StableBet
Professor Furlong and Pascal at the AI Lab
THE AI LAB
THE AI LAB DAILY · TODAY'S CARD

Tuesday 22 September 2026 racing: the card at Beverley, Listowel and three more

← Every day·The market on Monday, 21 September 2026·The day so far on Monday, 21 September 2026·Yesterday's results

The card for Tuesday 22 September 2026 runs to five meetings and 36 races: Listowel in Ireland, and Beverley, Warwick, Fontwell and Southwell in Britain. The first race is at Beverley at 13:30 and the last is at Southwell at 20:30. The model has identified a significant disagreement with the market in the Free Bets On attheraces.com Handicap at Southwell, suggesting that race might be of particular interest. The feature is the Old Gravel Pits Lodges EBF Novice Stakes at Beverley, due off at 14:05, and it has the next section to itself. Every runner, with the model's estimated chance beside the market's price, is on the racecards page, and the model's view of the whole day is on the day's predictions; the card was written before the first race, so nothing on it is a result.

The feature: the Old Gravel Pits Lodges EBF Novice Stakes at Beverley

Stablebet model (estimated win chances)

Beverley 14:05 · Old Gravel Pits Lodges EBF Novice Stakes · Class 3 · 1m100y

Full output →

The model rates Bayraktar its most likely winner at 31%, ahead of Stellar Quest (27%) and Austin (22%). It reads as a clear first choice, but beatable.

HorseSPModel chanceMarketvs Market
Bayraktar
Pierre-Louis Jamin / K R Burke
3.50
31.1%
26.4%Market 4.8pp longer
Stellar Quest
Jack Nicholls / K R Burke
21.00
27.3%
4.4%Market 22.9pp longer
Austin
T Eaves / K A Ryan
17.00
22.1%
5.4%Market 16.7pp longer
Arthurian
Jason Hart / A M Balding
1.44
19.5%
63.8%Market 44.4pp shorter

These are the model's estimated win chances, not tips. The model is calibrated to real results (when it says 25%, about 25% win) but it does not beat the market, so treat it as an independent second opinion for understanding the race. See how accurate it is →

Model: ensemble-v1.0 · Generated Tue, 22 Sep 2026The LabMethodology

The model rates Bayraktar its most likely winner, while the market has Arthurian shortest. The card above sets the model's estimated chance for each runner beside the chance implied by the market, and the gap between the two columns is the only thing the Lab claims to measure; it says which horse the model rates highest and nothing about what to do with that. The race is run over 1m100y, on ground declared as Good, with four priced runners. The course itself is covered in the Beverley guide.

Where the model disagrees with the market

The widest gap between the model's estimate and the market's price on this card is in the Free Bets On attheraces.com Handicap at Southwell (20:00), where the model gives Courageous Sioux a smaller chance than the price implies. The race was chosen by the size of that gap and by nothing else.

Stablebet model (estimated win chances)

Southwell 20:00 · Free Bets On attheraces.com Handicap · Class 5 · 4f214y

Full output →

The model rates Courageous Sioux its most likely winner at 19%, ahead of Hanney Girl (18%) and Managing Director (14%). It reads as a competitive, open race with no dominant runner.

HorseSPModel chanceMarketvs Market
Courageous Sioux
Rossa Ryan / C G Cox
2.00
19.4%
43.8%Market 24.3pp shorter
Hanney Girl
Ryan Sexton / T D Easterby
5.50
17.9%
15.9%Market 2.0pp longer
Managing Director
P J McDonald / G R Oldroyd
11.00
13.9%
8.0%Market 5.9pp longer
Goyard
Oisin Murphy / D O'Meara
11.00
13.1%
8.0%Market 5.1pp longer
Dance Of Angels
Nicola Currie / H Candy
8.00
12.0%
10.9%Market 1.1pp longer
Brazilian Belle
Sean Kirrane / M Dods
13.00
11.9%
6.7%Market 5.2pp longer
Rogue Bullet
L Morris / James Owen
13.00
11.8%
6.7%Market 5.0pp longer

