Five quantities a public record will order, and the one thing it will not

1

What an ordering needs before it can exist at all

A search phrased as a request for the best gambling sites is a request for an ordering, and an ordering is a specific mathematical object with requirements attached to it. Three of them, and all three have to hold before any list can claim to be in order rather than merely in sequence.

The first requirement is a common unit. Two quantities can be compared when they measure the same thing in the same units, and cannot otherwise, which is why nobody asks whether four kilograms beats a Tuesday. The second is a scale with known intervals. Knowing that one thing sits above another is weaker than knowing by how much, and the difference matters the moment anybody wants to average, weight or trade one criterion against another. The third is a published set of weights. Where several quantities feed one answer, something has to say how much each counts for, and if nothing does then whoever builds the list has supplied the weights out of their own head.

Run those three against what a public record actually hands over and the answer is honest and limited. The record is excellent evidence about a narrow set of facts, it is free, it takes minutes, and the party described wrote none of it. What it is not is a measuring instrument, because it was built to publish permissions rather than to rate the people holding them.

This leaf therefore does the useful half of the job and refuses the other half. Parts two to five name five quantities a reader can genuinely establish, each with the unit it carries, the source it comes from and the method for getting it. Part six sets those five against the three requirements above and shows which requirement each one fails. Part seven then prices the failure, because on this site a limitation that cannot be priced is only an opinion.

One thing is settled before any of that starts. No number produced anywhere on this page is attached to a named business, and this site publishes no rating of its own against anybody. The arithmetic below is about what an ordering can and cannot carry, worked on constructions this page declares as it goes, and the constructions exist to be tested rather than believed.

The honest summary, printed at the front so nobody has to reach part six for it, is this: five quantities can be established, four of them have no unit in common with any other, the fifth has an order but no interval, no published weights exist, and the two inputs most readers would weight heaviest are published nowhere. An ordering built on that material is a statement about its builder.

2

Two counts off a single row, and what each counts

The first two quantities come off one row and cost nothing beyond the minutes spent looking. Both are counts, both are integers, and both are frequently misread as measures of quality when they are measures of shape.

Quantity one is the number of activities a permission runs to. Its unit is a count of named activities. Its source is the published list of licensed businesses. Its method is a search on a company name or a web address, followed by reading the inventory of activities on the entry that comes back. What it settles is whether a particular product may lawfully be offered by that holder, which is a yes or a no about each activity on the list rather than a mark out of anything.

What it does not settle is worth stating because the temptation runs the other way. A holder with several activities is not thereby better than a holder with one. The count measures the breadth of a permission, so it tracks the size of a business and nothing else, and a narrow permission held impeccably outranks a broad one held badly on every criterion a reader actually cares about. Breadth is a fact about an estate.

Quantity two is the number of web addresses recorded against the same entry. Its unit is a count of addresses. Its source is the same row. Its method is reading the field and setting it beside whatever the site in front of you shows. The useful operation is a match rather than a tally: a reader asks whether the address they are standing on appears in the field, and that question has a clean answer in a way that the count never does.

Two cautions belong with both counts and they come from the publisher’s own silence. The front page of the record explains neither how a recorded address should be interpreted nor what the status words on a row are supposed to mean, and the page a reader arrives at carries no date of its own. A count taken on a Tuesday is therefore a count about that Tuesday, with the date supplied by whoever took it because nobody else supplies one.

One limitation runs beneath both counts and beneath everything in parts three to five, so it goes here rather than in a footnote at the end. Each of these quantities measures a permission and the party holding it. Not one of them measures the website a reader has actually opened. What an entry establishes is the holder of a permission together with the activities that permission covers, and it halts there, making no claim about whose hands are on a given address today, since permissions change owners and a front end can be run under an arrangement no record sets out. Every figure computed later on inherits that boundary, and no quantity of arithmetic shifts it.

Against the three requirements in part one, the two counts fail at the first hurdle rather than the last. Each has a perfectly good unit of its own. Neither shares a unit with the other, since activities and addresses are different kinds of thing, and neither shares a unit with anything in the three parts that follow. Two integers that cannot be added to one another are two facts, and two facts are not a position.

