The post "Testing on a larger scale" led me to create this thread. I'm wondering what others in the community would require of their algorithms to actually use them to invest significant sums. I'm kind of interested in specifics if possible along the lines of how many securities would an algo have to work for? Over what period? What kind of results? What amount of slippage (trading expenses) would you account for? Are you more comfortable with a broad ETF, sector ETFs? If you use individual stocks how many would you need to feel diversified? What kind of risk do you feel comfortable with? Do you lever or short? I'm trying to keep this general knowing that I come to the table with strong beliefs (actually fear and skepticism). Are you more comfortable with daily trading? Would you use minute to minute? Would you allow the computer to generate/place the trades without you? I'm not looking for anyone to answer all of these. Frankly I'm looking for some ideas. Also, this question relates to "significant sums". What I mean is what would it take for you to put enough of your capital at risk with an algorithm that a 35% drawdown would be uncomfortable?