Best gambling sites, measured in the only unit that compares them: the price of one spin

1

One unit, and why it has to be the spin

Ask which gambling sites are best and the answers arrive as adjectives. Generous. Trusted. Fast. Not one of those words can be subtracted from another, which is the whole difficulty: a reader is handed a pile of claims in a currency that has no arithmetic. So this sheet picks a unit with arithmetic in it, and the unit is the price of a single spin.

Fix the tally at 800 and the price at 10p and the whole comparison becomes three multiplications. Staking is 800 times 10p, which is £80.00. Expected return at a 96 per cent configuration is 96 per cent of that, which is £76.80. The gap between those two figures, £3.20, is what the arithmetic says the 800 presses cost on average, and it works out at four tenths of a penny each time the button goes down. Pace is the fourth figure: 2.5 seconds is the gap a licensed slot holds from the opening of one cycle to the opening of the next, so 800 of them fill 2,000 seconds, which is 33 minutes and 20 seconds of unbroken pressing.

Staking, expected return, cost and pace are the comparison. A site that hands over 800 spins has handed over 33 minutes and an expected £76.80, whatever the banner says, and a site that hands over 800 spins with conditions attached has handed over less than that, by an amount the conditions decide. Nothing on this page asks a reader to believe a ranking. It asks a reader to convert.

The conversion is also where most marketing quietly dies, because the tally is advertised and the price rarely is. Eight hundred spins sounds like one thing. Here is what it is, at seven prices a cashier might set.

Top row to bottom row is a factor of 500, and the tally never moved. That is the reason a count on its own ranks nothing: a thousand spins at a penny is worth less than a hundred at 10p, and the headline figure is bigger on the worse award. The ceiling in the last row is not a house guess either. An instrument made on 25 February 2025 holds an online slot stake to £5 once a player has turned 25, with £2 the figure between 18 and 24, and two guidance notes behind it carrying 9 April and 21 May 2025. The cap reaches slots and nothing else, so a table game or a live dealer round sits outside it.

For a reader aged 18 to 24 the last row of that table is unlawful on a licensed site, and the ceiling row becomes £2: 800 spins at £2 is £1,600.00 of staking, an expected £1,536.00 back, and £64.00 of expected cost. Two readers, the same award, a cost that differs by £96.00 because of a birthday. No adjective carries that information. The unit does.

800 spins, 10p each, £3.20 of expected cost, 33 min 20 s of pressing

2

A number is worth nothing until it can leave

An expected £76.80 is a figure about a balance, and a balance is not money. What turns one into the other is the withdrawal rule, and in Great Britain the rule is written in the player’s favour in terms blunter than most readers expect.

Withdrawal from a deposit balance, winnings and all, is a player’s entitlement at any moment. There must be no restriction on the largest sum a player can take out at any time. Reasonableness governs any cap on the size of a withdrawal, and any cap on how many are taken. Those three statements come off one guidance page, and that page shows no publication date, with nothing on it to say when the text was last revised, which is worth saying out loud: the rule can be cited, and a freshness date cannot be attached to it, because the page does not state one.

A fourth rule closes the back door. The technical standard on responsible product design denies consumers the option of cancelling a withdrawal request. Guidance beneath it adds two things: a customer who has asked to withdraw should not then be offered the chance to redeposit those same funds, and the way out ought to be smoothed of friction so far as an operator can manage it. The point of that requirement is simple and the arithmetic makes it plain: the moment a reader cancels a withdrawal of £76.80 and stakes it instead, the expected cost of the next 768 spins at 10p is £3.07, and the balance that was money is a balance again.

Put the two sides together and a comparison between sites gets its first real column. One side has a withdrawal entitlement with no stated maximum, a reasonableness test on any cap, and a prohibition on offering the cancel button. The other side has whatever its terms happen to say on the morning you read them.

