Late to the thread, chiming in - and joining the choir: Low correlation to the current robots (I was glad to see GDX and not SPY as the second one), enough volume so as not to see high spread, and underlying stock tickers making it fit to the methodology.
Best if we have several robots where the equity curves are not highly correlated, and the best way to get there before the robots are completed is to look at low-correlated underlying ETFs.
Concurring with earlier posts, probably XOP (Oil & Gas) should be next. And if we could see more in the future ( :-) ) - XLV (Healthcare), XLP (Staples), and XHB (Builders) might be contenders.

I place lower importance on the availability of x2 and x3 ETFs for the specific index or industry group. With the way margin requirements for leveraged ETFs are calculated now seems to me there is no advantage to trading those vs. creating leverage by margin (except in the context of non marginable retirement account). Also I am not sure how the peculiarities created by the daily reset of leveraged ETFs will affect robot results.

Pascal, Billy - thanks for your great work, and keeping us all in the loop.