AI betting predictions explained — what they are, what they cost, and how UK punters should use them

Updated October 2026

Licensed

Available in US

Fast payouts

18+ Only

AI has moved into the betting mainstream quickly. Search trends put “ai betting predictions” next to “ai football betting” and “ai betting tips” almost on the same page, and the apps themselves now ship prediction feeds, stat packs and “smart picks” sections next to the price. The question for a UK punter in 2026 is not whether the software exists but whether it makes money, what it sits beside (the bookmaker’s margin, a bonus, a stake limit), and whether anything it produces is what its marketing calls “a tip”.

This guide is a route through that question, and the angle that runs through it is cost — what the prediction, the offer and the licence frame actually take out of a punter’s wallet. The cover themes come first (responsible gambling, mobile, predictions, the comparison table, the legality frame), then the ten reviewed operators, then the workings of a single market and finally the questions readers keep asking.

Current as of 27 October 2026; licence and safer-gambling claims verified against the Gambling Commission’s public register and the operators’ own terms pages.

What a prediction actually is — and why no tip is ever certain

A prediction is a guess about a future event with a probability attached. Form, head-to-head records, line-ups, injuries, expected goals, weather and the bookmaker’s own prices all go into the input, and any one of those can be wrong before kick-off. The point of a prediction is not that it is true; the point is that it lets the punter judge whether the bookmaker’s price is too high, too low or about right. A prediction that does not feed back into the price is decoration.

The harder concept under that is the value bet. A price is fair when it equals the true probability of the outcome, in decimal form. Decimal odds of 2.00 mean a one-in-two chance, 3.00 a one-in-three, and so on. A value bet is one where the punter’s probability is higher than the implied probability of the price. Most bets the average punter places are not value bets once the bookmaker’s margin is stripped out, and that is by design. The bookmaker’s margin, also called the overround, is the cut the book keeps from every market, expressed as a percentage of turnover. On a two-way market at 1.90/1.90, the margin sits in a band the page works out below; on a three-way football market with a draw the band is wider, because an extra outcome leaves more room for the book to edge the prices.

No prediction changes that. A good model narrows the difference between the punter’s probability and the true probability of an event; it does not turn the bookmaker’s margin into a discount. The reader who carries away one sentence from this section: a prediction is a price-check, not a result.

The form of the prediction matters as well. Tips sites and tipsters rarely publish their model’s hit rate the way a fund manager would publish one. They publish recent winners, headline picks and frequency of posting. Both of those can be a marketing line that does not survive an audit. A punter who wants to take predictions seriously keeps their own record: stake, price, market, result, and the closing price at kick-off, because the gap between the price taken and the closing price is the sharpest measure of edge a small punter can capture.

Bankroll discipline is the unglamorous half. A punter decides in advance what unit to stake (often 1–3% of the bank), never chases a loss with a bigger next bet, and writes off any session that drifts above the loss limit for the day. The model, the price and the discipline are three legs of the same stool, and a single tip without the other two is what the bookmaker would call “a customer’s enthusiasm”.

The AI itself, for most of what a UK punter sees in 2026, is one of three things: a statistical model trained on past results, an automated analysis tool summarising news and stats, or an algorithmic scanner of in-play odds that flags drift. None of those has access to information the market does not already have; the edge they offer, where there is one, is speed and discipline, not certainty. A bettor who treats a prediction as one input among several, weights it against the price, and walks away from markets where the model and the book disagree is the bettor who comes out ahead in the long run.

Responsible gambling — the rules that bracket any AI prediction

AI predictions change nothing on the safer-gambling side. The rules that apply to a UK punter in 2026 apply just as much to a tip from an AI as to a tip from a friend, and several of them are statutory rather than optional. They belong at the start because they change what a “winning” pick is worth.

The Gambling Commission is the regulator for commercial betting in Great Britain, working under the Gambling Act 2005. Every operator taking remote bets from customers in Great Britain needs a remote licence, regardless of where the operator itself is based (sections 33 and 36 of the Act as amended by the Gambling (Licensing and Advertising) Act 2014). Northern Ireland is outside that Act, so the rules below apply to GB customers; the GB-licensed operator label points at the part of the UK the Act covers.

The identity check is the first thing the punter meets. Since 7 May 2019, name, address and date of birth are verified before a customer can deposit or gamble at a Commission-licensed site, including with a free bet. There is no anonymous play at a licensed bookmaker. That sounds like friction, and it is; it is also the lever that stops an underage player from getting past the deposit screen. The minimum age is 18 across the board.

