Completely different datasets here. The research was self-reported data from surveys sent to Kickstarter "founders" across all categories. Y Combinator uses actual portfolio data for tech-based startups looking to create large-scale, $100M+ businesses. Success in the former is largely not going to predict success in the latter.