The AI Lab · Silicon Tipster League
Grok: horse racing tips, tested
Grok · xAICurrently #5 of 8
Grok is xAI's contrarian, plugged into the live chatter of X. In the Silicon Tipster League it reads every UK and Irish racecard cold and names one horse to win, and we log the lot in public.
Research, not tips. Every pick logged before the off and settled at industry Starting Price, wins and losses alike. 18+ · please gamble responsibly.
Grok's scorecard
2,561 picks settledStaked (£1 stakes)
£2,561
Returned
£2,097
Profit
−£464
ROI
-18.1%
AI Picks the Winner
−£200(-13.8%)Never sees the odds, reads the race itself.
- Staked
- £1,445
- Returned
- £1,245
- Bets
- 1,445
- Win%
- 23%
AI Finds the Value Bets
−£265(-23.7%)Sees the market's implied chance, hunts the value it missed.
- Staked
- £1,116
- Returned
- £851
- Bets
- 1,116
- Win%
- 14%
Went its own way: 17 races where Grok was the lone dissenter while the rest of the field agreed on a different horse.
Grok's running profit: AI Picks the Winner vs AI Finds the Value Bets
One line per arm, £1 a bet, settled to Starting Price. The gap between them is the interesting bit: does seeing the market's view help Grok, or just pull it onto the favourite?
Grok's recent calls
The actual picks and one-line reasons Grok logged before each off, newest first. Read the logic and judge it for yourself. The board settles the argument at Starting Price.
| Race | Arm | Pick | Conf. | Reason | Result |
|---|---|---|---|---|---|
| Kempton Park21:00 | Value pick | Slack Bob | Medium | Trainer form overlooked by market | Lost |
| Kempton Park21:00 | Winner pick | Santiago Boy | Medium | Strong trainer-jockey pairing on polytrack | Lost |
| Kempton Park20:30 | Value pick | White Crown Star | Medium | Low market price ignores trainer's recent handicap wins | Lost |
| Kempton Park20:30 | Winner pick | Talis Evolvere | Medium | Best trainer-jockey combo on the card | Lost |
| Sligo20:18 | Value pick | Run For Fun | Medium | Mullins bumper specialist heavily undervalued | Lost |
| Sligo20:18 | Winner pick | Run For Fun | High | Mullins team dominant in bumpers | Lost |
| Worcester20:08 | Value pick | Cinnodin | Medium | Top yard/jockey combo underpriced | Lost |
| Worcester20:08 | Winner pick | Stiletto | Medium | Harry Cobden aboard top jockey | Lost |
| Kempton Park20:00 | Value pick | Starlight Time | Low | Varian handicap debutant likely underestimated | Won · SP 7.5 · +£6.50 |
| Kempton Park20:00 | Winner pick | Exposure | Medium | Top trainer with capable jockey aboard | Lost |
| Sligo19:48 | Value pick | Cornmarket | Medium | Market underestimates handicap potential | Lost |
| Sligo19:48 | Winner pick | Ringdufferin | Medium | Top trainer A L T Moore | Lost |
About Grok
Grok is the model built by xAI, Elon Musk's AI company, and it wears its personality on its sleeve. Where most assistants aim for the measured middle, Grok's public character is brash, contrarian and quick to back its own read against the room. It is also the one big model wired into X's real-time feed, so it has a reputation for reaching for the live take rather than the settled one. Whether that swagger is an edge or just noise on a racecard is exactly the kind of thing our league is built to find out.
In our league Grok runs on the current frontier model from xAI. We are not grading it on its X plumbing or its hot takes: here it gets the same plain racecard every other AI gets, the runners, going, class and distance. That levels the field and lets Grok's actual reasoning do the talking, one horse and one short line at a time. The character we are curious about is the betting one: does a model with a contrarian streak keep finding a different answer to the crowd, and does that independence hold up when the results are settled in cold, honest numbers?
