Nexus Panel BoardPanels of numbers this site works out for itself

How the numbers are made

Two sources, no others, and every metric on the board traces back to one of them in a way you can repeat yourself.

Tiles drawn across the site
276
stat tiles drawn across the site
Data sources in total
3
addresses published
Alphabet the set touches
32 of 32
symbols the set touches
Pages at each depth
56
1
2
3
pages at each depth, front page first

Nexus market addresses in this board

copy, do not retype
nexusb2l73qzjn4slhyfxa3jvpolw7fomiz5sgyyefnsdhikaqgborqd.onion
nexusma2iekjhhyenua3u4zlyfsj2ubwxr2nt6gdte5rvwukzze63fyd.onion
nexusabcdbnjw6or46y3hh6uicsl4xvu5cncfp27hkggznvtxhxlngad.onion

Printed as supplied, in no order, with nothing marked best or main. This board never opens an address, so nothing on this page says one of them is reachable for you now.

The rule one sentence long

read this first

If a number appears anywhere on this site, it was worked out at the moment you asked for the page, from either the three onion strings or the site own files. Nothing is stored, nothing is remembered between requests, and nothing came from anywhere else.

That is a narrow rule and it costs the site a lot. It rules out everything readers usually want from a page about a market: whether it is up, how busy it is, how long a withdrawal takes, what people think of it. None of that can be produced from a string or a file, so none of it is here. What is left is smaller and true.

Source one the three onion strings

The three addresses printed at the top of every page are the site only external input. They are stored once, as plain text, in a single configuration file, and they are never fetched, tested, or compared against any list. Every string metric on the board is ordinary arithmetic over those characters.

MetricExactly how it is worked out
LengthThe character count of the address with the suffix removed. Nothing more.
Digit and letter countsCount of characters matching a digit, and the remainder. Done per address and summed across the set.
Distinct symbolsThe address split into characters, duplicates dropped, the rest counted.
Alphabet coverageThe format allows thirty two symbols. The metric counts how many of those appear at least once, either in one address or across all three.
Shared openingCharacters compared position by position from the front until two addresses disagree, then the shortest of those results across all three pairs.
Divergence pointThe shared opening length plus one. That is the first position where the three are not all the same.
Position agreementFor each position, the three characters at that position are compared. The result is three counts: all three the same, exactly two the same, all three different.
Pair distanceFor a pair of addresses, the number of positions holding different characters.
Longest runThe longest stretch of one character repeated immediately, found with a regular expression over the string.
Longest repeated chunkThe longest substring that occurs twice inside the same address, found by trying lengths from long to short.
Repeated pairsEvery neighbouring pair of characters is counted, and the metric reports how many pairs occur more than once.
Vowels and vowel gapsCount of the five vowel letters, and the length of the longest stretch containing none of them.
Easily traded symbolsA fixed, named set of symbols that a tired reader confuses, counted per address. The set is chosen by this site and stated on the panel that uses it.
EighthsThe address cut into seven equal blocks, with digits, distinct symbols or vowels counted inside each. That gives the small line and bar pictures.

Source two this site own files

The second source is the page registry that builds this site. It is a set of PHP arrays, one entry per page, and the corpus metrics walk it the same way the router does.

MetricExactly how it is worked out
Page and panel countsEntries in the registry. A panel is any entry that sits inside a section and is not the section page itself.
Word countA page prose is flattened into one string, tags are stripped, placeholders count as one word, whitespace is collapsed, and the result is split on spaces.
Reading timeWord count divided by two hundred and twenty, rounded up, with a floor of one minute.
Average, shortest and longest panelThe same word count taken over every panel and reduced in the obvious way.
Words per sectionWord counts summed within each section, including the section page.
Tiles drawnThe number of tile definitions across every page in the registry.
Links between panelsThe count of hand placed related links across the registry. Navigation, breadcrumbs and the footer index are not counted, because those are generated.
Phrase coverageA phrase is lowercased and searched for in each page title, heading and flattened prose. The metric reports how many pages contain it.
DepthThe number of slashes in the page address plus one. It is an upper bound on clicks from the front, not a measured path.

