AI betting predictions: how they work, what they cost, who really offers them

Updated October 2026

Licensed

Available in US

Fast payouts

18+ Only

27 October 2026 · Checked against the Gambling Commission’s public register of remote operating licences

Anyone selling AI predictions has a price. That price is rarely the subscription. The price is the margin the bookmaker has already built into every quote the model points at, the turnover the welcome bonus demands, and the deadline the free bet expires against. The focus is simple: not whether AI predictions exist (they do, in a dozen guises), but what they cost the punter who follows them — the explicit fee, the implicit one, and the one that shows up in the odds themselves. The arithmetic here provides the answer. Recommendations are deliberately absent: no model has been proven to beat the market, and promising one would be the single dishonest path.

A prediction is one input into a betting decision, never the decision itself. A betting decision is a stake against a price, the price being the bookmaker’s view of the probability of an outcome minus its margin. The punter who follows AI predictions without understanding that second layer is paying twice: once to the model, once to the overround. The pages that follow look at each layer in turn — what the model can and cannot do, what the bookmaker’s margin does to any price the model produces, what the welcome offers actually demand of a new account, and which of the ten licensed operators the page surveys fit which kind of punter. None of that is a recommendation to bet. It is the work that has to happen before a bet is worth placing.

Safeguards the reader cannot do without

A page about predictions is, before anything else, a page about money that can be lost. Gambling in Great Britain is licensed and regulated for a reason, and the protections around it exist because the harms are real and well documented. The first section of this guide is therefore the one the rest assumes: the reader knows where the off-switches are, and is not betting money needed for rent, bills, or anything else.

The duty that sits with the operator, not the bettor

The Gambling Commission licenses every commercial bookmaker taking remote bets from customers in Great Britain, and the Gambling Act 2005 is the statute the licences are issued under. Section 33 of that Act makes it an offence to provide gambling facilities in Great Britain without the licence the Act requires — aimed at the operator, not the customer. Northern Ireland sits outside the Act and runs an older law, so a GB licence does not cover bets placed from there. None of that makes betting safe. It places a duty of care on the operator to identify the customer, verify age, offer self-exclusion, run affordability checks at higher spend, and route complaints through an independent adjudicator. The punter’s side of the same bargain is the safeguards below.

The four safeguards every licensed operator must offer

Every bookmaker in the ranking later in this article is licensed by the Gambling Commission and bound by the same social responsibility code. The four protections the punter encounters first are deposit limits, time-outs, self-exclusion, and reality checks. Deposit limits are set by the punter, not the operator: pick a ceiling, the system enforces it, and changing it upwards usually has a cooling-off period. Time-outs are short breaks — twenty-four hours, a week, a month — taken from inside the account. Self-exclusion is the heavier version: six months, one year, five years, or five years with auto-renewal, and after the minimum period the exclusion stays in force for a further seven years unless the punter asks GAMSTOP to remove it. Reality checks are the pop-up that interrupts a session after a chosen interval and reports net position. None of these is a moral judgement; all of them are tooling that exists because the alternative is harm.

GAMSTOP itself is the national multi-operator online self-exclusion scheme and a mandatory licence condition for online operators in Great Britain since 31 March 2020. One registration covers every licensed site and app. Betting shops run a separate scheme. A punter who has signed up to GAMSTOP cannot open an account at any licensed operator for the chosen period — which is why the Commission treats it as the central pillar of online protection. The same register is the test of whether a brand holds a licence at all, and the public register can be searched by business name, trading name, domain, or account number.

The financial vulnerability check, and the one that is still being phased in

Since 28 February 2025, a financial vulnerability check has applied to licensed remote betting when a punter’s deposits minus withdrawals exceed £150 in a rolling thirty-day window. The check is a public-record check (bankruptcy, county court judgments, similar), not an income check and not a credit reference check. It is the lightest of the affordability measures the Commission has signalled; the heavier “financial risk assessments” the press calls affordability checks were announced on 7 July 2026 in a staged plan with no start date. The phrasing matters: the announced measure is “staged, with no start date”, not in force, and a page that writes otherwise misleads on a control the punter may not yet be subject to. Deposit limits, meanwhile, are themselves in transition: a phased regime has been under consultation and the current status is best described in qualitative terms.

Help that is one tap away

The licensed operators named later in this guide link to the same set of help services. GamCare runs the National Gambling Helpline, twenty-four hours a day. GambleAware funds treatment and research. The NHS runs gambling clinics in several regions. Gamblers Anonymous runs meetings in person and online. GamCare, GAMSTOP, and the others are independent of any operator; an honest bookmaker does not steer a struggling punter away from them and an honest page does not steer a struggling punter towards betting.

Phones, apps, and the shape the bettor works with

Most bets at the operators surveyed are placed from a handset, not a desktop. The decision about which bookmaker to use is, increasingly, a decision about which app behaves well on the screen the punter actually has. That is a different question from which bookmaker has the deepest market, and it is the one this section answers.

What “app” means at a British bookmaker

A licensed bookmaker in Great Britain publishes three things that all live behind the same login: a website, an iOS app, and an Android app. The website is usually the same product as the app with a different shell, and the differences that matter are layout (touchscreen versus pointer), the use of push notifications, and whether features such as live streaming or cash out survive the cut-down mobile interface. Every operator in the ranking below publishes on both stores; the variations are in how the same product reads on the screen the punter has.

A handful of features are common to every credible British bookmaker app in 2026. Touch ID or Face ID to log back in. A bet slip that survives being minimised so the punter can check another market without losing the selection. Cash out as a button on open singles and accumulators, often with a partial cash out. Push notifications for goals, settled bets, and price boosts — usually tunable, ideally switched off for the kinds of alert that nudge a punter back when they should have stopped. None of that is exotic. What separates one app from another is how cleanly those features are wired, how many of them survive a flaky 4G connection, and how forgiving the bet slip is when the punter mis-taps.

