Before we go any further, let me ask you a question...
Robbed of your hard-earned cash on horses that looked certain to win...
Only to watch them fold in the final furlong and take a chunk of your betting bank with them?
Have you ever had that sickening moment where your horse hits the front, the race looks won, and something you've never heard of comes flying past a stride from the line?
You tell yourself it was rotten luck, and now and then that's the truth of it.
But a quiet voice keeps asking why it always seems to land on you, and never on them.
Do you ever feel like the bookies are ten moves ahead of you...
Playing chess on a board you can't even see, while you're still learning how the pieces move...
Then I've some news that might sting a little before it starts to feel a whole lot better.
That's how I felt, sat in the same worn-out armchair night after night for the best part of 20 years.
Hunched over the form pages with a cheap pen and a pad of paper, trying to out-think a building full of data scientists.
I'd have the radio on low and three betting apps open at once, flicking between them to squeeze an extra quarter of a point out of a price, convinced that was the clever bit.
It was never going to work.
There was no chance of ever making serious money doing it that way.
They had me right where they wanted me...
Just happy enough to keep betting, spending my wages on their sites and in their shops...
But never so fed up that I'd walk away for good.
Looking back, the worst part wasn't the money.
It was how much room the betting took up in my head.
I'd be sat in a meeting at work, nodding along, while half my brain was working out whether a horse at Wolverhampton would handle the drop back in trip.
Until one grey Tuesday morning I woke up and decided I'd finally had enough of handing it all over.
It wasn't a big dramatic moment.
I was stood at the bathroom mirror with a toothbrush in my mouth, and I did a sum I'd been dodging for years: roughly how much I'd handed the bookies since my 30th birthday.
I'm not going to print the figure here, but it had five digits and it didn't start with a one.
My job is fixing systems that don't work, and it hit me that I'd been running a broken one at home for two decades without once asking why it kept failing.
If the bookies knew something I didn't, then the people who worked for them knew it too.
Thanks to my job, I knew where to find them, because I'd spent years hiring contractors through LinkedIn and I knew people list their employer on their profile as proudly as a medal.
I started messaging anyone I could find on LinkedIn who worked in trading, pricing or risk for a UK bookmaker.
My proposal was simple.
I'd pay them £500 if, and only if, they agreed to get on a video call with me for an hour and tell me what really went on inside their firm.
It was a decent sum for an hour of someone's time, and I meant every penny of it.
I wrote each message by hand, mentioned something from the person's own profile, and kept it short enough to read on a phone in a lift.
As you'd imagine, almost everyone ignored me completely.
A handful told me where to go in no uncertain terms, and one man reported the message and blocked me inside a minute.
Another replied with a link to his firm's safer gambling page, which stung more than I let on.
LinkedIn eventually flagged my account for sending too many messages.
Three weeks in, with nothing to show for it but a warning and a bruised ego, I'd all but talked myself out of the idea.
When I woke up one morning to a new message sitting in my inbox.
It was from a man I'll call Ray.
I've changed his name to protect his identity, for reasons that'll become obvious.
He worked for one of the biggest bookmakers in the UK.
He sat in a small team that tested the pricing software before every update went live, checking each new version against the old one race by race.
But that meant he knew precisely what went on inside the engine room of that building.
His department had just been told to expect redundancies, and his mortgage deal was about to run out, so £500 landed at the right moment.
His first message was three short lines, and the last of them read, "I probably shouldn't, but I need the money, so let's talk."
We jumped on a call that morning and I confirmed the deal was real.
He kept his camera off for the first ten minutes, and twice I heard him get up to check the door was shut.
I sent half the money to his PayPal on the spot, with the rest to follow at the end of the hour.
What he told me answered every question I'd ever had about why I could never get ahead.
He described how the bookmakers had built and trained an AI system that weighed up every variable imaginable.
He told me their pricing hadn't been the work of a clever man in a back office for well over a decade.
A pricing engine built by a team of data scientists, the same sort of people I'd spent my whole career working alongside in IT.
Machine learning software that repriced millions of markets a day...
Spitting out odds that looked generous to the punter on the surface...
While methodically disguising which horses were most likely to win.
He put it in a way I've never forgotten.
The price on your screen isn't an offer, he said, it's a trap with a friendly face painted on it.
It's a number that's been calculated, tested against millions of past results, and tuned to take a slice off you on every bet you place.
He explained that the engine doesn't only ask how likely a horse is to win.
It also asks how much the firm stands to lose if it does, and how many punters will back it whatever the price.
A horse ridden by a famous jockey for a famous yard will be backed by thousands of people who've never studied its form, so the engine can offer it at a stingier price and still take the money.