These are the model's estimated win chances, not tips. The model is calibrated to real results (when it says 25%, about 25% win) but it does not beat the market, so treat it as an independent second opinion for understanding the race. See how accurate it is →

Model: ensemble-v1.0 · Generated Tue, 22 Sep 2026The LabMethodology

A gap is a difference of opinion, and a difference of opinion is all it is: the wide ones tend to fall on runners the model has the least evidence about, as often as on a horse the market has overlooked. The left column is the model's estimated chance, the right is the chance the price implies, and the track record shows how often the model's side of the argument has been right once every pick is settled at starting price.

What the card costs

Every price on the card carries the bookmaker's margin: add up the chances the prices imply and the total comes to more than a whole field, and the excess is what the book keeps whichever horse wins. The table below is the latest published margin for each book the Lab tracks. It is compiled from the previous evening's prices, so it describes the card as it was sold then rather than the morning's board, and the order of the table is a measurement, never a recommendation.

BookmakerRacesOverroundTakeout
1 SmarketsSBK62215.86%13.69%
2 BetTom54123.72%19.17%
3 Bet36579324.09%19.41%
4 LiveScore Bet shared feed62624.64%19.77%
= Virgin Bet shared feed62324.71%19.81%
6 William Hill82125.78%20.50%
7 Unibet81826.98%21.25%
= 7Bet67127.21%21.39%
= Quinn Bet80827.21%21.39%
= Hollywood Bets79627.32%21.46%
11 Betfred77527.58%21.61%
= Paddy Power84627.62%21.64%
= 10 Bet76227.75%21.72%
14 Coral shared feed84928.19%21.99%
= Ladbrokes shared feed84928.24%22.02%
= Star Sports84928.38%22.11%
17 Midnite82529.01%22.49%
= Boyle Sports83929.18%22.59%
19 Bet Victor shared feed59629.75%22.93%
= Betano shared feed59429.92%23.03%
= Dragon Bet85030.15%23.17%
= BetWright85030.19%23.19%
23 Betfair Sportsbook84530.51%23.37%
24 BresBet21732.04%24.27%
25 BetMGM85032.78%24.69%

First pre-race book of each day, 2026-08-27 to 2026-09-21, across 852 races every listed book priced in full. Exchanges and spread firms are excluded: they do not quote comparable win odds. Overround is how far past a fair book the prices sit; takeout is the share of stakes the book keeps. On 24 days of data, books marked = are level with the one above rather than behind it. Books marked "shared feed" are one half of a pair that does not price independently, so there are fewer independent books here than rows.

The margin is the reason the Lab's tested strategies lose over time even when a pick is sound; what betting costs has the working, and the Tipster League shows what it does to the AI tipsters over a season.

Elsewhere in the Lab: How Often Would I Need to Win Just to Break Even?.

Common questions

Which meetings race on Tuesday 22 September 2026?

Listowel in Ireland, and Beverley, Warwick, Fontwell and Southwell in Britain. Every card is on the racecards page with the model's estimated chance beside each runner, and the first race is at Beverley at 13:30.

When are the model's estimates published?

Each morning before the first race, from the declared runners, and they are not changed after the off. A card labels itself a pre-race estimate once its race has run.

What does the model's chance mean?

The model's own estimate of how likely each runner is to win, set beside the chance the market price implies. It is an opinion with a published record, not a tip, and the track record is where that record lives.

Why is a horse missing from a card?

Either it was withdrawn after the estimates were made, or the race was one the model did not price. The horse tracker follows any runner on the card to its next entry.

How are the AI tipsters doing?

The Tipster League settles every AI pick at starting price and publishes the running total, losses included.

A word on all of this

None of the above is a tip, and no system the Lab has tested has made money over the whole record. Bet only what you can afford to lose. If gambling has stopped being fun, BeGambleAware has free, confidential help.

Every figure on this page renders live from our published feeds and nothing here is a tip. For whether the model holds any edge, see our track record. 18+, please bet responsibly.

Previous cards