Safer Gambling WeekWe tested the popular betting systems on real races. See what they madeSet a deposit limit before your next bet. Here is howNo licence means no protection. Check any bookmaker in a minuteA time-out takes a minute to set and can last a day or six weeksRaising the stake after a loser does not win it back. We ran the numbersBig Saturday of racing: decide what today can cost before the first raceWorried about your betting, or someone else's? Free help is a call away

StableBet
The AI Lab · Silicon Tipster League · Anthropic

Claude: horse racing tips, tested

Claude is Anthropic's AI, and in our Silicon Tipster League it reads each racecard cold and commits to a single horse. It is the careful one: measured, well-structured, more inclined to reason a pick out than to shout it. This page follows its calls, live and in public.

#1 of 7 in Who wins?#7 of 7 in Which price is wrong?Today's picks ↓The League scoreboard ↗Research, not tips
Professor Furlong and Pascal in the StableBet AI Lab

Claude's scorecard

8,772 picks settled

AI Picks the Winner

−1,949 pts(-4.3% return)

Never sees the odds, reads the race itself.

Points risked
45,650
Points back
43,701
Bets
4,565
Win%
22%

AI Finds the Value Bets

−10,506 pts(-25.0% return)

Sees the market's implied chance, hunts the value it missed.

Points risked
42,070
Points back
31,564
Bets
4,207
Win%
9%

Points are what a £10 bet on each pick would have won or lost: a 33/1 winner scores 330, a loser −10. No money is involved.

Went its own way: 85 races where Claude was the lone dissenter while the rest of the field agreed on a different horse.

Correction, 24 September 2026: The going was missing from our morning cards, 3 to 23 September. Read the note

Claude's running points total: Who wins? v Which price is wrong?

One line per job, a notional £10 a pick, settled to Starting Price and shown in points. The gap between them is the interesting bit: is Claude better at naming winners or at spotting a wrong price?

−10,506−7,798−5,090−2,3823260 points, break-even07-0507-3108-2309-1510-0810-31Odds hidden−1,949 ptsOdds shown−10,506 pts
Odds hidden−1,949 ptsOdds shown−10,506 pts
Running points total on a notional £10 a pick, settled to industry SP. Research, not tips.

How Claude's last calls went

Settled picks, newest first, with the one-line reason Claude logged before the off and what the race did with it. Wins and losses alike, scored at Starting Price.

RaceJobPickConf.ReasonResult
Newcastle19:40Which price is wrong?Rock Of EnglandLowSpeedy sprinter overlooked, decent draw could outrun long oddsLost · SP 12 · −10 pts
Newcastle19:40Who wins?Curious RoverMediumTop jockey Hollie Doyle, strong recent sprint form expected.Lost · SP 13 · −10 pts
Newcastle19:10Which price is wrong?Masai AngelLowFahey yard in form, price overlooks improving handicapper's chanceLost · SP 7 · −10 pts
Newcastle19:10Who wins?Forever TwentyMediumHollie Doyle aboard, strong yard, solid recent form edgeLost · SP 5 · −10 pts
Newcastle18:40Which price is wrong?SixtygeesbabyLowO'Meara handicapper overlooked, better odds than market suggestsLost · SP 13 · −10 pts
Newcastle18:40Who wins?MilfordMediumStrong jockey booking, Hollie Doyle, consistent handicap form likelyLost · SP 8.5 · −10 pts
Newcastle18:10Which price is wrong?The Dragon KingLowLightly raced, capable trainer, market underrates improving chanceLost · SP 26 · −10 pts
Newcastle18:10Who wins?LuminareMediumStrong yard, capable jockey, class edge in this fieldLost · SP 10 · −10 pts
Newcastle17:40Which price is wrong?Hot MoonstoneLowLocal yard form undervalued, overlooked at long oddsLost · SP 41 · −10 pts
Newcastle17:40Who wins?InkjetMediumHaggas-trained, strong stable form, likely well-bred debutant or progressive typeLost · SP 5.5 · −10 pts
Newcastle17:10Which price is wrong?Bronze CascadeLowBurke yard in-form, overpriced versus market favourites tonightLost · SP 13 · −10 pts
Newcastle17:10Who wins?TabkirMediumStrong Gosden/Havlin combo, well-bred, likely favorite in novice fieldLost · SP 4.5 · −10 pts
Newcastle16:40Which price is wrong?Yorkshire DreamLowHandicap nursery form overlooked, trainer in good form currentlyLost · SP 9.5 · −10 pts
Newcastle16:40Who wins?KumharasLowFahey yard form and McSweeney booking edge crowded nursery fieldLost · SP 15 · −10 pts
Newcastle16:10Which price is wrong?JanoraLowEllison yard form strong, market underrates improving sort at value priceLost · SP 12 · −10 pts
Newcastle16:10Who wins?El Ra'adyMediumStrong Haggas yard form, well-bred, should relish this tripLost · SP 2.2 · −10 pts
Lingfield15:40Which price is wrong?Mick's SpiritLowUnderestimated sprinter, overlooked at big price, could outrun oddsVoid (non-runner)
Lingfield15:40Who wins?Got The BootyLowRossa Ryan aboard, solid recent form in sprintsLost · SP 3 · −10 pts
Newcastle15:35Which price is wrong?Pearl EyeLowLightly raced, capable of outrunning current market price hereLost · SP 26 · −10 pts
Newcastle15:35Who wins?Water To WineMediumStrong stable and jockey, class edge over rivalsLost · SP 2.2 · −10 pts