The flip side is the horse nobody's talking about, which can sit at a bigger price than it deserves because the engine has no reason to tighten it.
That was the moment on the call when I stopped taking notes and just listened.
The engine also watches how money moves through the market and shifts the odds in real time to protect the firm's margin.
The second a horse is backed by punters with a winning history, the price tightens before the rest of us have even noticed.
It was a finely tuned piece of software built to do one thing.
Take your money, and keep you coming back to hand over more
I asked him how many people in the building could explain the whole engine from end to end.
He laughed and said maybe four, and two of those had left.
He said the firm had a whole floor of clever people whose only job was to keep those odds tight and the punter's life hard.
Mathematicians and statisticians, all of them pointed at the same target.
Grown adults with letters after their names, coming in every day to make sure people like me never got a look in.
And every bet ever struck with the firm, win or lose, was just more fuel for the engine.
The more the country played, the better it learned to beat us.
The same engine keeps watch over every account on the books.
The moment you start winning at decent prices, a flag goes up against your name.
Your maximum stake gets cut, your account gets throttled, and before long you can't get more than a few quid on.
They don't ban you for losing.
Let that sink in for a second, because it changes how you see the whole game.
The one outcome they're built to punish is the exact one you've been chasing all these years.
Ray had watched accounts restricted after as few as a dozen bets, purely because the engine didn't like who they were backing and when.
He told me that for years he'd watched this software outperform every analyst in the building.
The AI age of betting had well and truly arrived, and it was cleaning up behind the scenes.
It had become a force of its own, banking record profits for the firm year after year.
He said that if I really wanted to compete, I needed access to the same kind of technology.
But he'd spent months of evenings rebuilding the core logic on his own private server.
He knew how it worked from the inside out, because testing it had been his job.
And what he'd built, he told me, was the closest thing to it that existed outside their walls.
Not a line of it had been copied, and he was firm about that, because every part of it had been written from scratch using what he'd learned watching the real thing at work.
He'd never once used it to bet, he said, because the moment he did he'd be breaking his contract and risking his job.
It had been sitting there for months, working perfectly, earning him precisely nothing.
He didn't want a salary, and he didn't want a one-off fee.
He wanted a share of whatever we made together.
A straight 50/50 split on any profits
It sounded fair to me.
He couldn't bet himself under his employment contract, and every family member's account had been closed when he took the job.
I could bet freely across dozens of accounts, and I had the technical background to keep his creation running smoothly
He brought the software, I brought the ability to use it, and neither of us was any use to the plan without the other.
We shook hands on it over video, and he logged off.
I sat there for a good while afterwards, staring at a blank screen and wondering whether I'd just been had.
That evening, the access details arrived from an anonymous email account.
I logged in, and what I saw stopped me in my tracks.
Selections were already flagged for the next morning's races.
There was a clean dashboard and a list of horses, with the model's own price for each one laid next to the best bookmaker price.
As an IT man, I knew straight away this was nothing like a homemade formula with a few sums bolted together.
This was a fully trained model, the sort of thing that takes serious money and serious know-how to build.
I've stood over enough half-baked systems in my career to know the difference between something knocked together in a weekend and something built with real care.
There were logs for every decision, timestamps on every data pull, and version notes going back months, which is how you can tell an engineer built something rather than a hobbyist.
I got three bets on before the odds shifted.
I woke up the next day with a churning mix of nerves and excitement.
The first pick lost and my heart sank.
Here we go again, I thought, another clever idea that falls apart the second real money is on it.
But I made myself see the day through.
The second race came, and that horse got up by a neck.
I finished the day up over £275.
The next day brought two winners from three for a profit of £315.
The first full week closed out with six winners from eleven bets.
That gave me a first-week profit of £1,150
I messaged Ray and told him the results.
"That's about what I expected," he wrote back, "give it a few months and you'll stop being shocked by it."
I sent his share over and he was delighted.
I didn't tell Mary for the first two months, which I'm not proud of.
After everything I'd put her through, I wanted the numbers to do the talking before I did.
I want to walk you through what happened next, because the numbers on their own don't tell you what it felt like.
That first week was in September 2025, and I ended my first full month with £3,840.50 in profit.
More than I'd ever made from betting in a single month.
By some considerable distance
Truth be told, I spent half of it braced for it all to collapse.
October 2025 added £4,420.30, and I started to relax into it.
Then November became my best month of that first winter.
I withdrew every penny and watched it land in my current account.
I still remember refreshing my banking app like a kid, half-expecting the money to vanish, and it just sat there, every penny of it.
That was the evening I finally sat Mary down, showed her the app, and told her the story from the first LinkedIn message onwards.
