StableBet
Professor Furlong and Pascal at the AI Lab
THE AI LAB
LAB NOTES · LAB NOTES

You v the AI at Cheltenham: The November Meeting Challenge

Cheltenham's November Meeting has its own You v the AI challenge this year. You pick one horse in every race on Friday, Saturday and Sunday, and your picks go on the same board as our racing model, the 5 AI chatbots in our Tipster League and the market favourite.

It is free. No money is staked and there are no prizes. What you get is a straight answer to a question most racing fans have asked themselves: over a whole meeting, would my picks have done better than a machine's, or better than just backing the favourite?

Play the Cheltenham November Meeting challenge

The meeting is a good place to ask it. Friday's best races are small fields of good horses, led by the Shloer Chase. Saturday and Sunday are built around two big handicaps, the Paddy Power Gold Cup and the Greatwood Handicap Hurdle, where few favourites have won in recent years. So the meeting asks for two different kinds of judgement.

Everyone on the board is scored the same way, at the official starting price, so nobody gets credit for a price they could not have had. That includes our own model. Its picks are published before racing and are locked from then on, and its results go on the board whether they flatter it or not.

How the challenge works

The races. Every race at Cheltenham on Friday, Saturday and Sunday, one horse in each. There is no shortlist of featured races as there is in the daily game: if it is run at the meeting, it is in the challenge.

When picks open. Each day's card opens the afternoon before, once the runners are declared. Until then the page shows the entries, and any picks you make on that provisional card are saved on your device and carried over to the final card. A pick stays open until its race is due off, so you can change your mind on the day.

Locking in. You lock in your card with a username. There is no email and no password.

The opposition. Our racing model makes its picks from the early-morning card on race day. The market favourite is taken from the same card, at the same moment. The chatbots are the entrants in our Silicon Tipster League, which pick before racing with the odds hidden from them.

The scoring. Every pick is treated as a notional tenner at starting price and shown as points, never as money. A winner scores ten points for each point of its starting price, so a winner at long odds is worth far more than a winner at short odds. A loser costs ten. A non-runner scores nothing.

The board. The meeting has its own leaderboard, separate from the daily game. You need to play on two of the meeting's days to be ranked on it. When the last race on Sunday is settled, whoever is top is the meeting's champion.

The daily game carries on. The Daily Five runs as usual alongside it. If a Cheltenham race is also one of that day's five, one pick counts on both boards.

What you are up against

It helps to know how good the opposition is before you start. We publish all of it.

Our model. Across 11,947 races, the model's top pick has won 24.5% of the time. Backing every one of those picks at starting price has returned −13.6%. It picks plenty of winners and it is not in profit, and the track record shows every race, losers included.

The favourite. Backing the favourite in every race is the simplest strategy there is. In our long-run test of 31,121 races it won 35% of the time and returned -9.00%. The favourite wins often and still loses money, because its price already allows for how often it wins. It is on the board as a baseline: if you cannot beat it over the meeting, your picks added nothing to what the market already knew.

The chatbots. The league has been running for 129 days and has settled 55,708 picks. 5 of its 5 chatbots are behind the favourite baseline. The league table has the standings as they are today.

None of them is unbeatable over one meeting. A long weekend is a small sample, and a single winner at a big price can put anyone top of the board, human or machine. That is worth remembering when you look at the final table: it will say who had the best meeting, and it will not say who is the best judge.

Friday's card is covered in our Cheltenham Friday preview, and the whole meeting is on the November Meeting hub.

Common questions

Does it cost anything to play?

No. It is a free prediction game. No money is staked, there are no prizes, and you need only a username to lock in your picks. It is for adults aged 18 and over.

How many days do I have to play?

Two, out of Friday, Saturday and Sunday, to be ranked on the meeting's board. You can still play a single day and see how your picks scored.

When can I make my picks?

Each day's card opens the afternoon before. Before that you can pick from the entries, and those picks carry over to the final card on the same device. Every pick stays open until its race is due off.

How are the picks scored?

At the official starting price, as a notional tenner on each pick, shown as points. A winner scores ten points for each point of its starting price and a loser costs ten.

Does a festival pick affect my Daily Five score?

Only if the race is also in that day's Daily Five, in which case the one pick counts on both boards. Otherwise the two games are scored separately.

Is the AI's pick a tip?

No. The model is a test we publish in full, including its losses. Nothing on the board is advice to bet.

A word on all of this

This is a game about judgement, not a way to make money from betting. Over time the odds favour the bookmaker. If your betting has stopped being fun, BeGambleAware has free, confidential help, and the National Gambling Helpline is on 0808 8020 133.

Every figure here is pulled live from our data and nothing beats the bookmaker's margin. For whether anyone holds a real edge, see our track record. 18+, please bet responsibly.

More from Lab Notes