files #54 !)

If this system works out well, we removed the need for bounties, which would open the way for #89 .