Deposits and withdrawals are the second frame. Credit cards are banned for any gambling transaction 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 the ban. The first £20 deposit and the £20 winning bet at Coral, for example, sit inside a debit-card-only frame; the £40 in free bets that follows is paid out the same way. The same pattern repeats across the comparison set below.

GAMSTOP is the national self-exclusion scheme for online gambling and a mandatory licence condition. Any customer can register with GAMSTOP and choose a minimum period of six months, one year, five years, or five years with auto-renewal; once the minimum is up the exclusion stays in place for a further seven years unless the customer asks for it to be lifted. Every GB-licensed online operator is part of GAMSTOP, and a punter who has registered is blocked from opening new accounts or accessing existing ones until the period ends. Betting shops (the high-street tier) use a separate retail self-exclusion. For the punter, GAMSTOP is the most direct of the protection tools on offer.

Financial checks have shifted in 2026. A financial vulnerability check (Licence Conditions and Codes of Practice, condition 3.4.4, in force since 30 August 2024) runs when a customer’s deposits minus withdrawals cross £150 in a rolling 30 days (since 28 February 2025). It is a public-record check (bankruptcy, county court judgments, the rest of the credit-visible record); it is not an income check and it is not a credit reference check. Financial risk assessments are announced in a staged plan (Commission article of 7 July 2026), with no start date — write announced, not in force. A punter who sees a request at the £150 mark in 2026 is seeing the vulnerability check fire, not a credit file pull.

Deposit limits came in in two phases. Operators must offer a gross deposit limit on a fixed time frame; the first set of changes took effect on 31 October 2025, a second phase was postponed to 30 September 2026, and whether that second phase is live at the moment of reading is not established. The limits are the customer’s choice, not the state’s; there is no statutory ceiling on stakes or losses in sports betting. A punter who wants to set a limit can pick daily, weekly or monthly; an operator that does not offer the choice is breaking its licence.

Bonus rules tighten in 2026. Wagering requirements on deposit-related bonuses are capped at 10x (a £10 bonus may not need more than £100 of qualifying bets); the cap took effect on 19 January 2026. Mixed-product promotions (an offer that requires gambling on two or more product types, for example “bet on football to get free spins”) are banned. The cap on free bets that carry no wagering condition, and any cap on the money value of a bonus, are not established in the rules and so are not carried forward as a fixed number. Promo codes are not printed on this page; the terms each operator publishes stand as the offer’s real text.

Help services sit alongside the framework. GamCare runs the National Gambling Helpline; GambleAware runs the wider information and treatment pathway; the NHS has gambling clinics in major cities; Gamblers Anonymous runs local meetings. The Unibet page carries links to all five, which is what a well-run site does. A punter who has lost more than the plan allowed, who has chased losses, or who is betting money needed for bills should reach out before the next deposit, not after.

The mobile layer — apps, handsets and where AI predictions actually live

Most AI predictions in 2026 are read on a phone, not at a desktop. The browser still does the heavy lifting for in-play and for live trackers, but the prediction feed itself, the price-boost list and the bet-slip all run on the handset. That has consequences the punter cannot see at the betting slip itself: notifications drain data, login is biometric by default, the price tick on a wagering bet updates faster than the eye on a small screen, and “cash out” sits one tap away, which is a feature and a hazard at the same time.

The mainstream UK apps shipped by the ten operators reviewed on this page all carry the same shape. They open to a home feed, show live and upcoming events, surface a bet builder on football and tennis, lay down price boosts at the top of the page, and let the punter cash out a settled win in place of waiting for the result. What changes between apps is depth: some (bet365, Sky Bet, Paddy Power) ship live streaming inside the app, others (Midnite, Coral, William Hill) stream on a per-event basis, and one (BetGoodwin) sits closest to a racing-and-football core without the casino add-on. Apps from the GB set let a punter set deposit limits, take a time-out and self-exclude without leaving the app — a GAMSTOP registration runs through the GAMSTOP site, not the operator’s.

Touchscreen friction is a real cost. The slip on a £10 wagering bet at minimum odds is one tap on the laptop and three taps on a phone, with the value of the slip changing between the first tap and the last on a fast market. AI predictions written for that interface must be read at the price the slip locks in, not the price on the screen at the moment the bettor flicked past. The slip’s price is the contract.