3

The one quantity a public record prints in pounds

Quantity three is the published record of regulatory action, and nothing else in this part of the subject arrives already denominated in money. Its unit is pounds, with dates attached. Its source is the regulator’s published enforcement listing, read alongside the actions page hanging off the holder’s own entry. Its method is to search both and to notice that they are not the same document.

That last point needs stating before any arithmetic, because it has caught this network out before. The two places carry different date fields for the same penalty, and one worked instance puts the same £650,000 penalty at 20 November 2025 in one place and at 30 October 2025 in the other, which is 21 days apart. Neither is wrong. They are different fields. A page naming a business has to take the date from that business’s own entry, and may cite the listing only for the shape of the record as a whole, which is precisely what this part does.

Take the shape, then. Twelve penalties stand on the listing covering roughly a year of actions, and they total £25,896,939. Divide by twelve and the average is £2,158,078.25. Sort them instead and the middle of the twelve falls at £737,500, halfway between the sixth and seventh figures. The average therefore stands at 2.93 times the middle, which is what a long tail looks like when it is written out in money.

Two further divisions finish the picture. The largest single penalty, at £10,000,000, is 38.61 per cent of everything on the listing, so one entry carries better than three pounds in eight. The smallest six entries added together come to £2,421,939, which is 9.35 per cent, so half the entries carry under a tenth of the money. And the extremes stand 81.41 to one apart, the smallest being £122,835 against that ten million.

What that distribution does to an ordering is the point of computing it. A quantity whose average sits near three times its middle cannot be averaged into anything without the largest member deciding the result, so a composite containing this one as a term is a composite about that one entry. Anybody wanting to use enforcement money in a comparison has to use a rank, a logarithm or a threshold instead of a mean, and choosing between those three is a decision nobody has published a rule for.

A second published series travels with the penalties and behaves differently. Licence suspensions appear on the same listing as five entries over the same period, and a suspension has no money attached to it at all. It is a state rather than a sum, which makes it the cleanest thing on the whole record to read and the hardest to combine with anything, since a state and a quantity have no arithmetic in common.

The deepest limitation here is the one nobody computes. An entry with no published action against it is not thereby an entry with a clean history. It may be an entry nobody has examined, and the record does not publish how many examinations produced nothing, so an absence of actions carries the same ambiguity that an absence of a row carries. Zero pounds of penalty is not a measurement of anything.

4

A name that is not a scope, and how to read the difference

Quantity four is the route out of an argument, and it arrives looking like a name while behaving like a list. Its unit is a named body together with the sectors that body is shown against. Its source is the regulator’s published roll of approved dispute entities. Its method is to find which body a business names at the close of its own complaints process, then to read the row standing beside that body on the roll.

Eight bodies carry approval and their sector coverage is uneven, which is the fact a reader needs and the fact a bare name hides. Five carry remote right across the sectors. A further two are shown against all sectors in that table, although betting alone appears in one of their own entries, which is a contradiction sitting on the page rather than anything this leaf resolves. The eighth is shown against premises-based categories only, with arcades, halls, family centres and licensed houses among them, and remote casino is not one of those categories.

Turn that into the quantity it actually is. A reader told nothing but the name of an approved body has learnt that the body sits somewhere among eight, and the probability that the named one covers the product being played is 5 in 8 on the strict reading of the table and 7 in 8 on the generous one. Those are 62.5 per cent and 87.5 per cent. Neither is 100, and the gap between them is caused by the page contradicting itself rather than by anything a reader has done wrong.

So the correct use of quantity four is not to score the choice of adjudicator but to check a scope. Two businesses naming two different approved bodies are not thereby ranked against one another, because the roll publishes no quality ordering over the eight and the regulator’s own obligation is only that an approved one is used. What can be checked is narrower and more useful: whether the body named covers the sector the argument is going to be about.

One published change on the same page is worth carrying because it shows the roll moving. A body previously on it has been removed, which the page records in its own update note. A roll that changes is a roll that has to be read on the day rather than remembered, and the date of reading belongs beside the result, in exactly the way part two required it for the two counts.

