The offshore and regulated bookmaker used to have the trading desk, the data feeds and the people who understood the numbers. The customer had a newspaper, a hunch and a mate who was absolutely certain about Saturday.
Now the customer has AI.
That sounds like the opening scene of a sportsbook’s nightmare. But before declaring that offshore bookmakers are panicking, a reality check: the evidence reviewed for this article does not establish a wave of operator complaints about AI-assisted customers.
What it does show is more interesting. Machine learning can improve how betting probabilities are assessed. Meanwhile, bookmakers are using the same technology to improve their own returns. Offshore sportsbooks are using the technology to bet into exchanges while regulated sportsbooks are using it to profile players and deal multiple line sets.
The race is game-on.
For bettors, the immediate attraction is convenience. An AI assistant can help organise supplied statistics, explain unfamiliar markets and turn a messy spreadsheet into something readable. Provided the underlying information is accurate, that can make research less laborious.
It fits a wider change explored in Today News’s coverage of why sports fans rely on more data than ever before. Watching a match increasingly comes with another screen full of numbers.
However, producing a convincing explanation and producing a profitable forecast are different achievements. A fluent paragraph about defensive weaknesses does not establish that the available price is worth taking.
One particularly relevant study by University of Bath researchers Conor Walsh and Alok Joshi investigated that distinction using NBA data.
They compared selecting machine-learning models for accuracy against selecting them for calibration: how closely predicted probabilities correspond to observed outcomes. A well-calibrated model’s 60% predictions should succeed approximately 60% of the time across enough comparable observations.
The researchers reported average returns of +34.69% for calibration-based selection versus −35.17% for accuracy-based selection in their experiments.
Those figures need their context attached. The betting simulations covered the 2018–19 NBA season using published Westgate closing odds. They were historical experiments, not a verified record of customers placing those bets today.
The research also tested trained predictive models, not somebody asking a general-purpose chatbot for tonight’s winners. Its useful lesson concerns evaluating probabilities; its headline returns are not a forecast for readers.
Here is why that distinction matters.
Imagine a hypothetical system that correctly picks winners 80% of the time. Impressive? Certainly sounds it.
But suppose every selection pays decimal odds of 1.20. Across 100 equal £10 bets, 80 winners produce £160 in profit, while 20 losses cost £200. The result is a £40 loss on £1,000 staked.
An 80% strike rate has produced a negative 4% return.
At those odds, the break-even rate is 83.33%. The system can be right most of the time and still leave its owner poorer. No amount of confident wording changes the arithmetic.
This is where price comparison earns its place. Today News’s guide to comparing betting odds explains why comparisons must concern the same selection and market.
For US sports, OddsTrader displays sportsbook prices alongside matchup information. That provides a practical reference when checking whether an analysis relates to the price actually available. Displayed odds still need confirming with the operator before placing a wager.
The uncomfortable part for anyone selling an “AI beats the bookies” story is that bookmakers have AI too.
Sportradar’s Alpha Odds performance figures report an 11.84% profit uplift against its own base odds across supported sports over the rolling 12 months to August 2026. The stated comparison covers 81 bookmakers and more than 2.2 billion accepted tickets.
These are vendor-reported figures, not an independent audit or evidence specifically about offshore operators. Nevertheless, they show what the supplier is selling: AI that strengthens the bookmaker’s business.
In an April 2025 announcement, Darren Small, then identified as Sportradar’s senior vice president of Managed Trading Services, described the technology as “giving operators sharper margins and bettors a more dynamic experience.”
The sharper margins deserve as much attention as the dynamic experience. His comments describe a commercial tool for operators, rather than an endorsement of customer profitability.
Offshore betting introduces another question that a prediction cannot answer: what happens after the bet?
A model’s estimated advantage says nothing about withdrawal conditions, identity checks, settlement disputes or the protection available in a customer’s jurisdiction.
For readers examining the US market, OddsTrader’s offshore sportsbook comparison provides a starting point for reviewing operators and their advertised terms. Its commercial rankings should be checked against the operator’s current conditions and relevant regulatory information.
An attractive number on a screen is only one part of that assessment.
Before trusting an AI betting service, ask for a dated record that includes every selection, losing periods and prices available when recommendations were issued. Ask whether results came from historical simulations or actual bets, and whether subscription costs are included.
Also ask what the system does when information is missing. “I don’t know” can be a more useful answer than a beautifully written guess.
AI can make research faster and probability analysis more accessible. It can also make weak reasoning sound remarkably polished. The sensible test remains whether its claims survive scrutiny.
A chatbot can explain a bet in seconds. It cannot make uncertainty disappear, and bookmakers are not leaving their own computers switched off.
David Prior
David Prior is the editor of Today News, responsible for the overall editorial strategy. He is an NCTJ-qualified journalist with over 20 years’ experience, and is also editor of the award-winning hyperlocal news title Altrincham Today. His LinkedIn profile is here.
