Should You Record Failed Trading Experiments?

An unsuccessful experiment may still reveal something useful about an assumption or market condition. Do you preserve failed ideas and their results, or do you mainly document methods that worked?