What return should we expect from a a good algorithm

Well, usually one wishes to maximize returns conditional on the circumstances. For example, one should minimize volatility when dealing with lower funding (in most cases) to avoid ruin (and thus it would be added as an additional objective). Indeed, tools like efficient frontier can thus be performed on much more than just sharpe alone. As a more concrete example, the latest competition provides some specific criterea for a "good" algorithm in its setting:

https://www.quantconnect.com/competitions/quant-league-9#rules

It depends on what you want in your algorithm. Usually, you want to maximize profits while keeping things like volatility and risk low. Or your algorithm could be intended for smooth but low gains that institutions can use with leverage. The criteria of what makes an algorithm good depend on what you want your algorithm to do, but as Aaron said Quant Leagues provide outlines for what they consider to be good. 

Hi Ehsan,

Strategies are typically evaluated by their Sharpe ratio instead of raw return since leverage can get applied to increase the raw return. To get an idea of what Sharpe ratio the top strategies on QC are generating, check out the Alpha Streams Market page or the Quant League leaderboard.

To see what we consider fit to be accepted into the market, refer to our Alpha Stream submission criteria.

Best,
Derek Melchin

As others have mentioned, looking at the Quant League/Alpha Streams leaderboards and criteria are good rule-of-thumbs. Generally you'd want to evaluate performance not just on returns alone, but the overall risk-return dynamics. Sharpe ratio (when treating "risk" as volatility) is a good measure, as is drawdown periods, maximum drawdowns, sortino ratio (when treating risk as downside volatility), etc. Scalability of the Alpha is important too, and with higher capital slippage/liquidity issues can be something to keep in mind.

If live trading on retail accounts, returns do become important in practice due to potentially limited capital and leverage regulations. In this case, you'd want to compare performance to a benchmark with similar risk exposures and consider opportunity costs as well.