I computed the ratio of good vs. bad results by adding up the number of responses to 'I tried this hack with good results, and I don't use it consistently, though I'd like to' and 'I tried this hack with good results, and now use it consistently' and dividing by the number of responses to 'I tried this hack with disappointing results'. Here are the results:

• Luke's algorithm: 0.67 [0.67 good results per bad result] (n=20)
• Pomodoro technique: 2.75 (n=45)
• Exercise: 4.86 (n=123)
• LeechBlock or similar: 0.87 (n=56)

(Please note that I haven't factored in the number of responses to 'I tried this hack with good results, but it stopped working', since it wasn't clear what weight should be given to them relative to the other two positive responses.)