Kirill Maltsev

Search Bench - The blind taste test for search engines

by
Most people have strong opinions about search engines but have never compared results without knowing which is which. Search Bench changes that. It shows you two anonymous result sets for any query you type, and you pick the one you prefer. No engine names until after you vote. Every comparison feeds a live public leaderboard — so you can see what people actually prefer when brand bias is removed.

Add a comment

Replies

Best
Kirill Maltsev
Hey everyone — I'm Kirill, the solo creator of Search Bench 👋 This started as a personal experiment. I realized I had very strong opinions about search engines, but had never actually compared results blind. Search Bench hides the engine names and uses pairwise comparisons instead of ratings. Each vote answers a simple question: which result set is better for this query? A few important notes: - This is not an objective ranking. Queries and voters are self-selected. - "Better" is intentionally undefined — relevance, freshness, usefulness, vibes… it's up to you. - The leaderboard is derived from a Bradley–Terry model (similar foundations to Elo, but better suited for pairwise data). Right now, the dataset is small, so the most valuable thing you can do is try a few comparisons with your own real queries and see if the results surprise you. I'm especially interested in feedback on: - biased sources I've missed - ways this could be gamed or misinterpreted - whether the signal holds at scale Happy to answer any questions — and thanks for checking it out.