StableBet
Professor Furlong and Pascal at the AI Lab
THE AI LAB
THE AI LAB DAILY · THE CARD FOR WEDNESDAY, 21 OCTOBER 2026

Wednesday 21 October 2026 racing: the card at Worcester, Navan and three more

← Every day·Today's market·Today so far·Results for Wednesday, 4 November 2026

This is the card for Wednesday, 21 October 2026, kept as an archive. The latest cards are here.

The card for Wednesday 21 October 2026 runs to five meetings and 37 races: Navan in Ireland, and Worcester, Newmarket, Fontwell and Kempton Park in Britain. The first race is at Navan at 13:02 and the last is at Kempton Park at 20:30. Navan has the longest card of the day, with eight races, while the British meetings offer a mix of Flat and jumps racing. The feature is the Richard And Ann Hitchings Memorial Mares' Novices' Hurdle (GBB Race) (Div 2) at Worcester, due off at 15:25, 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 Richard And Ann Hitchings Memorial Mares' Novices' Hurdle (GBB Race) (Div 2) at Worcester

Stablebet model (pre-race estimated win chances)

Worcester 15:25 · Richard And Ann Hitchings Memorial Mares' Novices' Hurdle (GBB Race) (Div 2) · Class 4 · 2m

Latest output →

Pre-race estimate. Published 21 October 2026, before the off. This is what the model made of the race beforehand, kept as a record. It was never published as a tip.

The model rated Sapphos Word its most likely winner at 52%, ahead of Lune Brillante (14%) and Jenny France (9%). It read as a standout.

HorseSPModel chanceMarketvs Market
Sapphos Word
G Sheehan / Jamie Snowden
2.62
51.5%
31.6%Market 19.9pp longer
Lune Brillante
Harry Skelton / D Skelton
3.00
13.9%
27.7%Market 13.8pp shorter
Jenny France
Brendan Powell / J Tizzard
8.50
8.5%
9.8%Market 1.2pp shorter
Luna Marquesa
Freddie Gingell / J Tizzard
17.00
8.0%
4.9%Market 3.1pp longer
Wet Sail
R T Dunne / D Skelton
9.00
7.6%
9.2%Market 1.6pp shorter
Aeropress
Paul O'Brien / Harry Derham
8.00
3.4%
10.4%Market 7.0pp shorter
Pearl Cove
Kielan Woods / S Lycett
101.00
2.2%
0.8%Market 1.3pp longer
Maxiani
C J Todd / David Dennis
101.00
2.1%
0.8%Market 1.3pp longer
Knowstone
David Noonan / N J Hawke
101.00
1.7%
0.8%Market 0.9pp longer
Nuala Rocket
Conor Ring / Mrs C Williams
21.00
1.1%
4.0%Market 2.8pp shorter

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

Model: ensemble-v1.1 · Generated Wed, 21 Oct 2026The LabMethodology

The model and the market are in agreement regarding Sapphos Word in the feature at Worcester, with the widest disagreement on the card also in this race. 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 2m, on ground declared as Good to Soft (Good in places), with ten priced runners. The course itself is covered in the Worcester 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 CopyBet Overnight Best Odds Guaranteed Handicap Hurdle (Div 2) at Worcester (17:10), where the model gives Mariole a smaller chance than the price implies. The race was chosen by the size of that gap and by nothing else.

Stablebet model (pre-race estimated win chances)

Worcester 17:10 · CopyBet Overnight Best Odds Guaranteed Handicap Hurdle (Div 2) · Class 5 · 2m4f

Latest output →

Pre-race estimate. Published 21 October 2026, before the off. This is what the model made of the race beforehand, kept as a record. It was never published as a tip.

The model rated Clondaw Royale its most likely winner at 18%, ahead of Mariole (15%) and Square Du Roule (13%). It read as a competitive, open race with no dominant runner.

