Does backing the outsider work?

Pascal has the slip in his hand and a faraway look in his eye. "One day," he says, "the 50/1 no-hoper romps home and I am set for life. And look at the price, all that upside for my fiver. Someone has to win it, so why not the one nobody fancies."

It is the most romantic bet in racing, and the logic sounds almost clever. The crowd piles onto the favourite, shortening it to poor value, so the value must be hiding at the other end of the book, where nobody is looking. Back the horse the market likes least, the reasoning goes, and you are buying the prices everybody else is too scared to take. One big result and you are square for the season.

The claim rests on the maths of the payout rather than the maths of the edge. A single 50/1 winner wipes out 50 losing bets in one go, so a punter convinces himself that one good day proves the whole method. Outsiders do win just often enough to keep that dream alive, and the human memory does the rest: you remember the 40/1 shock for years, and forget the thirty losing slips that paid for it.

There is also a contrarian flavour to it that feels smart. If backing the favourite loses money, surely doing the opposite must win. It sounds like the kind of thing a sharp punter would spot. The trouble is that it has the market exactly backwards. The long end of the book is not where the bookies have left value lying around. It is where they have buried the heaviest margin of all. We took the claim at face value and tested it on the cold numbers.

The pull of the outsider is the size of the prize, not the chance of landing it, and that is exactly why it works on people. A fiver on a 50/1 shot pays two hundred and fifty pounds plus your stake back. The brain fixes on the payout and quietly ignores the long run of races where the fiver simply vanishes. The payout is vivid and concrete; the long grind of losers is abstract and forgettable.

It is also a bet built around a story. Every punter has either landed a big-priced winner or watched a mate land one, and that single afternoon becomes the headline of the whole habit. The shock result of the festival, the 33/1 bolter that nobody saw coming, gets retold for years. Nobody ever recounts the hundred quiet slips in between that funded it, because there is no story in a beaten 40/1 shot. The wins are memorable and the losses are wallpaper.

Then there is the contrarian comfort. Backing the favourite feels like the mug's bet, the thing everybody does, so backing the outsider feels like thinking for yourself. It carries a whiff of being smarter than the crowd, of spotting the value the herd has overlooked. That self-image is seductive, and it is doing a lot of work to keep the punter staking.

Finally, the small stake makes it feel harmless. A fiver here and a fiver there does not feel like serious money, so the losses never quite register as losses. The dream stays cheap and the bill stays hidden. All of which feels like a sensible long-odds punt with limited downside. The data says it is the most expensive habit on the entire card.

Backing the outsider is the clearest place on the whole card to watch where your money actually goes, because the cost stacks up through three layers at once and they all pull in the same direction.

The first layer is the over-round, the bookmaker's built-in margin. Add up the chances implied by every price in a race and they come to more than 100%: the extra is about 12% per race, and it climbs towards 30% in big fields of 16-plus runners. You pay that cut on every single bet, whatever you back.

The second layer is the favourite-longshot bias, and this is the one that really catches the outsider. The over-round is not spread evenly across the card. Bookmakers load the heaviest part of it onto the big prices, because punters chronically over-back longshots chasing the life-changing payout. A 25/1 shot is rarely a genuine 25/1 chance. It is usually worse, priced shorter than its true odds, because the layer knows the dreamers will take it anyway. Backing the outsider means buying the most over-padded part of the book, every time.

The third layer is the brutal, simple fact that these horses do not win. The longest price on the card wins 3% of the time, roughly one race in every thirty or so. On top of that, plenty of these no-hopers fall or are pulled up over jumps, and a faller is a losing bet, your whole stake gone exactly like any other beaten horse.

Put the three together and the shape of the return is as telling as the size. Winners are rare and clumped, so you sit through long quiet runs waiting for the next big price to land, while the balance ticks down underneath. The wild variance hides the steady cost. A lucky cluster of winners can briefly put you in front and feel like proof, but it is variance flattering a few results, not an edge. Stretch the sample and the over-round, the bias and that tiny strike rate settle the question every time.