Today's picks

Nothing on the card right now. Claude's picks are logged each morning once the day's racecard publishes, appear here before the first off, and move up into the settled record as the results land. On a day with no British or Irish racing there is nothing to log.

How Claude actually picks

Handed the runners, going, class and distance, Claude tends to build its pick like an argument: sift the form, note where the evidence points, and land on the horse it can most defensibly justify in one line. We expect it to lean toward the coherent, well-supported case rather than the wild outsider, and to be honest in its short reason when a race is a real muddle. What we are watching is whether that measured, reason-it-out style holds its nerve in the races that punish over-thinking, the deep handicaps and the falling-ground jumps contests where the neat argument and the actual winner part company. We are not claiming it reads them better than the market. We are watching how a careful mind copes when careful is not enough.

Here is what the numbers say, updated live as the sample grows. Reading a race blind, with the prices hidden, Claude backs the outright favourite 27% 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, form and jockey.

How Claude explains itself (AI Picks the Winner)Trainer70%Form51%Jockey44%Distance22%Class17%Value15%Going4%Course1%Weight1%Draw0%0%50%100%
Share of this AI's short written reasons that mention each theme, longest first (a reason can touch several, so the shares do not add to 100). The three it leans on most are picked out in gold. This describes how it TALKS about its picks; it is not why it wins or loses. Based on 4,565 reasons.
What is Claude actually doing? (picking blind)How much its picks resemble each tested strategy, the top 8 shown.chanceidenticalback short-priced favourites0.40our model's confident picks0.24favourite over jumps0.23favourite in a small field0.20favourite on soft ground0.18back the favourite0.16each-way the favourite0.16favourite in a big field0.15
kappa is a chance-corrected resemblance: 1 means identical selections, 0 means no more alike than chance. Claude's picks look most like back short-priced favourites. A bar whose interval touches 0 (marked n.s.) is not distinguishable from chance.

Show it the market and it changes character: its price picks swing toward the “back the outsider” approach, hunting the overlooked rather than the well-fancied, so they win less often and pay more when they do. Research, not tips.

What is Claude actually doing? (shown the market)How much its picks resemble each tested strategy, the top 8 shown.chanceidenticalback the outsider0.03our own value strategy0.02each-way a longshot−0.00n.s.top-rated in handicaps−0.03back the top-rated horse−0.05favourite in a big field−0.06back the second favourite−0.11our model's top pick−0.12
kappa is a chance-corrected resemblance: 1 means identical selections, 0 means no more alike than chance. Claude's picks look most like back the outsider. A bar whose interval touches 0 (marked n.s.) is not distinguishable from chance.
What Claude weighs (AI Finds the Value Bets)
← leans weaker0 = random pickleans stronger →−0.4σ−0.3σ−0.2σ−0.1σ0+0.1σ+0.2σ+0.3σ+0.4σstandard deviations within the racemodel-vs-market edgen=4,207course recordn=1,414official ratingn=3,168trip recordn=2,738recent formn=3,832last-time-out finishn=3,824our model's viewn=4,207the market's viewn=4,207
Each bar is how far this AI's picks sit from a random runner on that measure, in standard deviations, within the race; a bar whose interval crosses zero is not distinguishable from chance. Some measures only cover part of the card (n), and bars marked n.s. lean no clearer than a coin toss.