Live betting on a phone is its own screen on the home screen

In-play is the most-used feature on a British bookmaker app and the most punishing for an interface. Prices move every second on a Premier League match; the screen has to update without locking; the cash out number has to keep up; the bet slip has to remain responsive while three other markets are open. Most apps pass that test for football, tennis, and basketball — the three sports the live product was built around. The weaker ones show their seams in cricket, where over-by-over pricing exposes any lag, and in lower-league football, where the data feed is thinnest.

Two design choices are worth checking before a deposit goes in. First, whether the in-play screen offers a one-click stake on the current price, or whether the punter has to open the bet slip, set the stake, and confirm — the extra taps are the difference between catching a price and missing it. Second, whether the bet slip allows the punter to set a default stake that survives across markets, so the punter is not typing £5 into a slip every time the price changes.

Live streaming that is actually free

Live streaming is the line between a serious bookmaker and a casual one, and the rules around it vary more than the marketing suggests. Some operators stream UK and Irish horse racing, selected football leagues, and a long tail of tennis and basketball free to anyone with a funded account. Others stream only to punter who has placed a bet in the last twenty-four hours, or only on races the punter has a wager on. The clearest version, which appears in the ranking later, requires nothing more than a logged-in account and an occasional bet in the previous day. The least clear version is the one where the schedule is hard to find and the eligibility criteria are tucked into a help page two menus deep.

The same logic applies to live trackers. The football equivalent of a shot map and a stat ticker is now standard at the top end of the market and is the cheapest way to keep an in-running bet readable when the picture is poor. Cricket trackers — wickets, run rate, required rate — are rarer and a clear sign that the operator takes the sport seriously. A punter who bets in-play should check the tracker before the stake, not after.

Browser, desktop, and the second screen

A British punter who bets in-play usually ends up with two screens: the app on the main handset, and a desktop browser showing a different market or a different bookmaker. The bookmaker that takes that workflow seriously supports both. The one that takes it less seriously forces the punter to use the app only and gives the website a stripped-down view, or vice versa. Neither is right or wrong, but the choice matters for the punter who wants to compare prices across two bookmakers during a match without tapping between two apps on the same phone.

How a prediction is built, and why no tip is ever certain

A prediction is a model output dressed as advice. The model can be anything from a regression on the last ten seasons of results to a recurrent neural net fed live possession and pass data, and the marketing around it changes faster than the substance. What does not change is the relationship between the model’s probability and the price the punter has to beat. That relationship is the only thing on the page that matters.

The four inputs every working prediction uses

Whatever the marketing says, every usable prediction rests on the same four inputs: form, statistics, line-ups, and the market’s own odds. Form is the short run of results; statistics are the longer-run metrics built from them; line-ups are the human inputs that move form and statistics at the margin (a rested striker, an injured keeper); and the market’s own odds are the consensus of every other punter and bookmaker expressed as a price. A prediction that uses one of those inputs alone is, in the language of the trade, a thin model; a prediction that uses all four is a thick one. The thickness, not the cleverness, is what beats the market — when anything does.

What an AI adds, when it adds anything, is the ability to weight and combine those inputs at a scale no individual punter can match. A neural net can absorb every completed match in a league, every shot, every expected-goals metric, and find patterns that a human analyst with a notebook will miss. The catch is that the patterns the network finds are not the patterns that pay the punter’s bills next week. They are the patterns in the data it was trained on. A model trained on five seasons of Premier League results will not see a transfer window that just closed; it will see the data it has. A punter who treats the model as a black box pays for the model’s blind spots.

The bookmaker’s own price is the hardest input to ignore

The bookmaker’s price is the only input on this list that already incorporates every other input every other bookmaker has. The exchange market, in particular, is a near-perfect aggregator: the back and lay prices at any moment are the live market view of probability. A prediction that disagrees with that view by a small amount is just noise; a prediction that disagrees with it by a large amount is either a genuine edge or a mistake, and the punter has no way to know which without doing the work the model claimed to do. Most edges in the literature turn out, on closer inspection, to be mistakes. The model is not the punter’s friend here; the discipline of asking whether the model’s probability and the market’s price really disagree by enough to justify the stake is.

Value is the difference between two numbers, and that is all it is

A value bet is a price whose implied probability, once the bookmaker’s margin is taken out, is below the punter’s estimate of the true probability. The arithmetic is mechanical. A bookmaker offers 5.0 on a tennis player. Implied probability, before margin, is one divided by 5.0, or twenty per cent. The bookmaker’s margin on that market is, say, five per cent. Strip the margin out across the other side and the true probability might be twenty-two per cent. The punter’s model puts the player’s chance at twenty-five per cent. The price offers value, by three percentage points, and that is the entire edge.

It is also a thin edge. Three percentage points sounds like a lot in the abstract. It is, in the real round, one winning bet in thirty-three that the punter would not otherwise have caught. A bankroll built on three percentage points needs a thousand bets to compound meaningfully and a hundred to register as anything other than variance. Most punter do not have a thousand bets in them on a single market. The model that promises three percentage points every week is, charitably, optimistic.

Staking is where most prediction-led bankrolls die

A prediction without staking is half a system. The punter who follows a model at a flat £10 stake on every tip will lose to the punter who follows the same model at one per cent of bankroll, because the second punter survives a losing run that bankrupts the first. The maths is the maths: a one per cent stake at decimal odds of 3.0 wins £2 for every £1 risked; a ten per cent stake wins £20 for the same bet, and the difference shows up only when the losing run arrives, which it will.

The staking plans in circulation — Martingale, Fibonacci, fixed percentage, level stakes — are not predictions and they do not make a bad prediction good. They are ways of managing the variance that a real prediction produces. The punter who reads a model’s tips and stakes half the bankroll on the first one has not understood the page. The punter who reads the same tips and stakes one per cent has.

Why no tip is ever certain

A tip is a probability, not a prediction in the strict sense of the word. A probability of seventy per cent is not a certainty; it is a seventy per cent chance, which loses three times in ten. A punter who reads a model output of “seventy per cent” and stakes the rent has misunderstood the sentence. The model is the most useful when its outputs are taken as probabilities and staked as probabilities. The model is the most dangerous when its outputs are taken as tips and staked as tips. The difference is the discipline of treating a percentage as a percentage, not as a guarantee.

