AI models are terrible at betting on soccer—especially xAI Grok

Systems from Google, OpenAI, Anthropic, and xAI struggle with the Premier League.

See full article...

 

[NinjaNerd56]

Yes, because the current "AI" models are nothing more than glorified predictive text and search systems.

They guess at what the most likely next text token is based on the knowledge they've trained on - they aren't "intelligent" in any meaning of the word.

Short version…shitty code, written by children.

 

[Erbium168]

As a Brit I call it football, but as a point of interest it’s called football because it’s played on foot, as opposed to on horseback.

So polo is called horseball? Got it.
Mildly amusing that you seem to have cited Google "AI"s summary of the Wikipedia article, which itself is much less definitive in its attribution.


The problem with that argument is that mediaeval Europe didn't go in for ball games on horseback. That reached is from Persia via India, which is why it is called Polo - a Tibetan language word just meaning "ball". So there was no contrast in Europe between ball games on horses or on foot.
The word "Polo" entered English after the game of Association Football was defined.

I hope you can come up with a more intelligent rebuttal than your one about the UK legal system.

 

[Erbium168]

And Sorceror's Stone vs Philosopher's Stone. Which I thought was weird because I remember reading Donald Duck comics featuring the Philosopher's Stone, which turned anything it touched into gold.

Indeed. "Philosopher's stone" dates from when prescience was called "Natural Philosophy" - a term which persisted for a long time into the actual scientific period. A sorcerer is a practitioner of magic.
For Harry Potter, Sorcerer is far more appropriate since in Rowling's world the wizards struggle to understand science.

 

[cinc]

Lenin says capitalism will never die... it reconstitutes into imperialism... then cultural commodity capitalism powered by fetishism, which then returns to cultural imperialism. You have to smash capitalism, which can only be done via class struggle.

The hope is if we have robot overlords, we will stop being racist xenophobes and unite against the capitalists. That's where I part with my namesake. I think ecological catastrophe is the only mechanism to break the existing system... but that's a different conversation.

Why does everyone say shibboleth again? Why is that word suddenly back into discourse? Very strange. Anyway.

Lenin was an autocrat mass murderer.

 

[markgo]

Much as I’d love to read this as a knock against AI capability, I seriously question the experimental design.

Humans use countless sources of information both consciously and unconsciously. In addition, betting information comes fast and furious in the period immediately before the game, including such critical things as injuries.

Without this kind of information, any bettor is doomed, not just AI.

 

[DrHow]

There needed to be controls that were making random bets. If the odds were fair, their expectations should be zero. However, the odds are not fair. There is vigorish. The random better's expectation is also negative. They went through their stash often enough that a substantial overall negative expectation is to be expected. It could only be compared meaningfully to the performance of random betters.

 

[House of Propane]

Short version…shitty code, written by children.

This is insane cope, the compensation at AI labs is incredible so they have pretty much absorbed the top talent.

 

[norton_I]

The results offer some comfort to white-collar professionals and businesses who are fretting that AI could take their jobs, as it roils the shares of industries from finance to marketing.

As long as those professions are successfully predicting something that is fundamentally unpredictable. I doubt many other people will be comforted.

As always, the usual way to make money in gamblingprediction markets is to have inside information, or set the rules in your favor. I don't doubt if you gave the AI tool access to threaten/bribe/spy on or otherwise cheat at sports betting that they would do as well as humans who have access to those tools.

 

[Kenjitsuka]

Oops, predictive text 2.0 fails yet again in anything other than writing slop and burning resources!

 

[norton_I]

Nate Silver has written on how ChatGPT is lousy at poker, an interesting read
“But if there are examples where LLMs already seem to have superhuman capabilities, they’re very far from it in poker. And I’d argue that poker is a better test of general intelligence than some of the more discrete tasks that ChatGPT performs so well.”

https://www.natesilver.net/p/chatgpt-is-shockingly-bad-at-poker

That's LLMs. I know everything thinks "AI == LLM" but there ML models that have been world class at poker for years. In fact it's generalizable: LLMs are bad at pretty much all strategy games, but neural networks trained for strategy are often great at them.