There is a second trapdoor in the licence conditions, and it is the one that catches people. An operator must verify a customer’s identity before that customer gambles, and the obligation is on the operator rather than on the player. Where verification has not been completed, the money a reader believes is theirs sits behind a door that the operator has to open before anything moves. A reader who treats £76.80 as cash before that step is finished has mistaken a figure on a screen for a payment.

One payment route is closed outright. From 14 April 2020 onward no licensee may take a credit card, whether directly or by way of a money service business that would in substance pass the same credit into the same gambling transaction. That rule removes a particular arithmetic from the table: 800 spins at 10p staked on borrowed money carries the £3.20 of expected cost and then whatever the card charges on top, which is a cost no return percentage touches.

No stated maximum, no cancel button, and £3.07 of cost in the cancel button alone

3

The complaints clock and the eight names behind it

Any ranking of sites is in part a ranking of what happens when something goes wrong, and on the licensed side that process is timed rather than promised. A complaint should be acknowledged by a licensee as soon as reasonably possible and inside 3 working days of its receipt, and inside 24 hours where that licensee offers gambling facilities at any hour of the day. Internal escalation included, the whole process may run eight weeks at most, counted from the day the complaint first landed, and the guidance page traces that period to code provision 6.1.1.2 on social responsibility.

The clock can be stopped, which is a detail most summaries drop. Leave a reasonable request for information unanswered for seven days and it may be reasonable for the operator to halt the clock until an answer arrives, at which point it should pick up where it paused rather than at the beginning. So eight weeks is eight weeks of the operator’s working time, not eight weeks of calendar time in every case.

The window for raising the complaint at all runs to at least six months from the date of the incident. An operator may encourage a reader to come forward sooner, and it may not write a term that bars a complaint inside that six-month window.

At the end of the internal process comes a final letter, and the guidance tells the operator what it must contain: the decision and the reason for it, a statement that this is the end of the operator’s own complaints process, and an explanation of how to escalate to an independent alternative dispute resolution entity. The reader is not left to find the next step unaided.

Which brings in the eight. Approval of the alternative dispute resolution entities rests with the regulator, only an approved one is open to a licensee, and eight stand on the published list: ADR Group; Pegasus ADR Service; Blexr Limited; Centre for Effective Dispute Resolution, shortened to CEDR; eCOGRA; IBAS; ProMediate; and Tattersalls Committee. The update note on that page records that Lindsays has been removed as a provider. That page went up on 17 May 2021 and was last revised on 15 January 2025, and it puts the obligation in a line: licensed operators are required only to use one or more entity which the regulator has approved.

Sector coverage is not uniform across those eight, and for a reader whose complaint is about a casino that difference decides who can hear it.

Five names in that first block, one qualified name, one name that needs both of its listings read, and one name that does not cover remote casino at all. The page describes IBAS as adjudicating on disputes between licensed gambling operators and their customers, and the word in that description doing the work is licensed.

On what it will not do the regulator is equally direct, and the phrase is its own: it is not an ombudsman. It says that this means the powers to resolve gambling complaints or to help consumers get their money back are not held; that a complaint about a gambling transaction, a problem with a withdrawal or a technical fault mid-game among them, is not one it can resolve; and that customers’ accounts are closed to it, so it cannot see what has happened when a reader is having problems. Its own direction sends a player to the gambling business with the complaint before anywhere else. That page is published 11 December 2023 and last updated 4 June 2024.

Priced in the unit of this sheet, the complaints machinery is worth the whole £76.80 and then some, because it is the only mechanism that gets a disputed balance looked at by somebody other than the operator holding it. An eight-week ceiling and a named list of adjudicators is not a bonus. It is the thing that makes the other figures mean anything.

3 working days, 24 hours, eight weeks, six months, and five of eight names reaching a casino

4

How big the thing being compared actually is

A page ranking gambling sites ought to say how large the field is, and the published figures are specific. Across April 2025 to March 2026 Great Britain’s gross gambling yield came in at £17.5 billion, standing 4.4 per cent above the twelve months preceding it. Strip out every reported lottery and the figure is £13.2 billion, up 4.7 per cent. The difference between those two, £4.3 billion, is the lottery share of the headline. Publication date was 17 September 2026.