How the bookmaker’s margin eats any prediction

Every price a bookmaker offers is the market’s view of probability minus the bookmaker’s margin. The margin is the cut the bookmaker takes on every pound staked, expressed as a percentage of the total pool. It is invisible in the marketing but visible in the math, and it is the single biggest cost the punter pays. The following arithmetic reveals what the predictions are competing against.

The two-way market as the cleanest example

The cleanest example is a two-way market, like a tennis match winner, where the bookmaker offers two prices — one on each player. Take a market priced at 1.90 on both sides, the decimal equivalent of 9/10, the kind of line that sits on a coin-toss game with a slight house preference. The implied probability of each side, ignoring margin, is one divided by 1.90, or roughly 52.63 per cent. Add the two together and the sum is 105.26 per cent. The five-point-twenty-six per cent above one hundred is the bookmaker’s margin on the market. The punter who bets on either side is paying it, and the price the punter receives has already been reduced by it.

That five-point-twenty-six per cent is the round number to remember. On a market at 1.90 / 1.90, the bookmaker keeps a little over five per cent of every pound staked. It is a typical margin for a major football or tennis market at a strong bookmaker. It is not the lowest margin available — sharp bookmakers and exchanges will run a tennis match closer to 103 per cent — and it is not the highest. It is the band within which most major-market prices sit, and the band against which any prediction has to clear.

Reading the margin from any set of prices

The same arithmetic works on any number of outcomes. Take a three-way football match, priced at 2.10 on the home side, 3.40 on the draw, and 3.60 on the away side. The implied probabilities are 47.62 per cent, 29.41 per cent, and 27.78 per cent, summing to 104.81 per cent. The margin is 4.81 per cent. Take a market the bookmaker has priced more aggressively — say 1.40 on the favourite, 7.00 on the outsider in a one-sided tennis match — and the implied probabilities sum to 71.43 plus 14.29, or 85.72 per cent, which is below one hundred and signals that the bookmaker has mispriced the market. That does not happen often. When it does, it is a genuine error the punter can exploit by laying the favourite at a sharper bookmaker.

The point of the arithmetic is not the calculation itself — most modern punters have a margin calculator in their pocket — but the habit of running it on every market before staking. A punter who checks the margin once a month on their main markets will spot the bookmaker who is consistently tighter than the field. A punter who never checks is paying the average margin on every bet, which is the highest tax they pay.

What the margin does to a model’s prediction

A model that puts a player’s chance of winning at sixty per cent, and a bookmaker price that implies sixty-three per cent before margin, looks like an edge. Strip the five per cent margin out and the book’s true implied probability is closer to sixty per cent, exactly matching the model. The edge was the margin. The punter who saw an edge and bet it has lost to the margin they did not see. This is the most common way a model looks good in retrospect and loses money in the account: the model is accurate, the bookmaker is accurate, and the gap between them is the margin.

The corollary is that the only prediction that pays is one whose probability differs from the market’s stripped-of-margin probability by more than the punter’s staking error. A one-percentage-point edge, after a one-percentage-point error in staking or commission, is no edge at all. A five-percentage-point edge is the kind of number the punter can actually compound, and the kind of number almost no published model produces reliably. The published model is, more often than not, the marketing.

The exchange as the cleanest price

A betting exchange runs on a different model. Back and lay prices are set by the punters, with the exchange taking a small commission on net winnings. The margin on an exchange market is the difference between the best back price and the best lay price, plus the commission. On a heavily-liquid football match the overround is closer to two per cent than five, and the commission is two to five per cent of net winnings. The price the punter sees on the exchange is the closest thing in the market to a true probability. A model that disagrees with the exchange price has to disagree with every other punter on the exchange. That is the standard to beat.

The exchange is not a free lunch. The commission is real, the liquidity at smaller events is poor, and the price the punter sees is the back price, which is always slightly worse than the mid-price. None of that makes the exchange worse than a traditional bookmaker on margin. It makes it better on margin and worse on liquidity, and the punter has to pick the trade-off for each market.

Why the welcome offer is not a free hit

The welcome offers attached to the ten licensed operators later in this guide are often the punter’s first encounter with a margin larger than five per cent. A free bet that pays only the winnings, not the stake, is the most common shape. A £10 free bet at odds of 3.0 returns £20 in profit, not the £30 a cash stake would return. The £10 the house kept is a margin dressed as a gift, and the punter who treats the offer as a free hit is paying it.

The compounding effect is real. A punter who claims ten welcome offers in a year, each at the headline value, may see £400 in bonus money clear through the account. The cost is the margin on the qualifying bets (often at minimum odds to count toward turnover) plus the value the free bet gives up by not returning the stake. Done well, the welcome offers are a small profit. Done badly, they are a marketing exercise in earnest. The punter who understands the margin sees the welcome offers as a transaction, not a gift. The punter who does not sees them as a discount on the next deposit.

Ten licensed bookmakers, ranked for the punter who follows models

The ten bookmakers surveyed here are drawn from the Gambling Commission’s public register of remote operating licences. Each one holds a licence for remote general betting standard (real events); each one is bound by the same social responsibility code, the same advertising rules, and the same financial vulnerability check above £150 in net deposits over a rolling thirty days. The page describes what each operator offers; it does not rank them by quality, because quality in this market is a function of the punter’s own habits, and what suits one punter’s staking plan is wrong for another’s. What follows is, instead, a working assessment of where each one fits in the punter’s day.

The column headed “Licence holder” carries the name on the Commission’s public register against the licence number in the next column. The “What the operator’s own terms state” column is drawn from each operator’s published terms and from the App Store listings published on 16 October 2026; it is what the operator says about itself, not what the page has concluded. The verdict at the end of each write-up is the page’s own, and it answers the question the section above it set up.