LLMs are really good at language. Humans do a lot of language, so that's impressive. But it's not everything. It's an easy mistake to make because LLMs are so popular that the flagship models are truly enormous. If you put a trillion parameter LLM with a significant fraction of the internet encoded into it, that can cover for a lot of missing capabilities.

 

[schnopp]

Forgive me but betting companies have to make money. This report is nothing to do with AI but rather the success of the businesses.

Guy

 

[RZetopan]

Yes, because the current "AI" models are nothing more than glorified predictive text and search systems.

They guess at what the most likely next text token is based on the knowledge they've trained on - they aren't "intelligent" in any meaning of the word.

That same description apparently applies to the shills promoting LLM AI so heavily.

 

[norton_I]

Even if you accept the principle that by carefully studying public information it's possible to gain a meaningful and durable edge in sports prediction, I don't think this experiment is very meaningful.

The models were not connected to the internet. They were fed data the researchers decided was relevant and asked to make decisions based on that. Maybe that's slightly interesting to compare the models with each other but for actually performing as a betting agent that's potentially a huge disadvantage. Realistically you want it to be able to access any news, forum discussions, and even weather forecasts that would have been available at the time when the bets would nominally be made. Since it's basically impossible to roll back the internet like that, I think the only sensible thing to do would be to run it forward looking for a season currently underway and track performance over a season.

 

[RZetopan]

Of course this is the top voted comment. Smh

And we all heard the sound of a single loose marble.

 

[RZetopan]

Keep screaming into the void. The comments on this site are more 'predictable' than Claude 4.6.

Once LLM weights are trained on sports gambling, they will start winning.

Go is a notoriously difficult game and machine learning was able to surpass humans despite years of 'experts' claiming it was impossible.

And you have proof of this, and would gladly show it, but your dog ate it. Funny how you totally miss the fact that they were trained, but flipping a coin is better.

 

[habilain]

Indeed. "Philosopher's stone" dates from when prescience was called "Natural Philosophy" - a term which persisted for a long time into the actual scientific period. A sorcerer is a practitioner of magic.
For Harry Potter, Sorcerer is far more appropriate since in Rowling's world the wizards struggle to understand science.

OK, as someone who knows mythology, the Philosopher's Stone is from Alchemy and keeps coming up in a lot of time periods throughout history, because it's basically the ultimate panacea. It was said to be a stone capable of transmuting lead into gold, mercury into silver, and also is the key ingredient of the Elixir of Life which granted immortality. While I have not personally read the Harry Potter books, a quick bit of research states that in Harry Potter the Philosopher's Stone grants the user immortality.

So that's a hard disagree. The Philosopher's Stone is referring to a specific thing in mythology, and the Harry Potter usage tracks with the myth. The Sorcerer's Stone is a made up thing that came into existence because an editor at Scholastic USA felt that Philsopher's Stone was too archaic.

 

[cleek]

OK, as someone who knows mythology, the Philosopher's Stone is from Alchemy and keeps coming up in a lot of time periods throughout history, because it's basically the ultimate panacea. It was said to be a stone capable of transmuting lead into gold, mercury into silver, and also is the key ingredient of the Elixir of Life which granted immortality. While I have not personally read the Harry Potter books, a quick bit of research states that in Harry Potter the Philosopher's Stone grants the user immortality.

So that's a hard disagree. The Philosopher's Stone is referring to a specific thing in mythology, and the Harry Potter usage tracks with the myth. The Sorcerer's Stone is a made up thing that came into existence because an editor at Scholastic USA felt that Philsopher's Stone was too archaic.

after which Rowling herself came up with the US title.

 

[RZetopan]

We read about this nonsense in grade school. It's John Henry all over again.

What's fascinating is that this site helped build the internet. Arstechnica was the place to get insight on how to build servers in the pre-Google search era. I wouldn't have been able to start our branch of Indymedia without it. Now it is by far the most reactionary place on the Internet.

Commenters here are still screaming about hallucinations... in 2000, they would have said PEBCAK and laughed at anyone who didn't check their work.

I wish the old Arstechnica was still around. Everyone has become bitter and old.