Browser play is not gone. Most reviewed operators run a desktop site that mirrors the app’s market depth, and several still let a punter watch a race on the larger screen while betting from the phone. The AI prediction services that are not built into an operator’s own app tend to live on a third-party site with a login, which is a different threat model: the punter’s stake and bank details sit at the prediction provider, and that provider is not on the Commission’s register the way an operator is. The punter who signs up to a third-party AI service to receive tips on bets they will then place elsewhere is betting on one licence and paying another, and the second one is rarely a UK one.

Touch ID and Face ID handle login on iOS; Android’s equivalents handle the same job. That is good practice and saves the punter from re-typing a long password on a pub wifi network, but it makes the biometric protection the customer’s, not the operator’s: a phone left unlocked is a betting account unlocked. A punter who keeps the app on a phone with biometric login should keep the phone locked.

Notifications are where a careful punter makes the easiest saving. Push alerts from bookmakers default to on, fire on every price boost and every accumulator push, and feed the loss-chasing pattern for a punter who has just missed a result. Most apps let the punter switch marketing alerts off while keeping bet-settled alerts on, which is the line to draw. Marketing alerts that are not bet-settled are selling-time, not information.

The AI prediction angle on mobile lives in two places. The first is inside the operator’s own app, where “smart picks” or “popular accas” are priced by the bookmaker and reflect the bookmaker’s margin — they are marketing, not advice. The second is the third-party prediction app, where the model is published without a price cross-check. A punter who reads both is the punter who keeps the lines separate: the operator’s “smart” picks are the operator’s marketing, and the third-party prediction is a number to check against the operator’s price.

Predictions and strategy — bankroll, forecasting and what no model can do

The discipline that turns a stream of predictions into something close to a strategy is bankroll management. The bankroll is the total sum a punter has set aside for betting, kept apart from bills, savings and the rent; the punter then stakes a unit (often 1–3% of the bank) on each bet and never wagers more than the limit per session. A prediction at the right edge still loses, and the unit size is what keeps a losing streak from becoming a month of regret.

Forecasting on a single match has three layers. The first is information: form, expected goals, head-to-head, line-ups and injuries. The second is the model’s own weighting, where a good model uses recent results more heavily than stale ones. The third is the market — every other bookmaker’s price. Punters who rely on a single source for the third layer get the bookmaker’s margin twice: once on the price they take, once on the cut the model leaves in. Cross-checking two or three prices for the same selection before placing is the smallest improvement a small-stakes punter can make.

The technical vocabulary a UK punter meets in 2026 runs short. A “single” is one bet, one outcome. A “double” combines two outcomes, the prices multiplying together so the combined price equals the two prices multiplied into one (the margin compounds as well, and is one reason a long accumulator’s published price is worth less than the headline). An “accumulator” or “acca” runs four or more outcomes into one bet, paid only if every leg wins; “acca insurance” refunds the stake as a free bet if one leg loses, and the refund itself is paid as free bets (whose stake is usually not returned with a win). A “Trixie” is four bets (three doubles and a treble) on three selections; a “Patent” adds singles on top; “Yankee”, “Lucky 15”, “Canadian” and “Heinz” run the same engine with more legs. The total stake on a Lucky 15 at £1 unit is £15, not £1, and that is the silent line that catches first-time punters.

Forecasts on horse racing have their own grammar. A straight forecast names first and second in the right order, a reverse forecast covers both orders, and a tricast covers first, second and third. Each-way betting doubles the stake and pays a fraction of the win price if the horse places; place terms are set by the race type and the field size. Best Odds Guaranteed, applied to UK and Irish racing by Coral, William Hill, BetVictor and a number of others, pays the higher of the price taken and the starting price.

In-play betting is the segment where AI predictions are loudest. A model can scan in-play prices for drift, can flag when a bookmaker is slow to move on a goal, and can fire a tip at the price before the market catches up. The frames still apply: a price taken is a price the slip locks in, the book’s margin is still inside the price, and the 5% to 8% extra cut many books apply to in-play markets is money a sharp punter accounts for before betting. The 90 Minute Payout on Betfair, which pays out winning single bets on football if the team goes two goals ahead, is not a market margin change; it is an offer built around a specific event shape and is read on its own terms.

The hardest prediction is the one to skip. A model that disagrees with the book is a flag, not a sign the model is wrong and not a sign the model is right. The discipline is to write the prediction down, watch the closing price at the moment the event starts, and learn which inputs helped and which did not. A punter who never looks at the closing price is a punter who never audits the model.

