The toplist, and who chooses not to be on it

stats4u.net has ranked its busiest counters since the beginning — a 2012 post already talks about improving it. The current version lives at /top, and it works differently underneath than it did back then: 25 counters by 30-day visitor count, a separate "fastest growing" list comparing this week against last, and a breakdown by country and by language. Filter by either and the page stops asking to be indexed — a per-country cut is a subset of the same 25 counters, not a new page worth a search engine's attention.

Ranked by visitors, not by hits

The main list sums unique visitors, not page views, over the last 30 days. A counter refreshed obsessively by its own owner doesn't climb the list for it; a counter with steady outside traffic does.

Growing, not just big

The second list isn't the same 25 counters sorted differently. It compares each counter's most recent 7 days against the 7 days before that and ranks by relative growth — with a floor: a counter needs at least 20 visits in the earlier week before it's eligible at all. Without that floor, a brand-new counter going from 1 visit to 4 would outrank a real trend, at 300% growth on a sample size too small to mean anything.

Who isn't there

As this is written, 2,193 counters had any traffic in the last 30 days. 2,010 of them are eligible for the toplist. The gap is entirely opt-outs: a counter has to be public, not disabled, not password-protected, and not explicitly marked "leave me off the toplist" — that last flag alone accounts for 20 counters whose owners want their traffic counted but not published in a ranking. All four conditions are checked on every request, not cached from whatever the flags said when the list was last rebuilt.

171 countries show up in the country breakdown over the same 30 days — filtered through the same visibility rules as the main list, so a private counter's visitors don't leak into a country's count either.

Nothing here is new functionality. It's a page that has quietly done its job since before this rebuild, checked against what it actually returns rather than what the code was assumed to do.