I started betting professionally in 2005. I started because it was a more interesting side hustle than poker, but it quickly became all-consuming. Betting on sport was a bit of math, a lot of sports knowledge, and good reasons to spend endless hours watching teams and optimising your takes. I got to sit on the couch and watch every match of the 2006 World Cup for “work reasons” and my wife couldn’t complain – life was grand.
The side hustle quickly turned into something real, which then turned into nearly a decade designing gambling products at one of the world’s largest offshore bookies. Then I spent a couple of seasons with Matthew Benham at Brentford, before eventually turning my sights on building the best gambling and football recruitment data in the world by founding StatsBomb.
Twenty years later, a lot has changed, but a surprising amount is still the same. Betting on sports – and football especially – is still a glorious problem-solving exercise that combines maths, sports knowledge and treasure-hunting into something highly profitable. And it’s still incredibly fun, especially when you are winning.
Gambling model outputs barely existed when I started but are now ubiquitous in the form of expected goals. Plenty of people assume that means the computers have taken over and all the advantage for human players has disappeared, but I can assure you that is not true. The fact of the matter is, the advantage has shifted from those who use models to those who understand what the models are capable of, and who make plays in the holes the robots leave behind.
Yes, lines are much sharper than they used to be. You’re no longer going to find lines that are wrong by 20 or 30 percentage points on which team is likely to win, across half the league, every single week. So in that sense, being a winning gambler is harder than before. But I have been heartened by the fact that there are still many wrong lines to play, even as expected goals as a concept has been adopted by pretty much every professional gambler, bookie and team I know.
To prove the concept, I spent last season documenting and analysing every play I made across three leagues in my newsletter, Variance Betting. People understandably hate touts, but this was me – a sometime professional gambler – trying to figure out if he could still win after a decade away from betting, for science. There may have been a bit of luck involved in that – there always is – but the successful patterns of old mostly repeated themselves, except this time with greater emphasis on initial model development, and much greater emphasis on what the models might not know.
People understandably hate touts, but this was me – a sometime professional gambler – trying to figure out if he could still win .
I learned three things:
That last bit is why I continued writing and educating novice punters this season. It just didn’t sit right with me that a lot of the products I helped pioneer had been turned into these monstrous profit-extraction machines. So by educating people from the lens of a professional, maybe I could help even the playing field just a bit, and make new friends along the way.
The lesson: even in the age of better data, machine learning and AI, smart punters can still win by combining sports knowledge with models to stay ahead of the game, and as someone who has been working with all of these elements for decades, I doubt that’s going to change any time soon.
Ted Knutson is a pro-punter and founder of StatsBomb
