When I ran the first test of a naive bot, that simply voted up whatever a list of voters voted up, I wasn't expecting great results. Changing things to vote up what certain whales vote up yields much better results, but not much better than mining. So far, my bots are failing pretty badly.
Although running the bot is far less CPU intensive, so technically it's cheaper to run a bot that follows whales votes, but if you vote up everything the whales vote up, you run out of voting power. Over a 24 hour period I can make about 1.5 SP, but then have to rest the account to regain voting power. If I only voted when the voting power was above a threshold, that would help somewhat, but only running some tests to see if the yield of curation rewards goes up will prove one way or another if it's viable. So, there's more to test there. If I had more Steem Power, then I think the naive whale following vote bot would actually yield a pretty decent return. Definitely more than mining.
Of course, when things change, like reputation rules, the bots have to get smarter. The naive bots can't make a ton of money. I am not sure if any bot can make very much. Other than Wang. Posting content and comments makes you a LOT more money than a naive bot will, that is for sure.
So, I guess I'm thinking about what to do next. I've been working on Steemwhistle (https://steemit.com/steem/@auxon/steemwhistle-a-net-library-that-exposes-the-steem-blockchain-as-reactive-extensions-observables), a reactive .NET library that observes the Steem blockchain and allows you to easily write asynchronous queries over the blockchain, and have made a lot of improvements and additions to it since I first posted about it. I've been testing my new bot experiments with it and things are looking pretty good, I think. I should have an update pushed to Github in a few days. It can be used for a lot more than bots, however, so I'm thinking about some other experiments.