How we measure

Every other directory takes a game’s own blurb, rewords it and publishes. That reads fine and tells you nothing. We measure instead — 5 things we stand behind today, one more still being validated — and put the date we recorded them on the page.

Still live

Every week

We request the page and record the status. Anything that stops responding is flagged and comes off the index.

Truly daily

Every week

We fingerprint the visible text on separate calendar days and compare. One sample proves the page is dynamic; it does not prove it is daily, so we wait for the second day rather than guess.

Signup

Every week

We look for a login that actually blocks play: a redirect to a sign-in page, a password field inside a covering dialog, or a page that is nothing but a login form. A newsletter box in the footer, a header login widget and a navigation drawer that lists “Log in” are all none of those.

Ad load

Every week

We count requests to known ad networks and measure how much of the screen ad frames occupy. Reported as none, a few, or heavy.

On a phone

Every week

We load the page at 390 px and check for horizontal overflow and tappable targets.

Real time

Held back — validatingNot yet

A median of sessions we have timed ourselves. We cannot time you playing on someone else's site, so this one waits for our own puzzles rather than repeating the game's own claim. Entries say “not checked yet” until then.

When we can’t tell

An entry shows “Not checked yet” whenever the evidence is not there — a puzzle we have only sampled once, a game we have no timing data for, a page that was unreachable on the last sweep. We would rather show a gap than a guess. A field left at its default value is not a measurement, and we do not present it as one.

What the signup check can and cannot see

Measured against a set of real pages, this check is right every time it says “Required” — and misses about a third of the walls that exist. So we publish the positive and hedge the negative: an entry that says “None detected” means we did not find a login blocking play, not that none exists.

Getting there took three false positives, each a different shape: a page wrapper that happened to carry a “modal” class, an inline login box on a site whose content was perfectly usable, and an off-canvas navigation drawer that listed “Log in” among a dozen menu items. The remaining blind spot is a plain login landing page — we could not find a reliable way to tell one from a working page that merely has a login box in its header, so we chose to miss those rather than accuse innocent sites.

Keeping the checks honest

Detectors drift. A site redesigns, a selector stops matching, and a check quietly starts returning the wrong answer for everything. Two things guard against that: a fixed set of test pages with known answers that runs alongside every sweep, and an alert when any check’s rate moves sharply across the whole index — if “signup required” jumps overnight, the likely explanation is our detector, not the entire web.

Found something wrong?

Tell us and we will re-run that entry. Corrections are the cheapest signal we get.