Staking by Kelly or fractional Kelly is the bridge between prediction and bankroll. A full Kelly stakes the edge divided by the price minus one (the formula’s result is the share of the bank to wager when the bettor has an edge above the bookmaker’s margin). A fractional Kelly (a half or a third) takes a less aggressive cut of that stake and is what most small punters use in practice, because the full Kelly’s variance is punishing over a small sample. Used honestly, fractional Kelly is the bridge between an edge and a stake; used badly, it becomes a recipe for over-betting on streaks.

The “sure bet” is the marketing line the punter should refuse. A bookmaker never offers a sure bet. What the book calls sure is the cut at the top of a price-boost list, and that cut is what the punter should ignore. The “risk-free” wording on a refund of a losing bet paid as a free bet is the same move in another key: the punter has to clear the free bet before the refund is theirs, the free bet typically expires in seven days, and the stake is not paid back with a win. The £40 free bet at Unibet or Coral, against a £20 first bet, sits across from these terms in plain language, and the conditions are the price the book has paid the punter to read.

The ten reviewed bookmakers at a glance

The ten operators named below hold a current remote general betting (real events) licence from the Gambling Commission, and each name appears on the Commission’s public register under the licensee noted beside it. Domain proves nothing — operators run .com and .co.uk interchangeably — but the account number does, and that is the test a punter should apply before depositing. The figures in the table are what each operator’s terms page states; they are not predictions or rankings. A licence is a permission to operate, not a guarantee.

A punt’s question on the table is not “which is best?” — the table does not say, and would not if it could — but “which licence matches my bet?”. The name is the brand the punter sees; the number is the figure the Commission’s public register returns when the punter types it in. Two names in the table sit on common ownership threads: BetVictor and talkSPORT BET both sit inside the BetVictor group, and Coral sits alongside Ladbrokes under the same parent, but the table does not pool them and the punter should not either. Several brands can sit on one licensee; the licence is the licencee’s.

The trade-off the table leaves to the prose is the price between these offers. The 50 free spins at Midnite, alongside the £25 free bets, are a casino add-on, and the punt that takes them is paying for the casino at the betting edge the same bookmaker is happy to offer on in-play. The 5p minimum bet at Sky Bet, sitting next to a £5 deposit minimum, is the lowest stake barrier in the set; it is also a marketing line aimed at the casual punter rather than a serious bettor. The £1,000 BOG cap at BetGoodwin hits the high-roller rail of racing punters and does not touch a Saturday-afternoon casual. The £10 in free bet tokens at Betway, expiring in seven days, are the shortest hook in the welcome column; the punter who wants time to clear is looking at a different welcome. There is no clean winner in any of those pairs; there is a punter-shaped answer in each.

A licence is more than a number — what the frame costs the punter

A UK remote licence is the price of doing business in Great Britain, and that price falls on the bookmaker. The punt that meets the price is the protection the licence buys. Three of those protections show up at the betting slip and never get named there.

Identity verification is the first. Name, address and date of birth are checked before the punter can deposit or gamble, including with a free bet or a bonus. That catches the credit-card-funded second account, the underage punter using a parent’s card, and the bare stolen-card profile; it is also the reason a serious punter has to upload a driving licence or passport on first deposit. The friction is the trade-off for the protection.

Self-exclusion is the second. GAMSTOP covers every GB-licensed online operator, blocks new account opening and sign-in at existing accounts for the period chosen, and stays in place for seven years after the minimum period ends. The retail self-exclusion that betting shops use is a different scheme; the punter who has registered with GAMSTOP online is not covered in a shop. A punter who has self-excluded has not, in industry jargon, taken a break — the period runs out only on application, and a punter who has self-excluded under a five-year auto-renewal plan has signed up for five years plus seven.

Bonus caps and mixed-product rules are the third. Wagering requirements are capped at 10x, mixed-product promotions are banned, and the offers above sit inside those rules. The £40 in free bets at Unibet on a £20 first bet has a clear £40 maximum on its own (no wagering condition on the free bet itself, in line with the terms the operator publishes; the cap on free-bet money value is not established in the rules and is not carried forward). The Sky Bet £30 in free bets credited as three £10 tokens expires on a deadline written into the terms the operator publishes, and the punter who does not use them in time loses them. None of this is the punt punishing the punter; it is the rules of engagement.

