The Showdown Score™

The only score that counts how often a product breaks, not just how owners feel about it.

Every product we card gets a Showdown Score out of 10. It sits on the card, right under the product's name, and on its own gear page. Today 680 products carry one. A star rating tells you how happy owners felt the week they wrote in. The Showdown Score asks the question you actually have before you spend the money: will this thing still be working next summer?

You worked hard for that money. Before you spend it, you deserve solid facts and the experience of people who already own the thing. This score puts that evidence in one place.

What goes into it

Three kinds of evidence, read from the product's own listings and dated on the day we read them. Owner ratings come from where the product is sold: Amazon for most, and the retailer or the maker's own store for products Amazon doesn't carry.

Owners

What owners rate it. A rating from a handful of buyers counts for less than the same rating from thousands, and a rating that is sliding counts for less than one that is holding. A rating pooled across a family of different models is not this product's own, so it is never ranked against one that is. A base seeded with free products is labelled on the card.

Failure record

How often it lets owners down, and how. We read the share of ratings at one or two stars, then the newest critical reviews themselves, and count what they describe. A unit that died counts harder than one that simply did not suit the buyer's setup, and dying in the first months counts harder than wearing out after years.

Hands-on test

Whether a credentialed tester backed this model. A test of this exact model can lift the score. A tester's pan keeps a product from winning. No test never costs a product a single point: plenty of good gear has never been on a test bench.

What a normal failure record looks like

Some gear draws more complaints than others by nature: a robot pool cleaner has motors, seals and software to fail, a fire pit has almost nothing. So we judge each product's failure record against other products of its kind. These are the typical figures we compare with: the share of owner ratings at one to three stars, with a one-star rating counting fully, a two-star 0.6 and a three-star 0.2. Each comes from every model of that kind we have read with 100+ ratings, measured 2026-11-06.

A kind of gear with fewer than eight well-reviewed models is compared with its broader family instead:

A product with only a few ratings is pulled toward its kind's typical figure until it has enough of its own, so a short spotless record is not mistaken for a proven one. The typical figures are built from the same owner-rating reads as the scores themselves.

What stays out

• Price. A cheap product that breaks is not a good product. Price is shown beside the score, and a value pick is judged on both.
• The warranty. A warranty does not stop a product breaking. It is shown beside the score, never in it.
• What a link pays us. Commission is not an input. Nothing about what a link pays enters the score.
• The marketing. Nothing in a maker's own marketing goes into the number.

What a score looks like

A live one: the Intex Easy Set 12 ft x 30 in Pool Set (28131EH). Tap See why it scored 4.9 to open the three parts. They always add up to the score shown.

The bands

Every score lands in a band, and the band word is printed beside the number.

How to read it

• The confidence word says how much evidence sits behind the number. A score on a deep, fully read base is high confidence; a thin one is marked early data and drawn with the range it could really be in.
• The stars line. When the score and the stars tell different stories, the card says so in one line. Today 117 cards carry one: products owners rate well that fail more often than the stars let on. The 4.0-star-and-up ones are all on one page: Highly rated on Amazon, but they don't hold up.
• "Not scored yet" means part of the evidence has not been read. We would rather show a question mark than a number we cannot stand behind.
• No score at all means an active safety recall, or a defect serious enough that we pulled the product. Neither gets a number.
• Every score is dated. When we re-read a product, its score moves, and the date moves with it.

Scores and our verdicts

The score and the winner we name are built on the same evidence, so they cannot quietly disagree. Wherever a winner was chosen among rivals we can score, it is checked against them. When new evidence puts a rival ahead, we re-run that showdown the same day, and if the rival holds up, the crown moves.

Why we don't publish the formula

We publish what counts and what does not. We keep the exact weights to ourselves for one reason: a published formula is a recipe for gaming it. If a seller knew exactly which reviews move the number, the number would stop meaning anything. What you can always see is the evidence itself: the owner base, the failure record and the test, dated, on every card.

More on how we check every claim a listing makes: How we test.