The basic idea for this algo is that it looks at companies that are fundamentally weak: they are not profitable, they have terrible returns on equity, and they are small (less than $50m) market caps.
It then waits until they spike -- their returns are in the top 5% of returns for all stocks over the last 10 days -- and then it shorts them.
The algo, as-is, is quite brittle and easy to break. It's not hard to find a time period during which the returns look far worse.
I'm turning the concept over to you guys to see if we can make something of it.