Table Stakes · Ranking Engine
A weighted composite of crowd ratings and editorial awards. Crowd scores are shrunk toward the city mean so a 5.0 with eleven reviews cannot outrank a 4.6 with three thousand. Every weight below is yours to move.
score = 100 × (wcrowd·crowd + wed·editorial + wcons·consensus) × (1 − penalties). Components are 0–1 before weighting and the weights are normalised, so only their ratio matters. Crowd uses Bayesian shrinkage toward the city mean. Editorial credit saturates as 1 − e−s·credits, so five weak listicles never out-vote one serious award. Settings persist in this browser.
Ratings and review counts are supplied per source and each chip links back to that platform's own listing: Yelp, Tripadvisor, and Places data powered by Google. Tripadvisor bubble ratings are shown using Tripadvisor's own rating image. The composite score is this site's own calculation and is not endorsed by, or attributable to, any of them. Editorial signal is hand-compiled from the Michelin Guide and James Beard Foundation awards and may contain errors. Dining tier comes from price level, with a Michelin star overriding it.