She went through it twice, handed the phone back, and asked how much I'd paid this Ray for his hour.
December added £6,380.60
For the first time in years, Christmas was paid for before the first card went up, with nothing left lurking on a statement for January.
January 2026 brought another £6,740.45, and the new year started the way the old one had finished.
Two cards I'd been shuffling money between for the best part of a decade, gone in a single afternoon.
February cleared £5,980.30
March was quieter, but still good for £4,290.85.
The cards were thinner as the jumps season wound down and the turf Flat hadn't found its feet, and the model sent fewer bets because fewer races met its standards.
I booked a 10-day holiday to Italy that Mary had been putting off for three years.
We'd never managed to go before.
She stopped rolling her eyes at me around this point.
Now, this is where it gets interesting, because everything up to that point was the software finding its feet.
By this stage the model had seven months of live racing behind it on top of all its history, and you could feel the difference.
April 2026 produced £6,540.90.
Each week's results were being checked against what the model had predicted, and you could see its judgement tightening month on month.
The picks came through with more conviction, the near-misses got rarer, and the quiet no-bet days started to feel like the software protecting me rather than letting me down.
May became my best single month to that point.
That was a month where I did nothing more strenuous than open an email and tap a few bets on before my second coffee.
June 2026 came in at £6,840.75.
That was the month I used to dread, the one that had cleaned me out more times than I care to count, and this year I watched it hand me money instead.
We spent most of that summer in the garden with the boys home from school, and for once I wasn't slipping indoors to check a result on my phone.
July held strong at £6,620.10
The ground baked hard through most of that month, and the model adjusted to firm-ground form without me lifting a finger, which is more than I ever managed in 20 years of trying.
Seven profitable months into 2026 and not a single wobble in sight, which after twenty years of systems falling apart on me still felt faintly unreal.
£6,998.60 in a single month, a fresh record, and the clearest sign yet that this thing keeps getting better the longer it runs.
That was the month we took the boys to Cornwall for a week, paid for out of the betting account, which is a sentence I never thought I'd write.
September 2026 brought £6,180.75 as the ground softened and the autumn cards filled out.
It was a touch below August, and I'd be lying if I said I didn't notice, but it was still more than £2,300 ahead of September the year before.
And October so far has added £5,640.50, with a week of racing still to go as I write this.
That's already well clear of the £4,420.30 I made in the whole of October last year, and the jumps season has barely got going.
The switch from the Flat to the jumps used to be the point in the calendar where every system I tried would come apart, because the form book changes so much from one month to the next.
This year the model went through it without a hiccup, because it had been through the same switch once already and it had learned from it.
That's the pattern I keep coming back to when people ask me whether the good run is about to end.
A lucky streak fades and a broken system falls off a cliff, but this has done neither in 14 months.
What this does instead is grind out a little more each season than the one before, because it finishes every year knowing more than it started it with.
That isn't a run of good luck, because luck doesn't improve month after month for over a year.
That's a machine learning its trade and getting better at it in the background while I get on with my life.
All withdrawn, all tax-free, from a few bets a day
Not one losing month in the entire run
Ray has received his share throughout.
He kept his job when the redundancies came, and we've never once fallen out.
The arrangement works, and we've no intention of changing it.
Two years ago I was hiding bank statements from Mary.
Now she asks me over breakfast which horses the software has flagged.
Bond Street Bets has gone from a dirty little secret to something the family is actually proud of.
No more staring at an empty bank account in the final days before payday.
Imagine the bills covered and the credit cards cleared for good.
Holidays whenever you choose, wherever you want to go.
A newer car sat on the drive, maybe that extension you keep talking about.
All of it within reach, and more besides.
That's what this has done for my life.
And it's what I want it to do for yours.
Let me tell you a little more about myself before we go any further.
My name is Will Chapman.
I'm 53 years old, married with three teenage boys, and I've lived in London my whole life.
I work in IT infrastructure
I've done it for 25 years, most of them for a large insurance company you'd probably recognise.
My team keeps several hundred servers running round the clock, and when something breaks in the night, it's usually my phone that rings.
I spend my days thinking about systems, logic, and how to make a clunky process run smoother.
If a server falls over at three in the morning, I'm the one tracing it back to the one line that broke.
Spotting faults and working back to their cause is the job, and I'm good at it
I mention all this because it's the reason this story has a happy ending.
The skills that make me useful at work are the very ones I never brought home to my betting.
For the better part of two decades, I left all of that at the office door.
A tip from a mate one week, a name that tickled me the next.
I'd spend the week troubleshooting million-pound systems for a living, then throw fifty quid at a grey because I fancied the look of it in the paddock.
