StableBet
Professor Furlong and Pascal at the AI Lab
THE AI LAB
Pascal turns up with a new system every single week. So we test every one of them, properly, against the market. Come and see how his ideas have actually held up.
And one of them is going to land any day now. Any day, I'm telling you.
THE AI LAB · THIS WEEK'S TEST

This week's betting system, put to the test

A new betting system goes under test here each week. Once it's tested, it graduates to the permanent Leaderboard and gets its own deep-dive.

Under test now: Four-fold on favourites, a real-money return of −30.6p per £1 staked. Read the full experiment →

Recent issues: Week 2 · Week 1

Updated 5 July 2026 · 27,421 races settled
PASCAL vs THE PROFESSOR

Betting systems, tested on real races

24
systems tested
27,421
real races
0
made a profit
−5p
best case, per £1
Professor Furlong
Professor Furlong

Pascal brings the hunches, I bring the maths. Every common way to bet, ranked by what it actually costs you, on the same 27,421 real races with the fallers counted as the losing bets they are. Open any system to hear his pitch, then the plain truth.

Professor Furlong explaining the numbers at the chalkboard

The whole field, ranked

Every common bet, measured the same way on the same 27,421real British races, with fallers and pulled-up horses counted honestly. See which staking and selection methods hold up best against the bookmaker's margin, and how much harder the flashier bets have to work.

SystemLong-run return, per £1 stakedPer £1Hit rate
1The AI's most-confident picksThe Professor−0.8%
2Odds-on favourites onlyThe systems−4.6%wins 60%
3Favourite in a big fieldThe systems−6.0%wins 24%
4Favourite in a small fieldThe systems−7.4%wins 46%
5Favourite over jumpsThe systems−7.8%wins 37%
6Festival favouritesThe systems−8.3%wins 32%
7Each-way the favouriteThe basics−8.6%places 62%
8Back the favouriteThe basics−8.7%wins 35%
9Favourite on soft groundThe systems−8.8%wins 35%
10Favourite in handicapsThe systems−8.8%wins 30%
11Favourite on the FlatThe systems−9.3%wins 34%
12Follow the AIThe Professor−11.4%
13Back the second favouriteThe basics−12.0%wins 21%
14Top-rated horseThe basics−15.9%wins 19%
15Top-rated in handicapsThe systems−16.4%wins 16%
16Double on two favouritesThe multiples−16.7%
17Lucky 15 on favouritesThe multiples−17.5%
18A random horseThe basics−21.6%wins 13%
19Back the lowest drawThe systems−25.1%wins 12%
20Back the old stagersThe systems−27.2%wins 8%
21Each-way an outsiderThe basics−29.9%wins 7%
22Four-fold on favouritesThe multiples−30.6%
23Back the outsiderThe basics−34.5%wins 3%
24Four-fold on random horsesThe multiples−62.1%
Pascal the punter unveiling his latest betting angle

Explore every system

Pascal reckons he has found the angle. He has 24 of them, truth be told. Filter by type, then tap any system to hear his pitch first, before the Professor shows you, in plain numbers, the verdict and the real-money graph.

The AI's most-confident picks−0.8%
−0.8p
Return per £1
—
Strike rate
361 bets
Sample
—
95% range
Pascal
Pascal “If the Professor's machine is this sure about a horse, that has to be where the money is. Wait for the big confident picks, pile in, and let the clever computer do the hard work for me.”
Professor
The Professor No. Backing only the picks the model rates 35 percent or better still lost money, about -0.8% to starting price over 361 bets. The reason is plain: the model being confident is not the same as the horse being a good price. By the time the model and the market both agree a horse should win, the bookies have already shortened it to match, so there is no spare value left to take and the overround quietly does its work. We land this close to break-even only because the sample is small and jumps-heavy, which is well within the swings of luck, not because anything has beaten the book. It is not a profit and it is not an edge, and the moment you take real-world prices and pay commission it slides back into a clear loss.
Starting price (SP) The odds a horse is at when the race starts, used to settle bets at the off.
Within noise A result close enough to break-even that it could just be luck, not a real edge.
Read the full analysis →
Professor Furlong and Pascal head to head over the ledgers

Where £200 a week actually goes

A real punter staking £200 a week, balance running week by week from January 2025. The lines wobble (a longshot lands, a favourite obliges) but they all drift the same way: down. Tap a system to add or remove it.

