Another idea: download the bots, do a backtest, than come back 3 months later and do another backtest. If the bot is still profitable than you probably have a better chance that the bot will stay profitable for some time.
In general these bots are quite over-optimized, with a high likelyhood of failure when starting live. The fact that they now indicate which part of the equity curve shows out of sample data is also helpfull in selecting bots that have a track record in live-conditions.
Daan Terstegge