How Grok actually picks
Handed a racecard with no odds, Grok has to build a case from the form, the going and the shape of the race. Its instinct, on reputation, is to look for the angle nobody else is on rather than to nod along with the obvious form pick. That is the trait we will be watching most closely. When Grok's one-line reason lands on an unfancied runner, is it seeing something in the going or the class drop that the field has underrated, or just reaching for a contrarian shape for its own sake? Its picks and its actual reason strings are published on the board, so you can read the argument it made and judge for yourself whether the boldness is insight or bravado. Naming a live one is a real skill; being paid enough when it wins to clear the bookmaker's margin is a different and much harder thing, and the board keeps those two honest.
Here is what the numbers say, updated live as the sample grows. Reading a race blind, with the prices hidden, Grok backs the outright favourite 28%of the time, and its picks most resemble the “back short-priced favourites” approach (backs the favourite only when it is odds-on). Its stated reasons lean on trainer, jockey and form.
Show it the market and it changes character: its value picks swing toward the “our own value strategy” approach, hunting the overlooked rather than the well-fancied. Any positive return here is over a young sample and is not yet a proven edge, research, not tips.
See how all five AIs rank → · Read the full write-up on how Grokpicks →
AI Picks the Winner vs AI Finds the Value Bets
Grok is the model where the winner-versus-value split could be most revealing. In AI Picks the Winner, never shown the price, its contrarian streak has room to run: it reads the race on its own terms and picks the horse it fancies, crowd be damned. The interesting question is what happens in AI Finds the Value Bets, when Grok can see the market. A truly independent thinker should use the price as one more input and still back its own judgement where the numbers disagree; a model that merely poses as contrarian might quietly fold and drift toward the favourite once the odds are in front of it. Watching whether the two Groks pick differently (and whether seeing the market sharpens its reasoning or just tames it into the consensus) is one of the better tests of how deep the model's independence actually goes.
What to watch on Grok's board
- Whether Grok is a true lone dissenter: the races where it 'went its own way' and picked a horse all four other AIs passed over
- How often its winner picks and value picks disagree, and whether seeing the market talks it out of its bolder calls
- Its short reason strings: is the contrarian pick backed by something in the going, class or distance, or just a bold shape?
- Whether that swagger holds up over a real sample once picks are settled at SP in plain numbers
The rest of the field
Grok is one of 8 on the board. See how the others are reading the same races:
Questions about Grok tips
Is Grok good at horse racing tips?
We are running the experiment in public to find out, with no thumb on the scale. Grok picks one horse per UK and Irish race from the racecard alone, and every call is logged before the off and settled at Starting Price, so its board shows exactly how it is doing. Naming likely winners is one thing; doing it well enough to beat the bookmaker's margin over a real sample is far harder, and our wider work suggests no method reliably manages it. Treat it as research and entertainment, not betting advice.
Does Grok's link to X give it an edge on the races?
Not here. In the Silicon Tipster League every AI, Grok included, is handed the same plain racecard (runners, going, class and distance) with no live feed from X or anywhere else. So the page reflects Grok's reasoning about the race in front of it, not its real-time plumbing. That is deliberate: it keeps the five models on a level field so any difference is down to how each one thinks.
Should I bet the horse Grok picks?
No, please don't treat any of this as a betting signal. The League is a public experiment in how five AIs read a racecard, published as research and entertainment. A named pick winning is not the same as being paid enough to come out ahead over time, and the bookmaker's margin is a formidable opponent. If you do bet, it is 18+, for fun, with money you can afford to lose; never chase, and see BeGambleAware.org.
Why does Grok sometimes pick a horse none of the other AIs fancy?
That is the 'went its own way' case, and Grok (being the contrarian of the group by reputation) is a natural candidate for it: races where it is the lone dissenter while the other four agree on a different horse. Sometimes independent thinking spots something real; sometimes it is just a bold call that doesn't come off. The board records both without flattering either, which is the whole point of running it in the open.
This page sits inside the AI Lab, where we test whether any betting system makes money (across 26,000+ races, none of them do), and ask the bigger question in does following an AI tipster work?
Gamble responsibly.This page is research and entertainment, not betting advice. No AI here beats the bookmaker's margin, and nothing on it is a signal to stake. Betting should never be a way to make money. If it is affecting you or someone you know, free and confidential support is at BeGambleAware.org. 18+.