Against the three requirements in part one, quantity four fails the first and the second together. Its unit is a name and a sector list, which cannot be added to a count of activities or to a sum in pounds, and there is no scale underneath it at all, since the roll is published as an approved set rather than as an ordered one.

5

Four steps with no interval published between them

Quantity five is the only genuinely ordered thing in this corner of the subject, and even it stops short of what an ordering needs. Its unit is a label drawn from a four-step scale the regulator fixes. Its source is the business’s own written terms, where a disclosure duty bites on any licensee holding money for customers: the terms have to set out what an insolvency would do to that money, how much of it would come through, and by what arrangement. Its method is to open the terms, locate the disclosure and copy the label.

Those labels are set strings and not descriptions anybody may rewrite. The lower pair both begin with the words not protected, separated by what follows: one continues to no segregation, the other to segregation of customer funds. Above that pair stand medium protection and then high protection. A holder selects the label and drafts its own prose around it, no exact form of words having been laid down for the explanation, so the rung is fixed while the paragraph beside it belongs to the holder.

Here is why it is the most useful item on the list and still not a measurement. Four rungs in a stated order is more than anything else here manages. What nobody has published is the height of a rung. Nothing says whether moving from the lower not protected label to the upper one is a larger or a smaller improvement than moving from medium to high, and nothing turns either move into pounds, into a probability, or into any quantity that could be averaged against a count of activities.

An ordinal scale with no spacing supports two operations and refuses the rest. It supports a comparison, in that one rung stands above another. It supports a threshold, in that a reader declines everything below a line they drew themselves. It supports no addition, no mean and no weighting, so a composite turning four rungs into one, two, three and four points has manufactured the spacing and then computed with it.

A clock travels with the lower pair and carries a date. Where the label in use starts with those two words, the holder owes the customer a reminder twice a year that an insolvency would not protect the money, a statement of the sum being held at that moment, and an acknowledgement taken before anything more is staked. 31 October 2025 is the day the paragraph imposing that started to bind, and the consultation answer standing behind it turned down every threshold put to it, so there is no balance too small for the reminder to reach.

Four ordered levels, nought published intervals between them, and one reminder every six months

6

Why five quantities refuse to become one

Now put the five together and watch the combination fail, item by item, against the three requirements part one set out. The failure is not a shortcoming of the record. It is a property of the quantities, and knowing which property fails is what tells a reader what each one is still good for.

Seven rows, five of which can be filled in and two of which cannot. Read the last column downwards and the five failures sort into three kinds. Two quantities have units that exist but match nothing. One has a unit that matches money everywhere and a distribution that destroys any average taken over it. One has no scale at all. One has a scale with no spacing. The two rows at the foot have nothing, which is a different problem and gets its own part.

The requirement that breaks hardest is the third, and it breaks silently. Suppose somebody solves every difficulty above by normalising each quantity onto a common range. They still have to say how much each of the five counts for, and no source anywhere publishes a weighting. Whoever supplies it has put their own judgement into the answer, and because the five inputs all arrived from public records the result looks checkable while the step that decided it was not.

That is the honest statement this leaf exists to make, and it matters most on a site whose whole subject is what can be checked. Five things can be established. They do not combine. No score follows from them, and any score that appears to follow has had weights smuggled into it. This page therefore prints no ordering of its own against anybody, which is not modesty but the conclusion of the arithmetic above.

What survives is better than nothing and worth saying plainly. Each of the five supports a threshold, which is a decision a reader makes rather than a comparison a publisher makes. Is the product permitted at all. Does this address appear in the field. Is there a published action, and how large. Does the named body cover this sector. Is the funds level above the line I have drawn. Five thresholds applied in sequence remove candidates without ever claiming that what remains is in order.

7

What one unpublished input does to a position, in numbers

Suppose somebody ignores part six and builds the composite anyway. The damage can be computed, and computing it is more use than objecting to it. What follows is this page’s own construction, declared here so that every figure after it can be checked or rejected: five inputs, each weighted equally at twenty points, giving a hundred points in total, with four of the five taken from public records and the fifth unpublished and therefore free to be anything inside its twenty.