Inside that total, £8.3 billion belongs to the sector grouping remote casino with betting and bingo, a 6.9 per cent increase. Land-based arcades, betting, bingo and casino together produced £4.9 billion, up 1.1 per cent. Premises holding a licence came to 8,081, a fall of 2.0 per cent, with 5,617 of them betting shops after a 3.6 per cent fall. Subtract the second of those from the first and 2,464 licensed premises are something other than a betting shop, which puts shops at a little under seven in every ten.

Two further figures come from the annual report itself and they are the two this site needs most: online casino games produced £5.7 billion of that £8.3 billion, and slots games produced £4.8 billion. That same report puts active accounts at 25.7 million as the final reporting quarter ended, and records for 31 March 2026 a register of 2,154 licensed operators whose 3,097 licensed activities come out at 1.44 apiece.

Now convert. One tally on this sheet costs an expected £3.20, so dividing a yield figure by £3.20 gives the number of such tallies the year’s yield would pay for. The slots figure of £4.8 billion comes to exactly 1.5 billion tallies of 800 spins. Online casino games at £5.7 billion come to 1,781,250,000. The whole remote sector at £8.3 billion comes to 2,593,750,000, and the £17.5 billion headline to 5,468,750,000. The land-based figure of £4.9 billion is 1,531,250,000, which is a little more than the slots figure despite covering four whole sectors.

One caution belongs on the headline, and it is a caution about a label rather than about a sum. Describing the same £17.5 billion, the statistics hub reaches for the market’s value in Great Britain and the news release for total gross gambling yield. Where a page needs to be precise it should use the release’s wording and say gross gambling yield, because that is what the figure is: stakes received less winnings paid out, not turnover and not revenue in the ordinary sense. Elsewhere the hub’s phrase is the regulator’s own too, so neither form is wrong, and only one of them is exact.

The distinction matters to the arithmetic on this sheet. Turnover and yield differ by the return percentage, and at 96 per cent they differ by a factor of twenty-five. The £80.00 staked across 800 spins produces £3.20 of yield, and £3.20 times twenty-five is £80.00 again. So a reader who reads a yield figure as though it were the amount staked has understated the staking by that same factor, and a reader who reads it as the amount lost by players has it right.

£4.8 billion of slots yield is 1.5 billion tallies of 800 spins

5

Per account, which is not per person

Divide the remote sector’s £8.3 billion by 25.7 million active accounts and £322.96 is the quotient. That is a year of net loss per account, averaged, and it converts straight into the unit of this sheet: priced at 10p with a 96 per cent configuration it is 3,229 spins, which is four of this sheet’s tallies and 29 spins over. At the lawful pace floor those 3,229 presses occupy 8,072 seconds, or 2 hours 14 minutes and 32 seconds of pressing spread across a whole year.

Two cautions sit on that quotient and both of them cut the same way. An account is not a person, because one reader may hold accounts at several operators and the same reader is then counted several times over. And the numerator’s remote sector takes in bingo and betting alongside casino, while the denominator counts accounts across that whole sector, so the figure describes the sector and not a casino player. Narrow the numerator to online casino games at £5.7 billion and the quotient falls to £221.79 per account, which is 2,217 spins, or 1 hour 32 minutes and 23 seconds of pressing.

The reason to run that division at all is that an average is the only honest way to answer the question a ranking cannot answer: what does a year of this cost. It costs, on the published figures, somewhere around two or three hundred pounds for the average account, and the arithmetic on this sheet says the same thing from the other end. Four tallies of 800 spins at 10p stake £320.00 and cost an expected £12.80. Losing £322.96 at a 10p price on a 96 per cent configuration takes 80,740 spins, which is 56 hours at the lawful floor, and that gap between £12.80 and £322.96 is the whole distance between an occasional tally and the average account.

