Waterhouse VC – AI provides catalyst for sharper and bigger betting syndicates

This is a republication of the April 2026 update from Waterhouse VC. It has been republished with permission. Subscribe to Waterhouse VC here to receive regular updates. 

Scottish football came closer than it had in decades to seeing what it had not seen since Aberdeen in 1985: a league champion outside Celtic or Rangers. 

Heart of Midlothian FC fell short on the final day, but the fact they got so close was extraordinary in itself. Their latest annual turnover was £24.4 million, compared with £94.1 million at Rangers and £143.6 million at Celtic.

Tony Bloom is central to the story. His £9.86 million investment for a 29 per cent non-voting stake was completed on June 25 2025, but the more important move came seven months earlier, when Hearts became the exclusive Scottish partner of Jamestown Analytics, the football data business in Bloom’s wider network.

Bloom, founder of Starlizard and widely regarded as the world’s best football bettor, is one of the few people to have carried a professional betting edge into mainstream sporting ownership. 

Brighton have gone from League One to an established Premier League side competing around the European places, while Union Saint-Gilloise won their first Belgian title in 90 years. 

His 19.1 per cent stake in Melbourne Victory suggests the same model is now being tested in Australia. 

Bloom’s football projects are the visible version of a much larger betting ecosystem where the same principles apply: better data, better models, going deeper, sharper prices and disciplined execution. 

The most successful betting groups turn over enormous sums but do not publish strategies, disclose models or seek attention. 

Waterhouse VC is seeing more individuals and teams attempting to build serious betting operations, with AI acting as the catalyst.

AI + square = sharp?

AI’s promise to the recreational bettor, or “square” in industry parlance, is to close the gap to the professional, or “sharp”. A growing number of AI-assisted betting tools help users compare prices, find arbitrage, identify positive-expected-value bets, track performance and understand market movement. 

They can be valuable to bettors who use them. With mass distribution, they become valuable businesses.

AI has also made it cheaper and faster to scrape data, write code, clean datasets, build models and produce presentable results. 

The more interesting question, for Waterhouse VC and for the incumbents we work with, is whether AI helps new teams get close enough to compete, or simply makes the strongest groups more efficient.

A benchmark is whether a team can win at scale: generating significant profit over time, in liquid markets, after accounting for limits, execution, costs and market movement. 

At the elite level, professional betting is a high-volume, low-margin business where small edges only matter if they can be repeated, protected and scaled.

AI may help more teams look sophisticated. It may even help some improve. Reaching that level remains a much harder test.

Age of information

Every new market tool arrives with a democratising promise. Poker training content made strategy easier to learn. Retail trading apps made markets easier to access. 

Betting exchanges and prediction markets made prices more transparent and participation more open. 

Football has been on the same curve: data-led recruitment is now standard practice, yet clubs with deeper analytical infrastructure, like those in Bloom’s network, have continued to pull ahead.

AI is, for now, an extension of that story. It makes research faster, modelling easier and analysis more presentable. It allows more people to build something that looks credible.

The biggest beneficiaries are not necessarily the new entrants. They are the groups that already understand the models, data and market structure well enough to use AI properly. 

In serious operations, AI can speed up cleaning, testing, monitoring and reporting. The people behind the process, not the software, still know what good output looks like.

The advantage comes from knowing how to interrogate the output, spot flawed assumptions and separate a genuine edge from a false positive.

For less experienced teams, the risk is the opposite. AI can make a weak model look finished, turning incomplete data into a polished dashboard and giving the user confidence they have not earned. 

It lowers the cost of producing analysis. It does not lower the cost of knowing whether the analysis is any good.

‘Detailed but still wrong’

We put the question to one of the betting groups in our investor base. Their answer was unambiguous.

“AI has made it much easier for people who think they have an edge. The reality is that those who couldn’t build a model before still can’t build one with AI. The AI produces an answer that is detailed but still wrong.”

That AI cannot turn weak modellers into strong ones is true today. The next months and years of AI progress will test that claim. 

The more durable point sits beneath the model: serious syndicates run on proprietary data sets built over years, clean, structured and indexed to outcomes. 

Newcomers do not have them, and AI cannot manufacture them.

Portfolio companies are already using AI to embed deeper advantage in the systems they own, widening leads that were already difficult to close.

Another syndicate in our network framed it more structurally: “Small data sets make it very hard for AI to beat traditional managed stats. That won’t change.”

In betting, a model only matters if it survives contact with the market. Their view was that anyone using AI to build serious betting models still needs an expert checking the work line by line. 

Not someone who understands prompts. Someone who has built neural models, the predictive machine-learning systems behind serious betting infrastructure, and knows where they fail.

One syndicate did concede that for someone young, intelligent and unable to afford a development team, AI is potentially a significant advantage. The constraint moves from capital to capability. 

For the average bettor without underlying expertise, AI is only as good as the person using it.

That means checking whether the data is clean, whether the backtest is honest, whether correlated outcomes are being mispriced and whether the assumed liquidity actually exists.

What actually changes?

Once a working model exists, AI becomes a force multiplier. It removes much of the grunt work around updating, reporting and monitoring. 

Professional betting operations rely on constant maintenance: refreshing datasets, tracking performance, monitoring prices, flagging anomalies and producing reports. AI makes that faster and cheaper.

“We are not replacing any analyst who leaves,” they told us. “We are putting in AI agents instead.”

The edge still sits in the original model, data and process. AI helps maintain the machine. The group is also careful about where its models live: downloaded to local machines, never publicly shared. 

Serious betting models are proprietary infrastructure, not data to be uploaded, queried and recycled. Once the model, data or process is exposed, the edge starts to disappear.

For weaker bettors, AI can make analysis more plausible, models more presentable and mistakes harder to spot. For strong groups, it compounds existing advantages.

David’s sling

It is a nice fantasy to think AI can turn ambitious bettors into Tony Bloom. The more realistic opportunity is wider and more practical: AI-embedded products that serve both the retail and professional ends of the market.

At the retail end, the best products will help bettors act more intelligently: comparing prices, tracking performance, identifying mispriced opportunities and understanding market movement. These businesses do not need to create professional bettors to be valuable. 

With a simple product and strong distribution, they can reach large audiences.

“AI has made it much easier for people who think they have an edge. The reality is that those who couldn’t build a model before still can’t build one with AI. The AI produces an answer that is detailed but still wrong.”

For operators and syndicates, the supplier tools within trading serve similar purposes. 

The ability to track prices across sharp books, execute quickly, manage risk and protect margin becomes more valuable as AI raises the floor across the market. More participants enter.

Existing retail bettors stretch each gambling dollar further with better tools, losing at lower margins and turning over more over a longer customer lifetime.

Both outcomes compound demand for the infrastructure that makes wagering function: data feeds, trading tools, market-making, surveillance, compliance and risk management.

This is a republication of the April 2026 update from Waterhouse VC. It has been republished with permission. Subscribe to Waterhouse VC here to receive regular updates.