−£5,000−£4,000−£3,000−£2,000−£1,000£0Jan ’25Apr ’25Jul ’25Oct ’25Jan ’26Apr ’26Jul ’26balance

Accumulators are left off this graph on purpose: over any short run their path is pure luck, so plotting one would mislead. The maths for them is in the table and on their own pages.

Betting systems: the questions everyone asks

Straight answers, from the data, to what people (and search engines) ask about these systems.

The basics

Do betting systems actually work?
No. We tested 24 common systems against 27,421 real British races, with fallers and pulled-up horses counted as the losing bets they are, and not one made a profit. The least-bad bet, odds-on favourites, still leaks about 4.6%, and most lose far more: backing the favourite is down 8.7%, a random horse 21.6%, a four-fold of favourites 30.6%. The bookmaker's built-in margin is in every price, and no fixed selection or staking rule removes it.
Can you beat the bookies long term?
Not with a system. Every one of the 24 systems we tested across 27,421 races lost money to starting price, and the least-bad of the lot, odds-on favourites, still leaks about 4.6%. The reason is the over-round, the margin baked into every price, which averages about 12% a race. We even built an AI model to find a fairer price and it was less accurate than the market itself (model Brier 0.1025 versus the market's 0.0927), which shows how hard the book is to beat.
What does it really mean that no system beats the bookmaker?
It means the loss is structural, not bad luck. The bookmaker prices every race so the runners' implied chances add up to more than 100%, so the average bet is a loser before the race is run, and no fixed rule strips that margin out. We confirmed it by testing all 24 systems on real results, counting fallers as losers, and by building a model that still could not out-price the market. The honest takeaway is to bet only for fun with money you can afford to lose, never as a way to make money.

The bookmaker edge

What is the over-round (the bookmaker's margin)?
The over-round is the bookmaker's built-in edge. Add up the implied chances of every runner from their prices and a fair book would total 100%, but a real book totals more, and that extra is the margin. Across our 27,421 races it averages about 12% per race, and it grows with field size: small fields are close to fair, while big handicaps of 16 or more runners carry around 30%. That margin is why the average bet loses before luck even comes into it.
How much does field size change the bookmaker's margin?
A lot. The over-round is close to fair in small fields, sits around 12% in mid-sized races, and climbs to roughly 30% in big fields of 16 or more runners. That is why large, competitive handicaps are the worst value on the card: more runners give the bookmaker more places to bury the margin. If value matters to you, small fields cost you the least, though they still cost you.
What is the favourite-longshot bias?
It is the long-standing pattern that short-priced horses are less bad value than long-priced ones, even though both lose on average. In our data the loss rate climbs steadily with the odds: odds-on favourites lose about 4.6%, while genuine outsiders are down about 34.5% and win just 2.6% of the time. Punters over-bet longshots chasing the big win, so bookmakers pad those prices the hardest. The bias means short prices are less bad value, not that any of them is good value.
Why do longshots lose more than favourites?
Because of the favourite-longshot bias. Punters are drawn to the big payout of a long price, so outsiders are over-bet relative to how often they actually win, which lets bookmakers pad those prices hardest. In our data backing the outsider loses about 34.5% and wins only 2.6% of the time, against about 4.6% lost on odds-on favourites. The longer the price, the more of the bookmaker's margin you quietly absorb.
Why does backing the favourite still lose if favourites win most often?
Because winning often is not the same as being priced fairly. Favourites do win the largest share of races, but their prices already build in the bookmaker's margin, so the payout when they win does not cover the times they lose. Spread across our 27,421 races that leaves backing the favourite about 8.7% down to starting price. Winning frequently protects you from big swings, but it cannot overcome a price set against you.