We wanted a number that no bookmaker, and no wishful punter, could argue with, so we kept the test plain and the rules strict.

The sample is 27,676 real British races. For every race we identified the outsider, the runner at the longest price in the field, and placed a level stake on it to win. Flat stakes throughout, the same unit every race, because progressive staking like doubling up only enlarges a negative edge, it never fixes one. We settled every bet at industry Starting Price, the official odds returned as the race goes off, with no commission and no early-price shopping. That is deliberately the kindest possible measure. A real punter taking real prices and paying exchange commission would see the cost run higher than this.

The rule that matters most is how we treated the horses that did not finish. Non-runners are voided, but fallers and pulled-up runners are counted as the losing bets they are. Your stake is gone whether the horse falls at the second-last or trails in last, so it belongs in the losing column, and a race only counts at all when it has a recorded winner. Counting the fallers honestly is what gives the outsider's loss its true figure rather than a flattering one.

Where two horses were joint-longest in the market we split the result so no phantom winner could leak in and lift the number. No filtering, no hand-picking favourable meetings, no quietly dropping the races that went badly. Every race in the sample, settled the same way, by the same rule. The point of the Lab is to publish the honest figure whether it flatters the system or not, so the method has to be the same for the systems that lose least and the systems that lose most. This one loses most.

Here is the figure with nothing dressed up. Backing the outsider in every race returned -34.70% to Starting Price across 27,676 real British races. For every £100 you staked you got back about £65.30. The 95% confidence range is [-40.9,-27.2], so even on the kindest reading of the variance it is a heavy loss, and on the harshest it is brutal. There is no version of this number that is close to break-even.

The strike rate tells you why. The outsider wins 3% of the time, roughly one race in every thirty or so. You sit through long, barren runs of losers waiting for a big price to land, and when it does the payout still does not cover the run that funded it. The leak is −34.7p of every £1 staked, with the wild swings hiding the steady cost underneath.

The comparison is where it really bites. Backing the favourite blind returns -8.79% to SP. A genuinely random horse, a pin in the paper with no judgement at all, returns -21.53%. Backing the outsider returns -34.70%, several times the cost of the favourite and a clear chunk worse than throwing darts. Read that twice. Deliberately picking the horse the market rates least likely is worse than picking with no information whatsoever, because a random pick at least lands on a favourite sometimes, while the outsider sits permanently at the most over-padded end of the book.

That is the whole result in one line. The favourite-longshot bias is not folklore, it is a measured drag, and at -34.7% the outsider is the most expensive single-pick line on the entire page. Past performance is no guide to the future, this is to SP with no commission, so real life is worse again.

So, does backing the outsider work? No, and it is worth knowing exactly why, because this is the single sharpest illustration in the Lab of where the bookmaker's margin really sits. It returned -34.70% to SP over 27,676 real British races, with the fallers and pulled-up horses honestly counted as the losing bets they are, the heaviest cost of any single-pick line we have measured.

It loses for the most basic reason there is. You are backing a horse that wins 3% of the time, at the most over-padded end of the book, and paying the bookmaker's margin on every single bet. The favourite-longshot bias sets the long-term cost, and the occasional fairytale winner only delays the result rather than reversing it. The dream is real, the price proves the maths is brutal, and the maths wins.

The honest small print does not rescue it. The damage per race is slightly less violent in small fields, where the longest price might be 8/1 rather than 50/1, simply because you are no longer betting at the very worst end of the curve. That is harm reduction, not a profit. There is no field size, no code and no condition where this turns positive.

The lesson is the one the whole Lab keeps proving from the other direction. If a system's entire logic is back the horse the market likes least, the market is usually right and the price already tells you so. The crowd is not leaving value at the long end of the book. The bookies have buried their fattest margin there precisely because that is where the dreamers look.

If you want to size bets sensibly, the honest tools are the dull ones: a flat stake you can afford to lose, and a hard look at whether you hold any edge at all. We publish ours either way, losses included, in the track record. Treat the outsider as a clear example of how a plausible-sounding rule loses heavily, never as anything to stake real cash on.