Read the bottom two rows together and the shape of the year appears. One tally is 800 presses and 33 minutes for £3.20. The slots average per account is £186.77, which is 58 such tallies, and the pressing time attached to the loss rather than to the tally is far longer, because the loss accumulates four tenths of a penny at a time. Nothing in a ranking can shorten that. Only a smaller price or a smaller tally can.

£322.96 a year per active account, which is 3,229 spins at 10p

6

Who the tax falls on, and at what rate

Readers comparing sites often arrive with a half-remembered belief that British gambling winnings are tax-free, and the published position is both narrower and more interesting than that. There is no gov.uk or HMRC page stating in terms that gambling winnings are not taxed. The position is reached by absence, through two pages of HMRC’s Business Income Manual that address a different question entirely: whether gambling is a trade.

On the first of those pages betting and gambling, as such, amount to no trading. It adds that an organised activity directed at making money out of the gambling public will normally amount to trading; that a spread-betting taxpayer is not normally carrying on a trade and so is taxable on no profits and takes relief for no losses; and that the bookmaker who organises the bet is taxable on what it makes. That page was published 22 November 2013 and last updated 4 August 2026.

The second page supplies the case law, and it is a case from 1925. In Graham v Green the man before the court lived entirely off starting-price bets on horses, and Rowlatt J held that no trade existed, in a sentence that has outlived everything around it: there is no tax on a habit. That manual records that the fact a taxpayer has a system by which they place their bets does not of itself make the activity a trade, that living on the proceeds does not either, and that having expertise or being systematic, studying form, is not enough. The same page marks out where a gambler can be trading, with money paid for a television appearance as its example, and it says that whether winnings of that sort are proceeds of a trade would turn on the facts.

So the honest formulation, and the only one a page should use, runs this way: HMRC publishes the position that betting and gambling do not of themselves constitute a trade, and trading income is therefore not what winnings are. Not that a rule declares winnings tax-free. No individual’s circumstances are addressed anywhere on this sheet, and tax advice is not what any of it is.

The charge falls somewhere, and it falls on the operator. Remote Gaming Duty attaches to gaming provider profits out of remote gaming that a customer whose usual home is the UK has played, those profits being stakes received less winnings paid out. Which is the same subtraction this sheet has been running all along. On 800 spins at 10p the operator receives £80.00 of stakes and expects to pay out £76.80, leaving £3.20 of profit, and the duty is a percentage of that £3.20.

The percentage has moved twice. Dated 1 April 2026 and onward, Remote Gaming Duty stands at 40 per cent, where 21 per cent had applied from 1 April 2019 onward and 15 per cent in the years before that. General Betting Duty, a different charge, is 15 per cent for fixed-odds and totalisator bets, 3 per cent for financial spread bets, 10 per cent for other spread bets, and 15 per cent of the commission a betting exchange charges its UK users. The guidance page is published 14 August 2014 and was last updated 1 April 2026.

Top row against bottom row, the Exchequer’s share of one 800-spin tally rises by 80p, and the operator’s share falls by that same 80p, while the player’s expected cost of £3.20 does not move at all. That is the shape of the whole charge: it divides the £3.20 and it does not create it. A reader who hopes a duty rise will be paid by the house rather than passed along should notice that the only two variables a site controls are the price of a spin and the configured return, and both of them sit in the player’s column.

At sector scale the same multiplication gives an order of magnitude rather than a tax bill. Forty per cent of the £4.8 billion slots yield is £1.92 billion, and forty per cent of the £5.7 billion online casino games yield is £2.28 billion. Those are illustrations and not receipts, since residence rather than a sector label decides where the duty bites, and the published yield figures are no tax base. The arithmetic is offered to show the size of the slice, not to compute it.