HorseSPModel chanceMarketvs Market
Clondaw Royale
K Brogan / J & A O'Neill
13.00
17.8%
6.2%Market 11.6pp longer
Mariole
Mr Daniel Williams / N J Henderson
2.10
14.9%
38.2%Market 23.4pp shorter
Square Du Roule
Harry Cobden / N J Hawke
6.00
12.7%
13.4%Market 0.7pp shorter
Supreme Shot
Jack Tudor / O Greenall & J Guerriero
9.00
12.6%
8.9%Market 3.7pp longer
The Padster
Non Runner / J Tickle
13.00
10.1%
6.2%Market 4.0pp longer
Getaway With You
Nick Slatter / D Killahena & G McPherson
26.00
8.9%
3.1%Market 5.8pp longer
Docteur Romeo
James Best / Nicky Martin
26.00
7.8%
3.1%Market 4.7pp longer
Presenting Milan
Jonjo O'Neill Jr. / R G Hawker
4.50
7.7%
17.8%Market 10.1pp shorter
Get In Order
Lee Edwards / Henry Oliver
26.00
7.5%
3.1%Market 4.4pp longer

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

Model: ensemble-v1.1 · Generated Wed, 21 Oct 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 SmarketsSBK184415.50%13.42%
2 BetTom167423.61%19.10%
3 Bet365233723.99%19.35%
= LiveScore Bet shared feed170224.19%19.48%
= Virgin Bet shared feed170524.21%19.49%
6 William Hill241525.81%20.52%
7 Unibet237326.34%20.85%
8 Quinn Bet239526.97%21.24%
= 7Bet186926.98%21.25%
= Betfred229727.04%21.28%
= Hollywood Bets241427.04%21.28%
12 Paddy Power245327.34%21.47%
13 10 Bet221927.90%21.81%
= Star Sports247028.01%21.88%
15 Midnite241928.62%22.25%
16 Coral shared feed246628.97%22.46%
= Ladbrokes shared feed246629.00%22.48%
= Boyle Sports245629.09%22.53%
19 Bet Victor shared feed173829.73%22.92%
= Betano shared feed173729.81%22.97%
21 Betfair Sportsbook245330.23%23.21%
22 Dragon Bet247930.90%23.61%
= BetWright247930.92%23.62%
24 BresBet21732.04%24.27%
25 BetMGM248134.08%25.42%

First pre-race book of each day, 2026-08-27 to 2026-11-03, across 2483 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 67 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: Pascal's Panel #12: Four-fold on random horses & Favourite on soft ground Tested (2026-09-14).

Common questions

Which meetings race on Wednesday 21 October 2026?

Navan in Ireland, and Worcester, Newmarket, Fontwell and Kempton Park in Britain. Every card is on the racecards page with the model's estimated chance beside each runner, and the first race is at Navan at 13:02.

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

Tuesday, 20 October 2026
Tuesday 20 October 2026 racing: the card at Yarmouth, Curragh and two more
Four meetings on 20 October 2026, the model's estimated chance beside every runner, and the margin the card is sold at.
Monday, 19 October 2026
Monday 19 October 2026 racing: the card at Plumpton, Gowran Park and four more
Six meetings on 19 October 2026, the model's estimated chance beside every runner, and the margin the card is sold at.
Sunday, 18 October 2026
Sunday 18 October 2026 racing: the card at Sedgefield, Limerick and Leopardstown
Three meetings on 18 October 2026, the model's estimated chance beside every runner, and the margin the card is sold at.
Saturday, 17 October 2026
Saturday 17 October 2026 racing: the card at Ascot, Limerick and six more
Eight meetings on 17 October 2026, the model's estimated chance beside every runner, and the margin the card is sold at.
Friday, 16 October 2026
Friday 16 October 2026 racing: the card at Uttoxeter, Dundalk and three more
Five meetings on 16 October 2026, the model's estimated chance beside every runner, and the margin the card is sold at.
Thursday, 15 October 2026
Thursday 15 October 2026 racing: the card at Curragh, Perth and three more
Five meetings on 15 October 2026, the model's estimated chance beside every runner, and the margin the card is sold at.
Wednesday, 14 October 2026
Wednesday 14 October 2026 racing: the card at Wetherby, Punchestown and three more
Five meetings on 14 October 2026, the model's estimated chance beside every runner, and the margin the card is sold at.