1. Random photos are better than screenshots
Everything is better than straight screenshots. You didn't show what you were using before, but at a dismal 5% conversion I am guessing it was straight screenshots that weren't "posed" on a device or anything? Photos that have relevant content and even a minor emotional effect will say more than your screenshots ever will. I don't think I would call a football and baseball image "random" on a sports site, though. They are more relevant to a sports fan than a screenshot.
2. Large Background Photos Convert Really Well
That is not universally true. The two screenshots you have provided demonstrate the effect of focus in a design, not the effect of background images. You have used a background that is not only relevant, but is high-contrast colors and blurred, which creates a "foreground" out of your form. That will make the user's eyes go directly to the form rather than getting lost in your boring headlines, etc.
3. Inciting curiosity works better than clarity
It's interesting that you interpreted this as "curiosity" rather than benefits vs. features. Your original headline was a classic mistake. "A social network for sports fans" is a useless statement. No benefit, no motivation, no reason to click. "Follow sports together" is inherently more appealing because it describes the benefit of the service and has a social motivation built in. THAT is what you have done. Curiosity is not relevant.
4. Focus on the value proposition for brand new users, not advanced users
Duh. I don't want to be disrespectful, but this is a pure numbers rationale. In any market there will always be a small minority of advanced users compared to the vast majority of basic users. This will be true for every service, ever, unless the only people that could possibly understand the offer are advanced users.
5. Allowing users to sign up with their email increases FB/Twitter sign-ups
This was a good one. It is usually less effective to give people an ultimatum (register or leave) than to give them options to compare (would you rather register with FB or email?). Users will focus on the choice rather than the "yes or no" if you give them useful options to choose from. I once read an example test where more people were willing to pay for an a set of partially broken dishes if they were compared to an incomplete set than if they were sold alone. Same principle.
6. Figure out and optimize for the target demographics that convert the best
In the age of analytics, it is futile to define target audiences before statistics are available. It is much better to do as you have done: launch it and let your target audience identify itself. Good lesson.