£1.28 of duty on £3.20 of expected profit, and 0.16p of it on every spin

7

The duty follows the player, the protection follows the licence

Here is the asymmetry that any list of gambling sites owes its reader, and it reads sharper stated with the tax rules in hand than in the vague language of legality.

Registration and payment are required of an operator based abroad once it offers remote gambling, spread betting excepted, to someone whose usual home is the UK. An operator outside a group has generally to appoint a representative approved by HMRC here, which falls away where its base is an exempt jurisdiction, an EU member state, Norway and Gibraltar being the examples given. The guidance is addressed to the operator throughout and describes no charge to the player at all. Which yields one precise sentence worth stating plainly: take money from a British resident and an offshore operator falls within the reach of British gaming duty, whether a licence exists or not.

Set that beside the pages above. The duty reaches across the border by following the player’s residence. The complaints clock does not. The eight approved adjudicators do not. The withdrawal entitlement with no stated maximum does not. The verification duty, the credit-card prohibition, the 2.5-second pace floor, the limit holding a bonus requirement to ten times the funds, the £5 and £2 slot caps: not one of them follows the player across that border. The obligation that travels is the one owed to the Exchequer, and the obligations owed to the reader stay behind with the licence.

So a reader choosing between a licensed site and an unlicensed one is not choosing between two regulated options with different features. They are choosing whether the rules in parts two, three and six of this sheet apply to the £76.80 at all. The arithmetic of the award does not change: 800 spins at 10p stake £80.00 and expect £76.80 back on either side of the line. What changes is whether anything compels the second figure to be paid.

One more point on the representative rule, because it is often read backwards. The exempt jurisdictions in that list are exempt from the requirement to appoint a UK representative for duty purposes. They are not British licensing jurisdictions, and an operator based in one of them is not thereby licensed in Great Britain. A reader who spots a familiar European address in a footer and reads it as a British permission has combined two different rules into one wrong conclusion.

The duty crosses the border, the eight-week clock does not

8

What the survey counted, and the figure to leave alone

The third annual Gambling Survey for Great Britain was published on 16 July 2026, and its headline is that 47 per cent of participants aged 18 and over told the survey they had participated in any form of gambling in the past four weeks. The statistics hub puts the period covered at January 2025 through January 2026. Leave out those who only play lottery draws and 27 per cent remains, the twenty points between the two figures being the lottery.

A third figure sits beneath them: the share of the population reaching a Problem Gambling Severity Index score of 8 or above was 2.4 per cent in 2025, which the regulator calls a stable level over the survey’s three years. Approximately 20,000 adults aged 18 and over are interviewed each year, and the report covers three consecutive years of official statistics from 2023 to 2025.

Those are the figures a comparison page may use and the arithmetic they support is modest. Of every thousand adults, roughly 470 gambled in some form in the four weeks before being asked, roughly 270 did something other than a lottery draw, and roughly 24 scored at or above the index threshold. In the unit of this sheet, 24 people in a thousand is a little under one in forty, and one in forty is close to the house edge itself, which is a coincidence of numbers rather than a finding and is worth nothing beyond being easy to remember.

There is one further published figure that must never be dressed up as a rate of adult participation. A 30 per cent figure attaches to young people who had spent money of their own on gambling inside the preceding 12 months, covering the first half of 2025 and released that November. It is a figure about young people. Set beside the 47 per cent or the 27 per cent it looks like a comparison and it is not one, because the population, the question and the period all differ.

Why a comparison page should care about a participation rate at all comes back to the unit. A rate describes how many people are in a room. It says nothing whatever about the price of a spin in that room, which is the only figure that decides what the room costs. Forty-seven per cent of adults gambling in four weeks at 1p a spin and 47 per cent gambling at £5.00 a spin are the same rate and a 500-fold difference in expected cost, which is exactly the point the first table on this sheet makes.