BetGoodwin — the specialist’s racing book

BetGoodwin sits on a Gambling Commission remote licence held by Goodwin Racing Limited and presents itself as a racing-led bookmaker. Its own published terms describe Best Odds Guaranteed on all UK and Irish horse and greyhound racing, up to £1,000 — a larger cap than most mainstream books, and the clearest signal of where the operator’s product is built. A racing specialist does not need to be the punter’s only bookmaker, but it is the bookmaker whose Best Odds Guaranteed ceiling and whose ante-post pricing the punter checks before a Saturday card. The shape suits a punter whose main interest is the racing side of the market and who is happy to keep a second account at a more generalist book for football and tennis. The licence register confirms the operator holds the relevant category; the racing focus is the operator’s own commercial choice.

Betway — the football-and-casino mainstream

Betway sits on a Commission licence held by Betway Limited and is the textbook mainstream bookmaker: a wide sportsbook, a deep casino, an app, and a welcome offer that pays £40 in free bet tokens (4 x £10) on settlement of a £10+ first bet at minimum odds 2.0, with a 7-day expiry and debit-card deposits only. The condition that limits the offer to debit card deposits is a small but real one: a punter who normally funds the account through an e-wallet or bank transfer will need to change the deposit method to claim the headline value. The free bet tokens are paid as four separate £10 bets, not as a single £40 free bet, which means each token is staked on its own market and the stake is not paid back with the win. The qualification is a straightforward entry point for anyone comfortable with a debit card, and it makes the app a solid one-stop shop for a player who doesn’t want to juggle accounts.

bet365 — the in-play benchmark

bet365 sits on a Commission licence held by Hillside (UK Sports) ENC and is the bookmaker other operators measure their in-play product against. The site describes in-play betting and live streaming of events worldwide, plus Bet Track, Match Live and a Live Golf Tracker for following bets in play. The breadth of the live-streaming catalogue is the single biggest reason a punter keeps a bet365 account: the football coverage runs from the Premier League down to lower leagues in dozens of countries, and the tennis schedule runs from the majors down to the ITF circuit. The in-play product is fast, the cash out is dependable, and the bet builder is competitive. The shape suits the punter who bets in-play on football or tennis and is willing to fund the account by debit card or bank transfer to claim the welcome offer.

William Hill — the high-street name with the streaming rule

William Hill sits on a Commission licence held by WHG (International) Limited and is the longest-established name in the list. Its own terms describe Best Odds Guaranteed on UK and Irish horse racing from 8am on the day of the race, on both single and multiple bets, with ante-post, tote and pari-mutuel bets excluded. The streaming rule is the more interesting constraint: WHTV requires a funded account or a bet in the last twenty-four hours, and live racing requires a bet of £1 or more on a qualifying race. The Betting TV service shows selected UK and US horse racing and UK greyhounds free. The shape suits the punter who bets racing primarily, who values the high-street brand’s consumer protections, and who is comfortable with the once-a-day bet that unlocks the live pictures.

Lottoland — the lottery-led hybrid

Lottoland sits on a Commission licence held by EU Lotto Limited and presents a different product from the rest of the list. The app combines casino, lottery betting, sports betting and scratchcards, and the operator states it is licensed by the Gambling Commission for customers in Great Britain. The shape is not a sportsbook with lottery bolted on; it is a lottery operator with a sportsbook attached. A punter who already has a main sportsbook and is looking for a second account to cover lottery products will find Lottoland useful. A punter whose primary interest is football or racing will find the sportsbook thinner than the specialist names. The licence is real; the product mix is the operator’s positioning.

VBet — the in-play feature stack

VBet sits on a Commission licence held by SCGO Limited and the App Store listing describes Bet Builder, pre-match and in-play betting, live streaming, and a live calendar with results and stats. The shape is the modern sportsbook template: a deep in-play product, a Bet Builder for same-game multiples, live streaming across a wide catalogue, and a results-and-stats overlay that turns the app into a research tool as well as a betting one. The shape suits a punter who wants a single app for pre-match and in-play, and who values the data layer that lets them run their own numbers alongside the bookmaker’s. The depth of the streaming catalogue varies by sport and event; the punter checks the schedule before staking.

Midnite — the new-entrant welcome

Midnite sits on a Commission licence held by Midnite Limited — one of the newer entries on the Commission’s register, with account number 101839. The app shows in-play events, an acca builder, popular accas and price boosts with maximum-stake limits, and the welcome offer pays £25 in free bets and 50 free spins to new customers who bet £5+ on pre-match odds of EVS (2.0) or greater within 14 days of sign-up, with the free bets and spins valid for 7 days. With the bonus paid as free bets and spins on a low entry stake, it is a low-risk way to test a new sportsbook and see how it performs compared to an established favourite.

Sky Bet — the broadcast-tied mainstream

Sky Bet sits on a Commission licence held by SBG Sports Limited and is the sportsbook arm of the broadcaster. The welcome offer pays £30 in free bets as three £10 tokens after a £5 deposit and a first single or each-way bet of 5p at odds of 1/1 or greater. The low qualifying threshold means a punter can secure the bonus with almost any bet, making it one of the most accessible offers on the market for a casual player.

Tote — the pool-betting specialist

Tote sits on a Commission licence held by TDCO Limited and is the pool-betting arm of the British tote system. The product is structurally different from a traditional bookmaker: the punter is betting into a pool with other punters, the dividend is declared after the race, and the house take is a fixed percentage of the pool rather than a margin built into every price. The shape suits a punter who already bets into pools at the track and wants the same product online, and who understands that pool betting is a different product from fixed-odds betting, not a substitute for it. The licence is for pool betting as well as fixed-odds; the brand is the pool.

Paddy Power — the features-and-markets bookmaker

Paddy Power sits on a Commission licence held by PPB Counterparty Services Limited and the App Store listing describes Bet Builder+, Popular Bet Builders, Super Sub (a bet rolls over to the replacement if a selected player is substituted), expanded Win More markets, and player and ACCA stats. Super Sub is the headline feature — a same-game accumulator that survives a substitution, which most bookmakers void on — and the depth of the Bet Builder product is the clearest signal that the operator takes the same-game market seriously. The shape suits the punter who bets accumulators on football, who values a Bet Builder that handles the corner cases, and who is comfortable with the Paddy Power brand voice on the marketing.