Specific bets

Is backing the favourite profitable?
No. Backing the favourite in every race loses about 8.7% of your stakes to starting price over the long run. Favourites win often, but the bookmaker's margin still grinds you down, and beaten favourites that fall or pull up cost you the whole stake. It is steadier than chasing big odds, but steady losing is still losing.
Is the second favourite a good bet?
No. Backing the second favourite loses about 12.0% to starting price over the long run, more than backing the favourite (8.7%). This follows from the favourite-longshot bias: as you step out to longer prices the value gets worse, not better. The idea that the second favourite is an overlooked sweet spot does not hold up in the data.
Is each-way betting better value?
No. Each-way betting is two bets in one, a win part and a place part, so it does not escape the bookmaker's margin, it just splits your stake across it. Each-way the favourite loses about 8.6% over the long run. The place part pays out more often, which softens the swings, but the standard place terms are set to keep the house edge intact. It manages variance, it does not turn a losing edge into a winning one.
Are accumulators (accas) worth it?
No, they are the worst value of all because they multiply the bookmaker's margin once for every leg. A four-fold of favourites loses about 30.6%, a Lucky 15 about 17.5%, and a four-fold of random horses about 62.1%, so well over half of every pound staked disappears. All the legs landing together is rare, which is exactly why the advertised payouts look so big. The size of the prize is a measure of how unlikely you are to collect.
Is the jumps favourite the one system that has made money?
No. Backing the favourite over jumps loses about 7.8%, but it is not the worst favourite bet, and it never made money. So many jumpers fall or pull up that you lose all those stakes, yet once everything is settled honestly the jumps favourite is actually one of the less-bad favourite bets: the favourite on the Flat (about 9.3%) and the favourite in handicaps (about 8.8%) are both worse. An earlier figure on this page overstated the jumps damage, because a settlement bug wrongly counted winnerless races as losses and made it look like the worst of the lot. We have corrected that. The fallers still cost you and it is still a clear loss, it is just not the steepest favourite bet.

Systems and tipsters

Do staking plans like the Martingale help?
No. Plans like the Martingale, where you double your stake after every loss to chase back what you lost, change the shape of your wins and losses but never the underlying edge. Every bet still carries the bookmaker's margin, so the long-run expectation stays negative. Worse, a normal losing run can balloon your stakes to amounts you cannot cover or that hit the bookmaker's maximum bet, turning a string of small losses into one big one.
Is following a tipster or model profitable?
Almost never once costs are included. A tipster or model can only profit by consistently finding horses priced longer than their true chance, and the market is very good at pricing, so that edge is rare and fragile. Our own AI model, trained on form, conditions, and weather, was less accurate than the market (a Brier score of about 0.1025 versus the market's 0.0927), and backed blind it lost about 11.4%, much like the favourite. Past results and confident write-ups are not evidence of a future edge.
Did your AI model find a way to win?
No. Backed blind in every race the model loses about 11.4%, much like backing the favourite. Its most-confident picks roughly break even, but that is on a small, jumps-heavy sample and sits within noise of zero, so it is not a profit. The model is actually less accurate than the market (Brier 0.1025 versus 0.0927), which is the clearest proof that the book is hard to beat: even a purpose-built model could not out-price it.