Feel free to invest everything you have in these scams, so that you can tell us "I told you so" when you finally win big time.* Is that you Sam? Shills and cult members are rarely ever aware that they are shills and cult members (and in the rare occurrence when that happens, they soon exit).**

*A date that will never come.

**
View: https://www.youtube.com/watch?v=t1GzfA95ZE4

 

[Arstotzka]

Maybe the reason that these "AI" suck at handicapping European football is that these things are American, as am I, and we Americans historically suck at "soccer". (Except for the occasional game, er .. match.) Because, I suspect, all our prime athletes play one of our other big 3 sports, which are also more rewarding in the US.

ps. I have always wondered whether the idea of the Philosopher's Stone, aka the alchemist search for the making of gold, came from the Buddhist concept of emptiness and enlightenment. Maybe.

The United States has won the Women's World Cup four times. It's only been held nine times since 1991. American women seem to be plenty good at soccer. Sadly, their accomplishments are under-appreciated and disappointingly unknown.

 

[RZetopan]

I go with “Alternative Intelligence”, it’s just as intelligent as Alternative Medicine is medicinal. It’s also full of alternative facts.

Especially all of those "facts" that LLM AI makes up on the spot.
https://www.morphllm.com/ai-hallucination-examples
https://mljourney.com/examples-of-llm-hallucinations/
https://thisweekinsciencenews.com/b...inations-types-examples-and-how-to-spot-them/
etc. Way too many to list all the links.

 

[RZetopan]

Yeah have to see as we add more data and the models get smarter, they might be better temporarily at certain games but betting/stock markets quickly normalise once someone starts beating the system.

As far as LLMs playing chess, latest Gemini now has an elo of around 2k which is already better than 99% of human chess players.

This will probably improve as models get smarter.

Probably? Pigs will probably fly before that happens. But the LLM AI shills will never improve, since that would result in them not shilling.

 

[jonbob_newcastle]

The United States has won the Women's World Cup four times. It's only been held nine times since 1991. American women seem to be plenty good at soccer. Sadly, their accomplishments are under-appreciated and disappointingly unknown.

That's a shame. Until a few years ago the English women's team barely got a mention either, but after they reached the final back in 2023 there was a groundswell of public support verging on honest to goodness patriotism and a lot more young women and girls took up an interest in the sport.

A cynic might say that the Lionesses only got that attention because the men's team weren't doing well. They might be right, but anything that encourages young people to go outside and be physically active is bound to be a good thing overall.

 

[ChrisMarshallNY]

Let me guess... Grok always bets on the team with the most blond-haired, blue-eyed players?

 

[jonbob_newcastle]

So polo is called horseball? Got it.
Mildly amusing that you seem to have cited Google "AI"s summary of the Wikipedia article, which itself is much less definitive in its attribution.

I didn't cite anything, nor did I mention polo.

Now, where's that ignore button? I have enough salt in my diet already.

 

[HuntingManatees]

As a Brit I call it football, but as a point of interest it’s called football because it’s played on foot, as opposed to on horseback.

Seems like I've been playing handball all wrong this whole time!

 

[issor]

I mean, I get how those comments are repetitive and it can be frustrating when every conversation goes the same way, but they're also true. LLMs can be useful in some ways as long as people understand their limitations, but there is no through line from LLMs to AGI.

That’s quite a goalpost shift though, saying if it isn’t AGI then calling it “glorified autocomplete” is accurate.

I don’t think it was reasonable to pit LLMs (or any machine learning for that matter) against something as unpredictable as sports betting when historical data only helps marginally. I get why people say “duh, LLMs can’t do this”. However, the completely dismissive and hyperbolic comments here do get grating. LLMs are way more useful than a “glorified autocomplete”.

 

[AI_Skeptic]

...And a hammer is terrible at cutting wood.

Why is this even a story?

 

[AI_Skeptic]

LLMs are way more useful than a “glorified autocomplete”.

Okay, like what? It's a glorified autocomplete. The only thing I can see it do is autocomplete things that are in the training data.

Please provide sources, and please identify how what these LLMs do generate value for individuals or businesses?

Please do NOT include anything that's related to programming in the sources.

 

[AI_Skeptic]

That same description apparently applies to the shills promoting LLM AI so heavily.