What happens when there is nothing to draw

Some metrics legitimately come back with nothing. When that happens the tile draws a small empty state: an outlined circle with a line through it, and a sentence saying what came back empty. It does not fall back to a plausible looking number, it does not hide, and it does not shrink the strip so you cannot tell a tile was there.

This is worth insisting on. The moment a board fills an empty result with something reasonable, every other number on it becomes a guess as well, because you can no longer tell which is which.

What is not measured and cannot become measured

  • Whether an address answers. The site never opens a connection, so there is no status, no uptime percentage and no last checked time anywhere on the board.
  • Anything about visitors. There is no analytics script, no counter, no log analysis and no cookie. The site does not know how many people read it and has no way to find out.
  • Anything inside the market. No prices, no listings, no vendor names, no policies, no balances, no timings.
  • Anything from another site. There are no external requests in the pages or in the stylesheet, so nothing here can be enriched by anybody else data.
  • Dates. The board uses relative language throughout, and the only date it ever emits is a machine readable modification stamp taken from the file timestamps.

Checking it yourself the point of the whole thing

Every string metric can be reproduced with a text editor and patience. Copy an address, count the characters, count the digits, compare two of them position by position. You will get the same answers this board gets, because there is nothing else in the calculation.

The corpus metrics take longer but work the same way. Open every page, count the words, add them up. If your total differs from the one shown, the difference will be in how you treat headings and list items, and the rule this board uses is written above.

That reproducibility is the only reason to trust anything here. It is not a claim about integrity. It is a property of arithmetic that is small enough to check.

Questions that come up on this panel

Does the board ever cache a number?

Only within one request. Some values are worked out once and reused while the page is being built, then thrown away. Nothing survives between requests.

Why compute at request time rather than storing results?

Because a stored number can drift away from the thing it describes without anybody noticing. Recomputing means a wrong number is a bug you can see rather than a stale file you cannot.

Could a metric here be wrong?

Yes, in the way arithmetic can be wrong: a mistake in the rule. What cannot happen is a number without a rule behind it, which is the failure mode this design exists to prevent.

Reading this panel wrong

3 ways it happens
  1. Reading this page as a promise of accuracy.It is a promise of provenance. Every figure has a stated rule, which is a different and smaller claim than every figure being useful.
  2. Assuming a computed number is a meaningful number.Character counts are trivially computable and mostly trivial. The prose on each panel exists to say which of its numbers actually matter.
  3. Thinking the two sources cover the subject.They cover a corner of it. Most of what a reader wants to know about a market is outside both, and the board says so rather than filling the gap.

What this panel is not: a methodology in the research sense. It is a list of arithmetic, written down so that nothing on the site can hide behind a chart.

Nearby panels

Every panel on the board

68 pages
Strings
How long each address is The symbols an address can contain The opening all three share Where the three part company Digits, and where they fall What repeats inside an address How far apart the three are The character they all end on
Reading
Symbols the eye trades What a front and back check proves The part nobody reads Reading an address in blocks Copying rather than typing The stretches you cannot say Reading an address out loud What a check cannot reach
Route
What the browser is doing The first load of the session When a page loads slowly Why the path keeps changing What each step can see The screens that mean failure Where the address lives between visits Where the address came from
Session
The name you choose The login screen How long a session lasts What stays behind Two tabs, one account Ending a session on purpose Losing access Encrypted text in a session
Money
The deposit address Waiting for confirmations What escrow holds Where the cost sits Rounding and leftovers Taking a balance back out The waiting, laid out The steps you cannot take back
Orders
Before the order exists Placing the order The delivery details Between placing and dispatch What tracking does not say Finalising When it goes to dispute The feedback you leave
Corpus
How much is on this board How the panels are distributed How long a panel is The tiles, counted Which phrases the board uses What is deliberately left out How the panels connect The tiles that come back empty
Reference
All panelsHow the numbers are madeThe addressesWhat this board covers