As requested, here is a tear sheet with round_trips = True.

You're using limit orders on low dollar volume stocks. In reality you won't get filled at those prices. Part of the issue is a bug in Quantopian's default slippage model, and part of the issue is that HFT firms will front-run your orders, place orders $0.01 better than yours, etc. Don't expect to get filled on low dollar volume stocks until the spread has crossed your limit price. Stocks on the move can have very wide spreads. In addition, many of these stocks many not actually be available to short, or the fees would be so high that the strategy would not be profitable.

For comparison, here's a long strategy of mine with similar stats. It only lost money in real trading.

@Viridian. Thanks for the info. There does appear to be some alpha there despite the potential pitfalls, so I'm going to keep on investigating the types of problems you've mentioned to see if there are any solutions. Never surrender.

I would also like to add that from my own observations, it's not very difficult to make a seemingly profitable short algo, even using market orders, that trades low dollar volume stocks. However in real trading, the spreads will be larger, and the borrow fees will be expensive -- that's if the stock is even available to short.

Just don't get too wrapped up in simulations. Nothing beats a good reality check -- trading with real money.

I can make over 50% profit a day on a simulator, but trading real money is 100000 times harder. Limit orders in the backtest are not realistic because you are the first in queue to excute the trade. In the real world, you are either behind several orders or you get partially filled. The market reacts to your limit orders so simulating on Quantopian is not realistic for intraday trading. The only way to test for sure is trade with real money in real time. My experience is the bigger the alpha you find the faster alpha disappears.