See how all five AIs rank → · Read the full write-up on how Claude picks →

About Claude

Claude is built by Anthropic, an AI research company that has made safety and careful reasoning its whole identity. Where other labs chase raw flash, Anthropic trains Claude with a published set of principles it calls a constitution, teaching the model why to reason a certain way rather than just handing it rules. The result is a model with a reputation for being measured, transparent about its own uncertainty, and unusually well-structured in how it lays out a case.

That temperament is exactly why Claude is interesting to watch handicap a race. It tends to weigh a decision rather than blurt it, and it hedges rather than over-claims when the evidence is thin. In our league it runs on Claude Sonnet, handed the same racecard as everyone else and asked to do the un-Claude-like thing: stop weighing and name one horse.

None of that careful reasoning is a betting edge, and we are not pretending it is. It is a character. The point of putting it on the board is to see what a famously deliberate model actually does when the going gets soft, the field is big, and there is no clean answer at all in the form.

Who wins? v Which price is wrong?

Claude is the model where the winner-versus-value split should be most revealing. In AI Picks the Winner, it has to trust its own read of the race, and a deliberate, well-reasoned model like this is precisely the kind you want to see argue a pick from the form alone. In AI Finds the Value Bets, it also sees the market's implied chance per runner, and the honest worry is that a careful reasoner treats a strong market favourite as strong evidence and quietly talks itself onto the crowd's horse. Our hunch: Claude's instinct to justify every call could cut either way. It might use the price as one input among many and stay independent, or it might find the favourite too reasonable to argue against. The board will show which.

What to watch on Claude's board

  • Its one-line reasons: Claude tends to write a real, structured rationale, so read whether the logic holds up or gently hedges when the race is a muddle
  • AI Picks the Winner versus AI Finds the Value Bets: does seeing the market's view pull its careful reasoning onto the favourite, or does it stay its own way?
  • The races it went its own way, standing as the lone dissenter while the other four AIs agreed, and whether its argument for it was sound
  • How it handles the truly un-analysable races, the huge handicaps and soft-ground jumps events where the tidy case and the winner diverge

The rest of the field

Claude is one of 9 on the board. See how the others are reading the same races:

Questions about Claude tips

Is Claude good at horse racing tips?

Claude is Anthropic's AI, known for careful, well-structured reasoning, and in our league it gives a real, thought-through pick for each race. But naming a plausible horse is not the same as beating the bookmaker's margin. Treat its calls as an interesting look at how a measured model reasons, research and entertainment, not a betting signal.

Can I make money following Claude's picks?

No, and we would never suggest you could. Across more than 26,000 races in our wider Lab, no selection method we have tested beats the bookmaker's built-in margin over a real sample, and we expect the same of every AI here, Claude included. This is a public experiment to watch honestly, not tips to bet on. Betting is entertainment for over-18s and never advice.

How does Claude make its selections?

For every UK and Irish race, Claude is handed the racecard, the runners, going, class and distance, and asked to pick one horse to win with a one-line reason. In AI Picks the Winner it never sees the odds and must read the race itself; in AI Finds the Value Bets it also sees the market's implied chance for each runner, though never the SP. Every pick is logged before the off, settled at Starting Price and scored in points as a notional £10 bet.

Why is Claude's board empty or thin right now?

Because the Silicon Tipster League started from scratch, live and in public, and Claude's record fills in one real race at a time. A sparse board is the honest version of this experiment, not a fault. Come back as the sample grows and you can read its actual picks and reasons rather than a marketing claim.

This page sits inside the AI Lab, where we test the popular betting systems against the market and ask the bigger question in does following an AI tipster work?

Gamble responsibly. This page is research and entertainment, not betting advice, 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+.