Two candidates are compared, and the composite on the four published inputs separates them by a gap of g points. The order comes out wrong whenever the unpublished contribution of the lower candidate exceeds that of the higher one by more than g. Each unpublished contribution runs anywhere from nought to twenty, so their difference runs from minus twenty to twenty, and it is most likely to land near nought and least likely near either end.

That gives a share for the reversal of (20 minus g) squared, divided by 800. At a gap of ten points the numerator is 100 and the share is 0.125, so the order is wrong one time in eight. At five points it is 225 over 800, or 28.125 per cent. At two points it is 324 over 800, or 40.5 per cent. At fifteen points it is 25 over 800, or 3.125 per cent. And at twenty points the numerator is nought, because a gap as wide as the whole unpublished input cannot be overturned by it.

The two money columns need their own declaration, because they rest on a construction and not on an observation. Take £120.00 as the sum a reader is deciding where to put, and treat a correct order as worth the whole of it and a reversed one as worth nothing. That is this page’s own simplification, chosen because it converts a probability into pounds without hiding a step. Under it, the worth of a position is £120.00 multiplied by the confidence in the column beside it.

Read off a position alone £120.00 Once one unpublished input runs at a ten point gap £105.00 What the hole in the composite is worth £15.00

Fifteen pounds comes out of two calculations that share no machinery, and their agreement is the check rather than an ornament. The first treats each unpublished contribution as running smoothly across its twenty points: their difference is then a triangle over minus twenty to twenty, the portion of it past a ten point gap is 100 over 800, and an eighth of £120.00 is £15.00. The second abandons smoothness and counts. Restrict each contribution to the 21 whole-point values from nought to twenty and there are 441 equally likely pairs; the pairs in which the lower candidate beats the higher by 11 or more number ten plus nine and so on down to one, which is 55; and 55 of 441 applied to £120.00 gives £14.97.

Three pence separates the two answers, and the three pence is the step size rather than a disagreement. A counted version of a smooth quantity lands near it and not on it, and the direction of the miss is set by where the boundary falls between whole points. Printing both figures rather than the tidier one is the point: a construction that survives being computed two ways is a construction worth arguing with.

Taking the drawing from its top row downwards, a practical rule falls out without any further arithmetic. Positions separated by a narrow margin on the published inputs are not separated at all, because a single hole in the composite is wide enough to swap them. Positions separated by most of the width of one input are reasonably safe. Nothing in between is reliable, and nothing anywhere tells a reader which case they are looking at unless the builder published the gaps, which is a thing no list does.

8

The two inputs a reader wants most, and where they are not

Ask anybody what they would rank on and two answers come back before any of the five above. How long does it take to get money out, and what happens when something goes wrong. Both are the right questions. Neither is carried by any source this site may lean on, and the reason differs between them.

The first has simply never been published. A hold is met on the way out, and no regulator, standard or register carries a figure for the length of time a payment is held before it goes, with no duty obliging anybody to produce one. A neighbouring duty does exist and it is about something else: a licensee may not hand a customer a route for undoing a withdrawal already requested. That governs reversal, not duration, and it says nothing at all about hours.

Any speed figure in circulation therefore came from somewhere that is not a primary source, and this page carries none. The honest treatment of the first absence is to write down the unit it would have carried, which is hours, note that nothing citable supplies it, and leave the row empty. An empty row entered on purpose is a finding. A row filled out of marketing material is an invention wearing a figure.

The second absence is the more interesting of the two, because the raw material exists while the total does not. Complaint timing is regulated in some detail: an internal process is capped at eight weeks counted from the day it lands, a complaint may still be opened up to half a year after whatever it concerns, and the running of those weeks may be stopped where the customer goes silent on a reasonable request. One offshore regulator goes further still, obliging its operators to send it complaint figures twice a year, on 15 January and 15 June, split by total, by outcome, by what is still open, by category, by referral onward and by whether anybody has gone to law.