The legal frame every punter needs before staking

The licence number on a bookmaker’s app is the first thing to check and the easiest thing to verify. The Commission’s public register can be searched by business name, trading name, domain or account number; a brand that does not appear there is unlicensed for the purposes of Great Britain, and the consequences of using an unlicensed brand are real even if no offence is aimed at the customer.

Why a GB licence is compulsory at the point of consumption

Since the Gambling (Licensing and Advertising) Act 2014, an operator whose remote facilities are used in Great Britain needs a Commission licence wherever it is based. Sections 33 and 36 of the Gambling Act 2005, as amended, set the offence. A Curaçao, Maltese, or Gibraltar licence is not a substitute. None of that makes a GB-licensed bookmaker safe — the operator can still refuse a bet, withhold winnings in a dispute, or go out of business — but it makes the operator subject to the social responsibility code, the bonus cap, the financial vulnerability check, the GAMSTOP obligation, and the Commission’s complaints route. The punter who uses a licensed operator has the protection of all of those. The punter who uses an unlicensed brand has none of them.

There is no state monopoly and no cap on the number of licences. Several brands can run under one licensee — Ladbrokes and Coral, Paddy Power and Betfair are the textbook cases — and a page that treats them as independent operators misleads. The licence register lists more brands than any article can feature, so no brand is “the only” licensed one.

The categories of licence and what they mean

The Commission issues licences in five categories relevant to a punter: remote general betting standard (real events), remote general betting standard (virtual events), remote general betting limited, remote pool betting, and remote betting intermediary (the licence for exchanges). The real-events standard licence is the one every operator in this guide holds and is the category that covers betting on football, tennis, horse racing, and the rest of the conventional sportsbook. Pool betting is the category Tote operates under for its core product. The exchange category covers Betfair and the smaller exchanges. None of these licences covers casino, which sits under a separate regime; a betting limit is never a casino limit, and a betting licence does not authorise the operator to take casino bets.

What may and may not be bet on

No statutory list of permitted bets or markets exists. An operator holds a licence for a category and offers markets under its own rules and the Licence Conditions and Codes of Practice. In-play betting is offered by GB bookmakers and is governed by the operator’s licence conditions and customer-protection duties; it is not legally unrestricted, and a punter who treats it as such has not read the small print. Esports are treated like any other event; virtual events sit in their own category; politics, TV, and novelty markets are offered and no statutory prohibition was found. Spread betting that is not a financial product is inside General Betting Duty. Pool betting on horse or dog racing sits under a separate category. No statutory cap on stakes in sports betting exists; operators set their own maximum stakes and may restrict customers.

Credit cards, identity, and the verification the punter cannot avoid

Credit cards have been banned for gambling in Great Britain since 14 April 2020, including credit-card-funded payments through an e-wallet; debit cards and bank-based payments are not covered by that ban. The punter who tries to fund a betting account with a credit card will be blocked at the deposit step. Identity verification before the first deposit has been compulsory since 7 May 2019 — name, address, date of birth — and the verification covers free bets and bonuses, not just real-money play. The punter who tries to register with a false name or a borrowed card will not get past the first deposit. Source-of-funds questions at higher spend are normal under the money-laundering rules.

Bonuses, advertising, and the cap the punter should know about

Since 19 January 2026, wagering requirements on bonuses are capped at 10x — a £10 bonus at 50x turnover needed £500 of bets before the cap, and under the cap no more than £100. Mixed-product promotions — an offer that requires gambling on two or more product types such as betting, casino, bingo, and lottery, for example “bet on football to get free spins” — are banned. Whether the cap applies to free bets with no wagering condition, and any cap on the money value of a bonus, are not established. The advertising rules sit with the ASA and the CAP Code, not the Commission: marketing must be socially responsible, must not be of strong appeal to under-18s, and no one who is or seems to be under 25 may be featured gambling or play a significant role. The “whistle-to-whistle” ban on TV ads around live sport before 9pm is an industry code, not law.

Tax the punter does and does not pay

The punter pays no tax on betting winnings in Great Britain; general information, not tax advice, and a professional gambler should ask HMRC or an adviser. General Betting Duty is the operator’s tax, currently 15%, and from 1 April 2027 a new remote betting rate of 25% applies within it; until that date the remote rate stays at 15%, and whether operators pass duty on through odds is not established. Spread betting has its own rates — 3% financial, 10% other. The Commission’s statutory levy is a separate charge, not a tax, and is paid by the operator. The punter’s net position is what the bookmaker pays out, less the stake.

Offshore, unlicensed, and what the punter gives up

A GB licence is compulsory at the point of consumption. Providing gambling facilities in Great Britain without the licence the Act requires is an offence under section 33 of the Gambling Act 2005; enforcement is disruption, not blocking — cease-and-desist notices, search-engine referrals, work with registrars, hosts, and payment providers, with court-order DNS blocking powers granted or about to be granted. No offence aimed at the customer who merely bets on an unlicensed site was found. What the customer gives up is the entire protection stack: no GAMSTOP coverage, no Commission complaints route, no approved dispute-resolution body, no British consumer redress if funds are withheld, no protection under the 10x bonus cap or the £150 vulnerability check. The honest framing is licensed = checks, limits, GAMSTOP, bonus cap, but also verification before the first bet; unlicensed = none of those protections and possible loss of funds. Neither is “the safe choice” by default.

Reading the betting slip, and what each line actually means

A betting slip is a contract, and the punter who treats it as a screen on the way to confirming a stake has not read the contract. Three elements on the slip deserve attention: the odds, the stake, and the return. The rest is decoration that does not change the price.

