Now that live-trading is gone, I have a bunch of algos that don't have any purpose any longer.
I briefly live-traded this algo before it crapped out my Robinhood account due to the order history becoming unmanageably huge, which in turn was crashing the RH-Q sync process. Real-life results weren't as great as in the backtest. (The backtest result is quite a bit inflated due to a bug I discovered in Q.) But it nonetheless looked promising. Backtests over other time periods (2008, etc.) look the same, but since it places so many trades the Q backtester would always poop out before completing a longer backtest.
At its heart it's a simple long-only algorithm that uses a SMA as a signal -- nothing special there, and definitely lots of room for improvement -- but I think the ingenuity had to do with timing entry and exits and dynamically managing risk and position size -- and I think those concepts could potentially be adapted to improve performance of a robust institutional-level trading signal. As is though this algorithm doesn't work at above $1m, nor is it really feasible on another platform (nor did it work on Q for more than a couple weeks), since it places hundreds and hundreds of small bets.
So... a dead end with a beautiful chart, and some lessons learned.