Enough has been written in the pages of this and other publications about many of the threats facing prediction markets (or as one LinkedIn user would insist, Unlicensed Betting Exchanges). They’ve generally focussed on the legal challenges, the ethical standing of some of the markets offered by some providers, player protection and the manipulation and insider trading possibilities – well, some would say evidence – that prediction markets present.

As anyone who remembers dial-up internet will know, those were all challenges faced by the original betting exchanges, from 20 years ago, and they survived fine well. However, I want to focus on the pure commercial challenge that they’ll face in any given geo over the medium term. For me, these focus on… are they fun?   The vast majority of bettors do so because it’s fun, and with fixed odds you always feel like the book is balancing up based on weight of money – you could at least tell yourself you only had to know better than the guys betting on the other side.  

That might still be true for prediction markets, but it feels like newbies are going to figure out pretty quickly that they’re outgunned on all sides. For a start, casual players are going to fall into two groups, those powered by AI and those who aren’t. It’s trivially easy to prompt up an edge over the guys who don’t use the standard AI/LLM tech available, and while of course there’s still room for niche knowledge, the tech is getting smarter faster than the humans using it. That said, there are levels of AI support to consider too. The guy with his spreadsheet is going to struggle to keep up with the Chat GPT/Claude user who can analyse a world of data and opinions, but they’re still behind someone using something cutting edge like ‘Inference Engine’ 51folds, perhaps the natural predator of the base level AI user.  

Just to get an idea of what a leader in the arms race looks like, we asked Bret Dvoracek, CEO & Co-Founder of 51Folds, and Betfair ace of old, how that might take shape. “We think the edge comes from how you reason, not just what you analyse. A standard LLM can summarise opinions and surface arguments, but it’s just sophisticated pattern-matching on words. What 51Folds does is convert your thesis into a working model; one that maps out how different factors connect and influence each other. So instead of asking ‘what do people think will happen?’, you’re asking ‘if X happens, how does that shift the odds of Y?’ The user armed with that isn’t just getting a smarter summary; they’re getting a tool that can pressure-test their logic and show exactly where their thinking might be wrong. That’s a different kind of weapon to wield that nobody has yet provided to prediction market users.” If the “I have a hunch” guy doesn’t read that and burst into tears and give up, he’s not been paying attention.

The problem is the battle noted above is a huge distance from what the (immense) PR assault is selling as the concept. “It’s a truth engine, we are democratising the truth!” The real truth, however, is going to be a good distance from that concept. If and when these markets get traction (or swerve the law) enough to be active in most states and many other countries, the liquidity is going to be gigantic – easily enough to bring in the really big sharks. Just look at Robinhood and its partnership with Susquehanna, which alone should be enough to demonstrate to the average bettor that they’ve a peashooter vs a battleship. If you go into this as a retail punter thinking you’ve got to outsmart Dave the plumber from Idaho, you’re in for a shock when in fact you’ve got to outsmart people – or people running analysis engines – that make you look stupider than your own shoes. The key point with this for me is the ability to automate these processes too – if it were just a case of the big boys eating up the edge, Starlizard style, in the big markets, that would be conclusive enough. However, once they can run their systems to wring out value from any and all markets, down to the smallest liquidity level, they’ve likely sponged up most of the value – and therefore most of the fun. Beating automation is one challenge, beating industrialised analysis is quite another.

Adam Shaw on LinkedIn posted a great analysis of prediction markets and their future, where he goes into a bit more detail about long-term elements required for prediction markets to thrive, with the two key groups being market makers/professional price setters, and recreational losers willing to give their money over time. Based on the way this has been sold, the former are queuing up to skin the latter, but are the latter available in almost unlimited supply? To assume so is to assume the prediction market product is equivalent to traditional sports betting – I don’t think it is.   The user experience is different, and the high tech look and feel might be appealing initially; but in the medium term it might reinforce the suspicion that, for most users, 
they’re outgunned.

Another element yet to be really figured out is an effective way to make a two-way user-generated parlay product. It’s clearly complex – otherwise Mr Planet Brain and his buddies would already have it humming away – but given algorithms for sportsbooks have pretty much ironed out all player edges in their bet builder products (or equivalent) then it can’t be insoluble. Some kind of parlay exchange/market must be feasible, but once again that just smells like something where whoever has the most processing power can bag all the value. This might be the area where prediction markets can flourish, but if it’s an automated market maker pricing up parlay products and building in a saucy overround, that looks a lot more like what sportsbooks do than something groundbreaking that you could fit in the same ballpark – or sport – as “democratising truth.”

While we are talking about fun vs money machines, you can’t avoid the question of timing and information. For me, this isn’t as big a threat as prediction markets’ opponents would have us believe. Someone is always going to have information first, whether that’s a courtsider on a wired fast internet connection (press box, anyone?) or a researcher who gets wind of an imminent political change that is exploitable. It feels to me that this will end up being a caveat emptor situation – if you’re getting eaten by people with quicker information, more fool you, just drop out of that market. The blatant manipulation stuff is a bit more worrying, but it’s in everyone’s interests to make that kind of thing stop. I say everyone’s interests – right now we’re in the part of the cycle that, to some operators, all PR is good PR. Old Mr ‘outraged of Chelmsford’ bleating about how he got skinned on political betting by an insider has, I reckon, more positive value to operators in getting the market’s name in the press than negative value in suggesting the market is too tough to beat.

You also have to add in the critical fact that sooner or later the operators are going to have to introduce a proper margin for themselves (as opposed to front running/participating in the market as an insider with privileged information possibly via a linked but cloaked subsidiary to make an income – nobody would do that, right ?) and that just magnifies the effect – Joe Punter can’t win.

This product is clearly here to stay, but 20 years ago you’d say Betfair was the future of sports (and all) betting in the UK, and it’s now somewhere between a minority and a niche product. With liquidity key to a market of this kind, it’s hard to imagine more than 1 or 2 operators really making this pay, which for my money would be something to think seriously about when 10-digit valuations are being bandied around. There’s clearly a honeymoon period (possibly forever, I couldn’t say) where they have some state markets more or less to themselves. But once it all settles out, attracting mass losers into an AI/elite human intelligence money threshing machine doesn’t look to me like it’s going to displace traditional sports betting.