Single bets: return, profit, and the stake the house keeps

A single bet is a stake against a price on a single outcome. The return is the stake multiplied by the odds. A £10 stake at decimal odds of 3.0 returns £30. The profit is the return less the stake, or £20. The stake is paid back with the win. That last clause matters because it does not apply to a free bet, where the stake is usually not paid back with the win; a £10 free bet at 3.0 returns £20 in profit, not £30. The bookmaker’s terms state the position for each offer; the punter checks before claiming.

Each-way bets: two stakes of equal size

An each-way bet is two bets of equal stakes: one on the win, one on the place. The stake doubles. The place terms — typically a quarter of the win odds for places one to four in a flat race, or one fifth for places one to five in a handicap — vary by race and by operator. The punter who reads “each-way £5” is staking £10 in total. The return calculation runs twice: the win stake at the win odds, and the place stake at the place odds. The punter who wants the each-way version of a bet reads the place terms on the operator’s site before confirming.

Accumulators: combined odds, combined stakes

An accumulator combines the odds of the legs into one price and stakes the punter’s single stake against the combined price. Three legs at 2.0, 3.0, and 4.0 produce combined odds of 24.0. A £10 stake returns £240 if all three win, and nothing if any one loses. The bookmaker’s margin on the accumulator is the compounded margin on each leg, which is why accumulators are systematically worse value than singles; the punter who builds accumulators for the thrill is paying the compounded margin for the privilege. Some operators offer acca insurance — a stake back as a free bet if one leg lets the punter down — and the punter reads the insurance terms before claiming.

Cash out and the price the house offers

Cash out is the bookmaker’s offer to buy the bet back before it settles. The price is set by the operator and is always worse than the true value of the bet at the current market price; the gap is the operator’s cut on the cash out. The punter who cashes out a winning bet at the ninetieth minute locks in a profit rather than waiting for the final whistle, and pays the operator for the lock-in. The punter who lets the bet run accepts the variance. Neither is right or wrong. The cash out is a tool, and the punter who uses it without checking the offered price against the live market has paid more than the tool costs.

What a welcome offer actually demands, and what it actually pays

A welcome offer is a marketing hook attached to a marketing calculation, and the punter who reads only the headline number has read half the offer. The arithmetic the offer hides is in three places: the qualifying bet, the wagering requirement, and the value of the free bet itself.

The qualifying bet the offer demands

The qualifying bet is the stake the punter has to place, at the minimum price the offer names, before the free bet lands. A £10 free bet at minimum odds of 2.0, on a first bet of £10 or more, is the standard pattern. The £10 stake has to win or lose at 2.0 or greater; the punter cannot place a £10 at 1.5 and claim the free bet. The qualifying bet is also the bet on which the bookmaker’s margin sits, and the punter who treats the qualifying bet as a free hit is paying the margin on it. The offer is a transaction, not a gift, and the qualifying bet is the entry fee.

The wagering requirement the offer demands, when it has one

Some welcome offers carry a wagering requirement on the free bet winnings — turn the winnings over a set number of times before they withdraw. The 10x cap on wagering requirements since 19 January 2026 limits how heavy that turnover can be; a £10 bonus at 50x needed £500 of bets before the cap, and under the cap no more than £100. The cap is a protection the punter benefits from; the wagering requirement is the cost the punter pays. A free bet with no wagering condition is the cleanest offer; a free bet with a 10x condition on the winnings is a less clean offer with a smaller headline value.

The free bet’s stake is usually not paid back with the win

A £10 free bet at decimal odds of 3.0 returns £20 in profit, not £30. The £10 the house kept is the cost of the free bet, and it is the reason a free bet’s headline value is not the punter’s profit. The free bet is worth, on a value-bet selection at odds of 3.0 with a five per cent margin, around £10 × (3 − 1) ÷ 3 × 0.95, or about £6.33. That is the value the punter should price the offer at, not the £10 headline. A punter who chases welcome offers without pricing them is paying more in qualifying-bet margin than the free bet returns.

Stacking, sequencing, and the punter’s annual haul

A punter who claims the welcome offer at each of the ten operators in this guide, and runs the qualifying bets and free bets cleanly, will collect somewhere in the region of £300 to £400 in headline bonus value over a year. The real number is lower after the qualifying-bet margin and the free-bet stake haircut. The real number is positive, after cost, for a punter who is opening new accounts anyway. The real number is negative, after opportunity cost, for a punter who is opening accounts just for the offers. The page does not recommend either route; it states the arithmetic so the punter can decide.

Building a bankroll that survives the variance

A bankroll is the money the punter has set aside for betting, and the discipline of managing it is the difference between a punter who is still betting in five years and one who is not. The arithmetic is mechanical; the discipline is the human part.

The staking plan, and why one per cent is the default

A one-per-cent stake at decimal odds of 2.0 wins £1 for every £1 risked. A ten-per-cent stake at 2.0 wins £10 for the same bet. The difference shows up only when the losing run arrives, which it will: a punter staking ten per cent of a £100 bankroll at 2.0 can be wiped out in seven consecutive losses; a punter staking one per cent of the same bankroll needs ninety-nine consecutive losses to wipe out, and a losing run of ninety-nine at 2.0 is statistically improbable. The one-per-cent stake is the default because it is the largest stake that survives a normal losing run without breaking the bankroll.

The Martingale plan — double the stake after each loss — is the staking plan the marketing sells hardest and the staking plan that bankrupts the punter fastest. A six-loss losing run at a starting £5 stake requires the seventh stake to be £320, and a seven-loss losing run is not rare at the markets the plan is sold for. The Fibonacci plan is a slower version of the same trap. The fixed-percentage plan is the only staking plan that survives the losing run, because the stake falls with the bankroll and rises with it.

The record the punter keeps, and what it shows

A punter who keeps a record — every bet, the market, the price, the stake, the closing price, the result — has the only tool that distinguishes a lucky streak from a good model. The closing price is the price at the off; the difference between the punter’s price and the closing price is the closing-line value, which is the literature’s best proxy for whether the punter beat the market or the market beat the punter. A punter who beats the closing line over a thousand bets has found an edge. A punter who loses to the closing line over a thousand bets has not.

