"So a Hacker News reception has good predictive power, estimated around the OLS's R² = 14%."<p>14% R-squared is the measure for 'good predictive power'? Second, using linear regression for binary variables feels kind of icky to me. A suggestion for an improve or fix or maybe even another blog post is to change the model to something more appropriate like a Logit/Probit model.<p>/econometric rant<p>Cool article though!