A lot of these LLM shills strike me as being the same people who were shilling "the blockchain" and "bitcoins" a few years ago.

 

[TheBaconson]

I'm not surprised. I don't think AI could do my job.

But I think AI can convince my management that it could do my job. That's what I'm worried about.

Where I’m currently doing some work at the moment there is a guy on the QA team that keeps coming up with ways AI can assist us, the answer every time is always, we already have a system in place, it is cheaper and it doesn’t require a subscription, or retraining.

 

[TheBaconson]

So you're argument is that LLMs won't improve at chess or poker?

Pigs will fly if they do? If you said the same thing 2 years ago when they couldn't play chess at all but now they're at 2000 elo (better than 99% of chess players), you wouldn't feel even the slightest hint of self doubt?

Of course not - far easier to be a luddite ostrich who shouts 'na na na' and hides from reality and facts.

The funny part is that everyone is waking up to AI competence and utility - that's why they're taking over everywhere. You guys here can keep on yelling about 'how the world is being swindled' but that's just not mapping with the real experience of people.

I almost feel bad for you... Almost.

I am yet to see anywhere in my daily life where AI has taken over anything, I’ll add willingly AI in notepad was not exactly a me decision.
The only place was my bank who used AI for customer support, that lasted all of a few weeks.

 

[kohukohunui]

leaving the rest of us to finish the job of making it really work, maintaining it, figuring out how it plugs into other things, updating it for new standards, getting rid of obsolete bits of it

Except Claude Code is actually making those things significantly faster and easier. It's reactionary to assume it's only used to generate swathes of unverified vibe code for greenfield dev, and nothing else.

 

[torp]

Except Claude Code is actually making those things significantly faster and easier. It's reactionary to assume it's only used to generate swathes of unverified vibe code for greenfield dev, and nothing else.

Interesting choice of term: "reactionary". Why do all "AI" evangelists use terms and arguments that come from religious and dictatorship brainwashing directions?

 

[ridergk]

On paper we were suppose to beat Southampton but it didn’t happen. As any fan knows, it’s bloody football and the unpredictability of it and that any side has the potential to be able to beat any other side is what makes it fun to watch and nail bitingly tense especially if you’re a Gunner.

 

[Pasqui23]

I mean it's not like there is much written about sports betting online.I doubt anyone who isn't deeply invested (and most who are) into it would fare better

 

[crepuscularbrolly]

Except Claude Code is actually making those things significantly faster and easier. It's reactionary to assume it's only used to generate swathes of unverified vibe code for greenfield dev, and nothing else.

Yeah, it's making you unwilling to learn and to think.

 

[crepuscularbrolly]

Only 94% of the world calls it football, it's hardly a supermajority.

While I give the US the fact that cricket is deeply confusing to the uninitiated (and don't let's start on croquet or Real Tennis) we Right Pondians have difficulty understanding why the US refers to an entertainment as "football" when, in fact, the ball hardly ever gets kicked at all. Though the US can justifiably point out that basebal explicitly refers to the key components of that game, while the British name "rounders" doesn't.

Two nations separated by languages in which 95% of the words look the same but 70% sound different and 40% have different meanings.

Let me introduce you to Rugby football. Or maybe Aussie rules football.

As for rounders, you go around the bases. Same sort of etymology as "soccer", and "rugger": using "-er" as a diminutive.

 

[The Lurker Beneath]

Is 'mystical' the US spelling for 'mythical'?

Phoenix probably qualifies as both.

 

[The Lurker Beneath]

That's a shame. Until a few years ago the English women's team barely got a mention either, but after they reached the final back in 2023 there was a groundswell of public support verging on honest to goodness patriotism and a lot more young women and girls took up an interest in the sport.

A cynic might say that the Lionesses only got that attention because the men's team weren't doing well. They might be right, but anything that encourages young people to go outside and be physically active is bound to be a good thing overall.

It only started to become a thing in Ireland around that time too - though of course Irish soccer rides a lot on the back of the British game in that the best players tend to join teams in the UK.

Irish women have been relatively well-served in Gaelic sports, though. Camogie (women's version of hurling) has been popular since its invention in 1904.