The record is also the tool that distinguishes a good market from a bad one. A punter who wins on football but loses on tennis has a sport-specific edge, not a general one, and the staking plan should reflect it. The punter who wins on each-way bets but loses on accumulators has a market-type edge, not a staking one. None of this is visible without the record.

The bookmaker comparison that turns the bankroll up

A bookmaker comparison is the punter’s cheapest edge. The same match is priced at 2.10 on the home side at one bookmaker and 2.20 at another. The 0.10 difference is four cents a pound staked, and the punter who takes 2.20 has won four pence for every pound the punter at 2.10 has not. Over a year, the punter who shops the line adds several per cent to the bottom line — more than most staking plans and most model outputs. The discipline is to keep accounts at three or four operators, check the line on every selection, and stake the highest. Three operators is enough to make the comparison meaningful; ten is too many to manage.

The session the punter ends on time

A session is a bounded period of betting — an hour, an afternoon, a Saturday — and the punter who ends the session on time has done the hardest thing in the discipline. The alternative is the tilt: a losing bet leads to a larger bet, which leads to a larger losing bet, which leads to a stake the punter would not have placed an hour earlier. The tilt is the most expensive thing a punter does, and the only defence is the off-switch. The deposit limit is one off-switch; the time-out is another; the session end is the third. None of them is optional for a punter who is still betting in five years.

Picking a model — or deciding not to follow one

A punter who follows a model is delegating the probability estimate to a third party. The delegation is not free, and the punter who treats it as free is the punter the model is built for. The discipline below is the work the punter does before paying for, or following, any model.

What a model can and cannot do

A model can absorb more data than a human, weight inputs more consistently, and avoid the cognitive biases that distort a human estimate. A model cannot see a transfer that hasn’t happened, judge the mood of a dressing room, or update on a news break the way a human can. The work the model does well is the work of pattern recognition across long histories; the work the model does badly is the work of edge cases. A punter who follows a model on the patterns and overrules it on the edge cases has the best of both. A punter who follows the model on everything has outsourced judgement the model cannot provide.

The model that publishes its record is the model worth following

A model that publishes its record — every tip, every price, every result — is auditable. A model that publishes only its wins is not. The published record is the punter’s only check on whether the model is honest about its losses; the unpublished record is the model’s marketing. The punter who pays for a model with no published record is paying for a marketing product.

The record has to be long enough to be meaningful and recent enough to be current. A model that published a thousand tips at the start of the season and has not updated since is using out-of-sample data the punter cannot verify. The punter wants a record that runs across the season and is updated weekly. Anything less is marketing.

The subscription, the free version, and what each actually costs

A model subscription is a real cost, and the punter who treats it as a free input has mispriced the offer. A £20-a-month subscription is £240 a year, and the value of the tips has to clear that £240 in expected profit before the subscription pays. A model with a published record at five per cent return on turnover clears the £240 easily. A model with a published record at one per cent return does not. The punter reads the record before paying the subscription.

A free model is not free. The free model is funded by affiliate revenue, by advertising, or by the model’s owners taking the other side of the bets the model recommends. The affiliate-funded model is the most honest, because the model’s revenue is the punter’s stake at the bookmaker, not the punter’s outcome. The advertising-funded model is less honest, because the model’s revenue is the click, not the punter’s outcome. The “we take the other side” model is the least honest, because the model’s revenue and the punter’s outcome are on opposite sides of the bet.

The decision to follow no model at all

A punter who decides to follow no model is not making a mistake. The decision is a legitimate one if the punter enjoys the work of forming the probability estimate, has the time to do it, and is honest about the biases the work introduces. The punter who follows no model has no subscription to pay and no record to keep, beyond the punter’s own. The discipline below applies the same way: a record of every bet, the closing line as the benchmark, and a bankroll that survives the variance. The model is a tool, and the punter who uses no tool has decided the work is the reward.

The punter’s end-of-day checklist

A punter who has finished the session and is about to close the app has six questions to answer. The questions take thirty seconds and are the only discipline that compounds.

Did the bankroll grow or shrink, and by how much

The first question is the day’s net position. The punter who does not know the day’s net position is the punter who cannot tell whether the staking plan is working. The number is what it is; the question is whether the punter is honest enough to write it down.

Did the closing line beat the punter’s price, or vice versa

The second question is the closing-line check. The punter who beat the closing line on most bets is beating the market; the punter who lost to the closing line is losing to it. The check is the difference between a working edge and a losing streak the punter has dressed as an edge.

Was the stake within the staking plan, or did the tilt take over

The third question is the tilt check. The punter who placed a stake outside the staking plan has paid the tilt tax, and the punter who cannot remember whether the stake was inside the plan has not kept the record well enough. The plan is the only protection against the tilt; the punter who leaves the plan unprotected has left the discipline unprotected.

Was the welcome offer claimed at the right bookmaker

The fourth question is the bookmaker comparison. The punter who checked the line at three operators and staked the highest has done the cheapest work on the page. The punter who staked at the first operator without checking has paid the difference between the highest and the first price, which compounds over a year into a real number.

Were the safer-gambling tools used, or were they left untouched

The fifth question is the safer-gambling check. The punter who set a deposit limit at the start of the month and stayed inside it has done the most important thing on this page. The punter who did not set a deposit limit has left the bankroll unprotected, and the next losing run is the test the bankroll may not pass.

What the next session will look like

The sixth question is the forward-looking check. The punter who has a plan for the next session — the markets to focus on, the stakes to use, the lines to shop — has done the work before the work begins. The punter who has no plan is staking into the next session the way the punter staked into the last one, which is the way the losing run arrived.

The calculation that anchors every margin on the page

The arithmetic below shows the cost of every prediction. The inputs are illustrative, not a quoted price; the result is a band, not a single figure, and the condition sits beside the result.

The setup: an illustrative two-way market

Take an illustrative two-way market priced at 1.90 on both sides — the decimal equivalent of 9/10, the kind of line that appears on a coin-toss game with a slight house preference. The implied probability of each side, ignoring margin, is one divided by 1.90, or 52.63 per cent. Add the two sides together and the sum is 105.26 per cent. The five-point-twenty-six per cent above one hundred is the bookmaker’s margin on the market. The punter who bets on either side is paying it.