47 per cent, 27 per cent, 2.4 per cent, and a figure about young people that is none of them

9

What an offshore offer sits outside of

An offer in this bracket has no Great Britain licence behind it. That sentence is short and the consequences are not, so here they are as a list of named absences rather than as a warning.

Gone is the acknowledgement inside 3 working days and the 24-hour acknowledgement where gambling runs around the clock. Gone is the eight-week ceiling on the whole complaints process and the six-month window for bringing one. Gone is the final letter that has to explain the decision and name the route onward. Gone is the list of eight approved adjudicators, five of which reach a remote casino dispute. Gone is the withdrawal entitlement with no stated maximum and the reasonableness test on any limit. Gone is the prohibition on offering a cancel button over a withdrawal request. Gone is the limit that holds a bonus requirement to ten times the funds. Gone is the £5 slot ceiling, along with the £2 one that applies between 18 and 24. Gone is the 2.5-second floor between cycles. Gone too is that line about not being an ombudsman, there being no regulator in the picture to say it.

What is not gone is the duty position in part six. An operator outside the UK that offers remote gambling to a person who usually lives here is within the scope of British gaming duty regardless of its licence, so the border that stops every protection above does not stop the charge. That is the whole asymmetry in one sentence, and it is the only honest reason to put the ranking behind the arithmetic rather than in front of it.

Also unchanged is the national self-exclusion scheme’s reach, or rather its limit. Belonging is compulsory for each online gambling firm that holds a British licence, so one registration covers every licensed account. An unlicensed site is outside that scheme entirely, and the scheme itself says so. Registration does not carry a reader as far as an unlicensed operator, and no sentence on this sheet will assist anyone in defeating a self-exclusion, a blocking tool or a block set by a bank. If a block is in place, the right move is the helpline in the panel below this sheet rather than a list of sites.

Read the ranking, if it is read at all, as a record of what a particular kind of offer looks like when none of parts two, three and six applies to it. The £76.80 is still £76.80 of arithmetic. It is no longer £76.80 that anything obliges anybody to pay.

Eleven named absences on one side, and the duty charge on both

10

The sheet in one column

Everything above reduces to a short list of figures, and a reader who keeps only the list has kept the sheet.

Eight hundred spins at 10p is £80.00 of staking. Set a 96 per cent configuration against that and the expectation is £76.80 back with £3.20 of cost, four tenths of a penny each press. At the lawful pace floor it is 33 minutes and 20 seconds. Change the price and every one of those figures moves by the same factor; change the tally and they move in proportion; change the return by one percentage point and the cost moves by 80p across the tally, because one point of 800 spins at 10p is exactly 80p.

That last relationship is the most useful one on the sheet, so here it is in full. A return of 96 per cent costs £3.20 across the tally. Ninety-five costs £4.00. Ninety-four costs £4.80. Ninety-seven costs £2.40, and ninety-eight £1.60. Six points of band, 92 to 98 per cent, carry the cost from £6.40 down to £1.60, a spread of £4.80, and no licensed site is obliged to publish which end of that band a given game sits at. Four alternatives satisfy the information duty, only one of them is a percentage, and the word between them is or.

Against that spread of £4.80, moving the spin price from 10p to 25p moves the cost from £3.20 to £8.00, a spread of £4.80 as well. The two levers are the same size at those settings, and one of them is printed on the screen in front of the reader while the other may not be published anywhere. A comparison that reads the published lever and ignores the unpublished one is still a better comparison than one built on adjectives, which is the modest claim this sheet makes for itself.

For the smallest payments a cashier will take, the three deposit sheets run the same arithmetic downward: a one pound deposit buys ten spins at this price, three pounds buys thirty, and five pounds buys fifty. For the other direction, the high roller sheet prices the ceiling and the clock together. The register itself, and what a start date on it does and does not prove, is on the new sites sheet, and the duties attached to a small screen are on the mobile sheet. All of them sit under what a spin is worth.

One percentage point of return is 80p across 800 spins at 10p