The bottom line

What is the safest way to bet?
There is no genuinely safe way to bet, because every bet carries the bookmaker's margin and the long-run expectation is negative. If you do bet, the least-bad value is short prices in small fields, where the over-round is lowest and the favourite-longshot bias hurts least, while accumulators and big-field longshots are the worst. The only truly safe choice is to treat any stake as money you are happy to lose for entertainment, and to set a fixed budget. Betting is not a way to make money.
So what should I take away from all this?
That the maths is settled and it is not in your favour. Across 27,421 real races not one of the 24 systems we tested made a profit, the least-bad still leaks about 4.6%, and the bookmaker's roughly 12% margin per race is the reason. No selection rule, staking plan, tipster, or AI removes it. Bet only for fun, with money you can afford to lose, and never as a plan to get ahead.
The full working, including the faller-counting bug we caught, is on the methodology page. Follow the model day to day on the track record, or read how it is built in the model write-up. 18+ only. This is research and education, not betting advice or a tipping service. Gambling can be addictive; please play responsibly.
Glossary: the words punters throw around ▾
Favourite Racing slang for the favourite, the horse with the shortest odds in the race.
Overround (the margin) The bookmaker's built-in cut: the prices in a race add up to more than 100%, and that extra is the edge you pay on every leg.
Starting price (SP) The odds a horse is at when the race starts, used to settle bets at the off.
Overround The bookmaker's built-in margin: add up the implied chances of every runner and it comes to more than 100%, and that extra slice is the house edge you pay on every bet.
Longshot / rag A runner at a very long price that the market gives almost no chance of winning.
Favourite-longshot bias Favourites lose the least and big outsiders lose the most, because punters overbet long prices and the bookies price that in.
Each-way (E/W) Two bets in one: half your stake on the horse to win, half on it to finish placed (usually top 3 or 4), so the total outlay is double your unit stake.
Place terms The bookmaker only pays a fraction of the win odds for a place, commonly a fifth or a quarter, which is why placing on a short-priced favourite returns very little.
Official rating (OR) A number the official handicapper gives each horse for its ability; higher means rated better.
Each-way Two bets in one: half your stake on the horse to win, half on it to finish placed (usually top 3 or 4), so you stake double your unit.
Odds-on A price shorter than even money, where you stake more than you stand to win (e.g. 4/6 returns £4 profit on a £6 bet).
To SP / starting price The odds at the off, before commission. The real prices you actually get are usually a touch worse.
Faller / pulled up A jumper that falls, unseats, or is eased to a stop over hurdles or fences. The horse does not finish, so the bet loses and your whole stake is gone.
Handicap A race where the better-rated horses are made to carry more weight, so all the runners theoretically have an equal chance.
Official rating A number the handicapper gives each horse for its ability; higher means rated better, but it also earns more weight to carry.
Going The official description of how the ground is riding, from firm through good down to soft and heavy. The softer it is, the harder the slog.
Mudlark A horse that runs particularly well on soft or heavy ground — a going preference the whole market can read in its form figures.
Draw The numbered starting stall a horse breaks from in a Flat race; stall one is usually the closest to the inside rail.
Draw bias A course-and-distance quirk where certain stalls win more often than chance — usually the bend, or faster ground on one side of the track.
Festival A flagship multi-day meeting — Cheltenham, Aintree, Royal Ascot — where the best horses meet and the prize money is at its highest.
Prize money The purse a race pays its finishers. Bigger purses attract stronger fields, sharper betting markets and far more scrutiny of the form.
Old stager A veteran racehorse — here aged ten or older — often a long-serving handicapper the regulars know by name.
Handicap mark The official rating that sets the weight a horse carries. A fading veteran's mark keeps falling, which tempts punters into thinking it is thrown in.
Win double One bet on two selections where both must win for it to pay out; the first winner's returns roll onto the second.
To SP Settled at the starting price, the official odds at the off, before any bookie commission or shorter early prices.
Accumulator (acca) One bet on several horses where every leg must win; the winnings from each roll onto the next.
Lucky 15 One bet covering four selections as 4 singles, 6 doubles, 4 trebles and 1 four-fold, fifteen bets in total.
Margin (overround) The bookmaker's built-in edge, the slice taken out of every price so the book pays back less than the true odds.
ROI to SP Profit or loss as a percentage of money staked, settled at the official starting price. -11.36% means you lose about £11 of every £100 you put on.
Over-round The bookmaker's built-in margin: add up the chances implied by every price in a race and they total more than 100%, and that extra slice is the edge stacked against you.
Within noise A result close enough to break-even that it could just be luck, not a real edge.

See what each costs a real punter →

NEW · PASCAL'S PANEL
The Betting Systems Leaderboard →
Most betting-system advice comes from someone with something to sell. This doesn't. Every common system, tested on real races and ranked by real-world return, and why the bookmaker's margin is so hard to beat.

Browse every weekly issue of Pascal's Panel →