The general formula, and what it produces

For any market with n outcomes priced at odds o₁, o₂, …, oₙ, the payout share is the reciprocal of the sum of the reciprocals of the odds, and the margin is one minus the payout share. On the 1.90 / 1.90 example, the payout share is one divided by (one over 1.90 plus one over 1.90), or one divided by 1.0526, which is 95.00 per cent. The margin is one minus 95.00 per cent, or 5.00 per cent — close enough to the 5.26 per cent of the implied-probability sum, with the small difference reflecting rounding. The headline result is the band: a typical major-market bookmaker runs a margin in the region of 4.5 to 5.5 per cent on a two-way market priced around 1.90 / 1.90, with the band widening to 6 to 8 per cent on less liquid markets and narrowing to 2 to 3 per cent on the most liquid exchange events. The condition is that the result describes the bookmaker’s cut on every pound staked, not the punter’s expected return on a single bet.

What the result means for the punter

A punter who bets £100 on a market with a 5.00 per cent margin is paying £5 to the bookmaker on that stake, in expectation, before the punter’s own edge is counted. A model that produces a 3 per cent edge over the market’s stripped-of-margin probability leaves the punter with a 2 per cent expected return on turnover — £2 for every £100 staked, before subscription costs and the tilt tax. A model that produces a 1 per cent edge leaves the punter with a 4 per cent expected loss on the margin alone. The arithmetic is the same arithmetic that produced the welcome-offer calculation earlier on the page; the only difference is the venue. The punter who knows the margin sees the cost of every prediction before paying for it. The punter who does not knows only after the bankroll tells them.

A short note on the operators outside this guide

The ten operators named above are the operators the page features, not the only operators licensed by the Gambling Commission. The Commission’s public register lists hundreds of remote operating licences, and the brands named on the register that are not in this guide are not lesser for the absence. Several well-known names do not appear on this page because the page’s selection is a working set, not a complete catalogue. Spreadex, for instance, is regulated by the Financial Conduct Authority for spread betting and by the Gambling Commission for fixed-odds betting, and states that spread betting losses can exceed the deposit; the spread-betting product is a different one from the fixed-odds product on the rest of the page, and the punter who wants to read about spread betting reads about spread betting on a page about spread betting.

The other named operators in the wider market — betfair, Ladbrokes, Coral, BetVictor, talkSPORT BET, 10bet, Betmaster, Copybet, BetWright, London Bet, Greenlight Bet, the operators behind a dozen further apps — are all licensed for the relevant categories and all are governed by the same Commission rules. The features vary; the licence is the same. The punter who is comparing operators across a wider set runs the comparison with the same column structure the table used, the same licence check, and the same look at what each operator’s own terms state. The page does not feature them because the plan does not feature them, and the plan’s selection is a working set.

A closing note on what this page is not

This manual covers the market mechanics. It does not name a selection, predict a result, or promise a win. The arithmetic on margins, on welcome offers, on staking plans, and on bankroll discipline is honest about the costs, and the operators named above are described as their own terms state them, not as recommendations. The punter who wants a model to follow should look at the arithmetic to know what the model has to beat. The arithmetic is the tool; the betting decision belongs to the punter. Nothing else on the page carries the same weight.

Frequently asked questions about AI predictions and the betting market

Can AI help with betting?

AI can help with the parts of betting that scale — pattern recognition across long runs of data, weighting the inputs that go into a probability estimate, and avoiding the cognitive biases that distort a human judgement. AI cannot see a transfer that just happened, judge the mood of a dressing room, or guarantee any single bet. The work the punter delegates to AI is the work the punter does not have to do; the work the punter keeps is the work the model cannot do. The discipline is to know which is which on every bet.

Can AI help with sports betting specifically?

Sports betting is the area where AI has the cleanest data: long histories of results, structured statistics, line-ups published in advance. The patterns are real, the data is rich, and the work of pattern recognition is what AI does well. The bookmaker’s margin is the same as it is on any market, the welcome-offer arithmetic is the same, and the staking discipline is the same. The punter who uses AI on sports betting still pays the margin, still needs the bankroll discipline, and still cannot predict a single result with certainty.

Can AI predict the outcome of a sports event?

AI can put a probability on an outcome; it cannot predict it with certainty. A model that puts a player’s chance of winning at seventy per cent is right three times in ten and wrong seven. A punter who stakes the rent on a seventy per cent tip has misunderstood the sentence. The output is a probability; the stake is a stake; the outcome is uncertain until the result.

What is a value bet and why is no tip ever certain?

A value bet is a price whose implied probability, once the bookmaker’s margin is taken out, is below the punter’s own estimate of the true probability. It is the difference between two numbers. No tip is ever certain because every tip is a probability, and a probability of seventy per cent loses three times in ten. The certainty the marketing sells is not in the data; the certainty the marketing sells is the headline, and the punter reads the data, not the headline.

What is the bookmaker’s margin and how can it be seen in the odds?

The bookmaker’s margin is the cut the bookmaker takes on every pound staked, expressed as the gap between the implied probabilities of the outcomes and one hundred per cent. On a market priced at 1.90 / 1.90, the implied probabilities sum to 105.26 per cent and the margin is 5.26 per cent. The punter who adds the implied probabilities of every outcome on a market has read the margin; the punter who has read just the headline price has read half the page.

How is the return on a bet worked out from the stake and the odds?

A single bet’s return is the stake multiplied by the decimal odds. A £10 stake at 3.0 returns £30. The profit is the return less the stake, or £20. A combination bet multiplies the odds of the legs into one price, and the stake is multiplied by the combined price; three legs at 2.0, 3.0, and 4.0 produce combined odds of 24.0 and a £10 stake returns £240 if all three legs win. The free bet does not pay the stake back with the win, and the punter reads the operator’s terms before claiming.

Content created by the «Fast Payout» team