Age, identity, deposits and bonuses are the bundle. The punter pays for it in time at first deposit (verification can take an hour and can take a day) and in the kind of products on offer (an unlicensed site has fewer identity checks and so can sell a bigger free bet, which is the very thing that makes it unlicensed). The reader who wants the protection is the reader who wants the licence; the reader who wants the bonus is asking the wrong question of the right system.

What it costs to play — the margin inside every price

Every price a bookmaker quotes has the margin inside it, the cut the book keeps from every unit bet on that market. The margin is the difference between what the punter pays in and what the book expects to pay back out, averaged over every outcome. The lower the margin the better the punter’s long-run return; a 5% margin is sharper for the punter than an 8% margin.

On a two-way market where both outcomes sit at decimal odds of 1.90 (fractional 9/10), the implied probabilities add to 1 ÷ 1.90 + 1 ÷ 1.90, which is 0.526 + 0.526 = 1.053. The book is returning 105.3% of what the punter pays in, the difference (5.3%) being the overround. The margin is the bookmaker’s cut, written as the difference between the implied probabilities and 100%, and sits at roughly 5.3% on the symmetric 1.90/1.90 example.

The same math on an asymmetric market (decimal 1.40 for one side, 3.10 for the other) adds to 1 ÷ 1.40 + 1 ÷ 3.10 = 0.714 + 0.323 = 1.037, an overround of about 3.7% (a margin of about 3.6%). On a three-way football market at 2.10 / 3.40 / 3.60, the sum is 0.476 + 0.294 + 0.278 = 1.048, an overround of about 4.8% (a margin of about 4.6%). The 5.3% band on the symmetric 1.90/1.90 is not a typical margin across the market; it is a worked example. Real two-way margins sit anywhere from low single digits to over 8%, and three-way football markets sit closer to 5% as a rounded average across the GB set. Each operator’s terms pages do not show their margins as such; the punter who wants to know reads the prices the operator offers on a chosen market and runs the math.

The band the calculation yields, rather than a single figure, is the point. A punter who treats one event’s margin as the market’s overall margin is reading one race’s price as if it described every race, and is over-betting on markets they have not measured. Three things move the band’s centre. The depth of the market (the more bettors, the closer to fair). The market’s anticipated shape (a heavy favourite pulls the margin up on the favourite’s side). The bookmaker’s own book (a book that already has too much on one side widens its price on that side to balance the action).

In-play markets widen the margin further. Many books publish a 5% to 8% additional margin on in-play prices, on top of the pre-match margin, to offset the speed of price moves. The punter betting on a live market with an AI prediction is paying the book for the privilege of moving at the book’s pace; the prediction is the entry ticket, the extra margin is the cover charge.

Margins are not the same as odds. A bookmaker can quote a high price and still hold a wide margin, by widening both sides of the market; it can also quote a low price and a thin margin on a niche market. The punter who follows predictions with discipline should also follow margins — by comparing the same selection across two or three books before placing the bet.

A worked example sharpens the picture. Take the symmetric example above: 1.90 / 1.90, decimal odds. The book keeps 5.3% of every unit bet on this market. A punter placing ten £10 bets in a session at this two-way market stakes £100 in total; the long-run return on those bets at the margin above is £100 × (1 − 0.053) = £94.70, the punter’s expected loss is £5.30, and that is the cost of playing on this market averaged over outcomes. The punter who is at a sharper book paying 1.95 / 1.95 (decimal) keeps more of the £100 — implied probabilities sum to 1.026, a 2.6% margin, and the session’s expected loss halves to £2.60. The price difference between one book and another is the punter’s wallet.

The lesson the calculation hands the reader: every price a bookmaker quotes carries a margin, every margin narrows the punter’s long-run return, and the AI prediction does not change that. The prediction names a probability; the price names the probability the book is willing to pay; the gap between the two is the punter’s edge, before the margin is taken out of it.

Reading a UK licence — register, numbers and the test that matters

The Commission’s public register is the test of who holds a UK licence. It is searchable by business name, trading name, account number or domain, and the entries show the licence categories the holder operates under. Remote general betting standard (real events) is the headline category on every operator in the ten-name set above. That is the licence a sportsbook needs to take bets on real sporting events; a separate licence covers virtual events, and another covers betting exchanges (Paddy Power and Betfair, for example, sit under the same parent group’s exchange and sportsbook licences). Pool betting on horse and dog racing is its own licence and duty; Tote, the pool-betting specialist in the table, holds the pool licence rather than the general one for the racing core of its offer.