Sending figures to a regulator is not the same act as publishing them, and that gap is where this input disappears. No total assembled from those returns could be found, so neither a volume nor a proportion upheld can be cited for that jurisdiction by this site or by anybody else working from primary sources. The British regulator sets out its own position in three sentences: it is no ombudsman, it holds no power to settle one individual complaint or to recover anybody’s money, and nothing gives it sight of an account. Enforcement outcomes get published, which is quantity three. A ratio does not.

Price those two absences on the construction in part seven and the result is worse than one hole. Two unpublished inputs out of five leave forty of the hundred points free to move rather than twenty, which doubles the width a gap has to clear before an order can be trusted. Under that arrangement a ten point gap on the three published inputs is no longer worth 87.5 per cent, since the difference of two free contributions spreads twice as far, and the gap that buys certainty rises from twenty points to forty. An ordering of two things on three published inputs is close to a coin unless they are a long way apart.

9

What stops being recorded, and what stops along with it

Everything in parts two to five assumes a British grant exists, because four of the five quantities are published only as a consequence of one. Offers reached this way stand outside the British licence, which takes away those four quantities and the duties generating them in a single movement, so on a page about measurement the duties worth naming are the ones that would otherwise be producing numbers for a reader to read.

Take those away and the arithmetic in part seven does not get worse in the way a reader might expect. It stops applying altogether. The construction there assumed four published inputs and one hole. Outside the British grant the enforcement listing carries nothing, the approved roll carries nothing, the activity inventory and the recorded addresses do not exist, and the funds level is a disclosure nobody is obliged to make. Four of the five quantities go at once, which leaves one hole and four, not four and one.

Running the same construction on that arrangement gives the honest ending. With eighty of the hundred points unpublished, the difference between two candidates on the twenty points that remain is almost never large enough to matter, and the confidence in any position converges on the fifty per cent standing at the head of part seven’s table. That is not a rhetorical flourish. It is the same formula with the free range widened, and a coin is what it produces.

Which leaves one thing worth doing rather than none, and it is a threshold rather than an ordering. A reader can still establish whether a British grant exists at all, which is a yes or a no and needs none of the apparatus above. Everything in this leaf past that first answer is addressed to the case where the answer was yes, and where it is no, the five quantities are not hard to obtain so much as absent.

10

The five checks written out as a sheet anybody can fill

The procedure below is the usable residue of everything above, and it is written as a sheet with five rows because the result is a sheet and not a verdict. Each row carries what to look for, where to look, the unit the answer comes in, and a date. The whole thing takes a few minutes and produces no ordering whatsoever, which is the intended outcome rather than a shortfall.

Row one, the activity. Search the list of licensed businesses on a company name and again on the web address, write down the inventory of activities that comes back, and record whether the product being considered is among them. The mechanics of the lookup, field by field, live on the step-by-step page for this section, since performing it badly produces an answer that looks exactly like performing it well.

Row two, the address. Read the recorded web addresses off the same entry and record whether the one in front of you is among them. Write yes, no, or not checked, and never write a count on its own, since a count of addresses measures the size of an estate and the question being asked is about a match.

Row three, the money. Search the enforcement listing and the actions page attached to the entry, and copy both the figure and the date, taking the date from the holder’s own page where the two differ. Record nought pounds as nought pounds and not as a clean history, because an entry with no published action may be an entry nobody has looked at.

Row four, the route out. Find the body named at the end of the complaints process in the business’s own documents, then read its sector row on the approved roll and record whether the sector covers the product. The answer is a yes or a no about scope, not a mark against the body, and the roll changes, so the date belongs in the row beside it.

Row five, the level. Open the written terms, find the disclosure about what happens to customer money if the business fails, and copy the label exactly as it is given, since the labels are fixed strings and the explanation around them is the holder’s own prose. Decide in advance which levels you will accept and apply that as a threshold rather than converting the four levels into points.

Then stop, and stop deliberately. Two rows at the foot of the sheet stay empty, those being how long a payment takes to leave and what share of complaints end in the customer’s favour, and the right thing to write in them is the word unpublished and the date it was looked for. A sheet with five rows filled, two rows honestly blank and a date at the top is a better object than any list in order, because every line on it can be checked by somebody else, and because the construction in part seven shows what the two blanks would have done to the order if anybody had tried to supply them.