About Five Best Places
Five Best Places is a neighborhood-by-neighborhood guide to the local spots actually worth your time — restaurants, bars, coffee shops, bakeries, and the rest of the places you turn to when you want something good.
We started this project because finding “the best” anywhere online had become genuinely hard. Big aggregators like Yelp, Eater, and Time Out optimize for breadth, not honesty — their lists are partly sponsored, partly editor-driven, and almost never specific enough. A search for “best tacos in Hollywood” returns a list of the best tacos in all of LA, half of which closed two years ago.
We wanted something more specific, more honest, and more grounded in real signal. So we built it.
How we rank
Every place on this site is scored by a Bayesian weighted formula combined with a logarithmic volume bonus:
score = (v / (v + m)) × R + (m / (v + m)) × C + γ × log₁₀(v)
Where R is the place's average rating, v is its review count, C is the mean rating across candidates, γ is a small constant rewarding review volume, and m is the area-adaptive prior weight — set to the median review count among local candidates and clamped between 100 and 2,000.
The adaptive prior is important. A “200-review” place is small in the West Village (median ~1,500) but solid in Carlsbad (median ~125). A fixed prior would over-promote big-city novelties and under-recognize suburban favorites. Using the local median fixes that automatically.
On top of the formula, a dominance override kicks in when one place has at least 3× the reviews of any other on the list (and at least 1,000 total, with a 4.2+ rating). That place is moved to #1 regardless of formula score — popularity at extreme scale is its own signal.
We also collapse chain duplicates: if multiple locations of the same brand show up, we keep only the highest-scored one and mark it as having multiple locations.
The practical effect: a 4.9-star place with 12 reviews scores lower than a 4.6-star place with 800 reviews. Tiny-sample places — where one bad day or a handful of friends-of-the-owner reviews can swing the average — get pulled toward the local average. Hundreds of slightly-imperfect ratings beat a dozen perfect ones.
The Bayesian half is the same statistical idea that powers IMDb's Top 250 movie list, adapted for restaurant ratings and tuned for local geography.
Where the data comes from
All ratings, review counts, addresses, and photos come from the Google Places API. Descriptions are written by an AI model that paraphrases and synthesizes recent review snippets — we never republish review text verbatim, and we're explicit when we don't have enough data to write a detailed description.
Data is refreshed regularly. The “last updated” date at the top of every ranking tells you how fresh the numbers on that page are. Restaurants change constantly, so always verify hours and current status before visiting.
No paid placements
We don't take payment for listings, rankings, or removals. Every spot earns its position through the public Google rating data. If a place wants to rank higher, the only path is to be a better business and earn more positive reviews.
Eventually this site may display ads to cover hosting and data costs, but ads are kept clearly separate from rankings.
Cities we cover
Get in touch
Feedback, corrections, press, or just want to say hi? hello@fivebestplaces.com