An operator that holds a GB remote licence is a Commission licensee, regardless of where the operator itself sits. Curaçao, Malta, Gibraltar and the Isle of Man do not license operators for GB customers in place of the Commission; the point-of-consumption rule (the Gambling (Licensing and Advertising) Act 2014) means an operator whose site is used in Great Britain needs a Commission licence, full stop. Several names in the ten-name set are run out of Gibraltar or Malta, and their UK pages list a Commission licence as the gate. The licence status is checked against the register before deposit.

Domain proves nothing. Operators run .com and .co.uk interchangeably; the Commission register records the licence, not the address the punter types in. A punter who sees an offer on a .com site from a brand not on the register should walk past it; the brand is not licensed for GB customers and the protections on this page do not apply.

The punter’s checklist at first deposit is short. The licence number on the site’s “about” page or terms should match the register. The site’s payment screen should not accept credit cards (the LCCP has banned them since 14 April 2020). The site’s safer-gambling page should link to GamCare and GAMSTOP. The “18+” badge should sit on the homepage, not buried in two menus deep. The site’s deposit-limit screen should be reachable from the bettor’s account page without contacting support. If any of those is missing, the punter is not at a licensed book — or the book is breaking its licence.

Compliance is not static. The Commission has powers that show up in headlines — cease-and-desist notices, search-engine referrals, payment-provider work — and court-order DNS blocking powers are in the pipeline or recently granted; ISP blocking on demand is not in operation at the time of writing and is not carried forward as a fact. The punter who sees a brand pitched on social media with no licence should report the brand to the Commission and pick another.

The penalty for a customer who bets on an unlicensed site is the loss of every protection the licensed site offers. There is no Commission complaints route; no approved dispute-resolution body; no GAMSTOP coverage; no British consumer redress if funds are withheld; no 10x bonus cap and no £150 vulnerability check. The punter’s account can be closed by the operator without notice, and the funds inside the account can be removed under the operator’s own rules. The honest framing is licensed = the friction and the protection, unlicensed = none of either.

Frequently asked questions

Can AI help with betting?

Yes, in a narrow way. An AI can crunch past results, weight form, and flag where a bookmaker’s price drifts against the punter’s own probability estimate, which is the work a value punter does by hand. It cannot predict outcomes with certainty, it does not get inside the dressing room, and it cannot beat the bookmaker’s margin inside a market. A prediction that does not change the punter’s price check is decoration.

Can AI help with sports betting?

In much the same way as above. A model can scan in-play prices faster than a person can, summarise team news across leagues, and surface the bet builder’s combined price on a same-game multi. The bookmaker’s margin still sits inside every price, and a model that ignores it loses the punt’s money over a season. The punt’s discipline — recording every bet, comparing closing prices and limiting the bank — is what turns the model’s tips into a strategy.

Can AI predict sports betting?

Predict, loosely — name probabilities and flag value. Certainty, no. A prediction is a probability distribution, not a result; a model that names the correct scoreline twice in a row is not guaranteed a third. The honest AI prediction product publishes its hit rate, not its next tip.

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

A value bet is one priced higher than the punter’s own probability estimate, after the bookmaker’s margin is stripped out. If the price is 3.00 (decimal), the implied probability is one-in-three; if the punter’s model gives the outcome a one-in-two chance, the bet has positive expected value. No tip is ever certain because the model is never complete (line-ups move, weather moves, the pitch moves), and the bookmaker’s margin narrows the value window a tip must beat. A predicted win is not a paid win.

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

The margin is the cut the book keeps from every unit bet on the market, written as the difference between the implied probabilities of all outcomes and 100%. On a two-way market at 1.90/1.90 (decimal), the implied probabilities add to 1.053, so the margin sits around 5.3%. On a three-way football market, the implied probabilities of home, draw and away add up to roughly 1.05 (a margin around 5%). A lower margin is sharper for the punt; a higher margin narrows the long-run return.

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

Decimal odds: return = stake × odds, net profit = stake × (odds − 1). A £10 wagering bet at decimal odds 3.00 pays £30 (the £10 stake plus £20 profit). On a combination bet the prices of the legs multiply together: a £10 double at 3.00 and 4.00 prices carries odds of 12.00, return £120 (£10 stake plus £110 profit) if both legs win. The bookmaker’s margin is taken from the price before the stake is even placed, so the long-run return is the stake net of the margin.

Content created by the «Best betting exchange» team