Every website is different, and every customer segment is different. You aren't going to find any rock solid data that will give you the best captcha for your own site. The best way will be to split test a few different ideas and optimise for the best conversion rate for your product.
A honeypot field is the least intrusive to your user but it generally falls victim to autocomplete forms and screen readers. A honeypot captcha should slide past the form autocomplete scripts but may still cause problems with users using screen readers.
A simple checkbox (like the reCAPTCHA 'I am not a robot' above) is the next least amount of invested effort a user will have to do. You mentioned that it did not work on your site, but verify if that is because of the concept itself or just reCAPTCHA's implementation. You could roll your own that tests for the user physically clicking on the button that should work well.
A regular captcha is usually effective, but can come at the cost of user conversions. In order to make this less intrusive, you could hide the captcha for trusted users. These would include first time users and users finding the site via search or a reputable link.
With effective split testing, you can find out which captcha converts the most users, when and why you need a captcha, and even if the captcha is the reason your users aren't converting in the first place. As Steve Blank says, no facts exist inside the building; only opinions -- and the only way to know for sure is to test.