The man who wouldn't dream of deploying a system without testing it first would happily bet a week's wages on a gut feeling and a good name.
My wife Mary thought it was harmless enough in the early years.
A tenner here and twenty quid there didn't worry anyone
It only became a problem when I started taking it seriously.
I followed tipsters, bought systems, and built my own rating sheets late into the night.
I had folders of them, colour-coded, full of formulas I was convinced held the answer.
Nothing ever worked for long enough to matter
A system would fly for a fortnight, suck me in, then fall apart without warning and take my bank with it.
There were months where I spent more online with the bookies than I did running the car.
One stands out, a Saturday night after a big festival had wiped out my account, both of us sat at the dining table not saying much at all.
She didn't shout, which was worse, she just asked me, very softly, whether I even enjoyed it any more.
I couldn't give her a straight answer, and that scared me more than the money ever had.
Promises made and broken, more times than I want to remember
I knew deep down I'd never quit entirely.
My grandad took me to my first meeting at Brighton racecourse when I was nine, and I can still smell the wet grass and the hot dogs.
There's something about a big field thundering towards the stands that has never stopped making the hairs on my arms stand up.
Walking away from racing was never going to happen, so the only real option was to stop losing at it.
But something had to change
I couldn't keep bleeding money into an industry that was never going to let me win.
That's what led me to the bathroom mirror, and to the LinkedIn plan.
I want to spend a moment on this, because the software itself is the whole heart of the story.
When Ray first described what he'd done, I half expected to be underwhelmed once I actually got my hands on it.
It didn't underwhelm me for a second
He'd started with the one thing money can't easily buy, which is real knowledge of how the firm's own engine was put together.
He knew which factors it leaned on, how it weighed public money, and where it deliberately cut corners on the smaller races nobody was watching.
Then he spent the best part of a year, evenings and weekends, rebuilding that logic from scratch on a server in his garage.
He fed it years of historical racing, every result he could lay his hands on, and let it learn the patterns for itself.
By the time he'd finished, it had studied more than 15 years of British racing, hundreds of thousands of races, with every runner's details stored alongside the result.
It would flag horses that looked wrong even to him, and he'd have to work out whether the software had spotted something he'd missed or just made a mistake.
Most of the time, unnervingly, the machine was right.
By the time it reached me it was already impressive, but this is where my side of things came in.
Twenty-five years in IT infrastructure meant I knew how to feed a hungry system clean data and keep it running without falling over.
A model is only ever as good as what you pour into it, and one bad feed can poison everything downstream without anyone noticing.
We tightened up where the racing data came from, cross-checked it against a second source, and built in alerts so that if anything looked off first thing in the morning, we'd know before a single bet went on.
A jockey change that hadn't filtered through, or a going report that was hours out of date, now gets caught before the model ever sees it.
A brilliant selection is worthless if the price has already gone by the time you back it.
So we tuned the system to run early and fast, producing its picks while the overnight prices were still soft and the big professional money hadn't yet moved in.
Then I brought in a habit from my day job that Ray had never thought of.
In my world, you never put a new version of anything live until it's run alongside the old one and proved itself.
So every change Ray makes to the model now runs in shadow for a full month, pricing the same races as the live version without a penny staked on its picks.
Only when the new version has beaten the old one on real races does it take over the morning email, and more than a few of Ray's bright ideas have failed that test.
He grumbled about it at first, but he's since admitted it saved us from at least two changes that would have cost real money.
Every evening, once the day's results are in, it looks back over everything it priced and asks itself a blunt question.
When I said a horse had a one-in-four chance, did horses like that win roughly one time in four?
If its 25% horses have been winning less often than that, it's been too generous, and it reins itself in.
If they've been winning more often, it's been too timid, and it loosens up.
It then adjusts the weight it gives each factor, and starts the next morning a fraction better calibrated than the day before.
We set that check to run automatically, so the model keeps itself in line while the pair of us are asleep.
One of the biggest leaps came when we taught it to read the early market as one more piece of evidence, never the deciding vote.
If a horse the model rated highly was also being backed by professional money first thing, that agreement counted for something.
Any type of race where its own record had been poor went on that list, two-year-old maidens on the all-weather in midwinter being one example.
It wasn't allowed to bet in those races again until its shadow prices proved it had got the hang of them.
A machine that knows what it doesn't know is worth ten that pretend to have an answer for everything.
Over 14 months, those small daily adjustments have added up to something you can see in the results.
The 2026 months aren't ahead of the 2025 ones by accident
They're ahead because the software has had more data to learn from, and it never once forgets a lesson.
People assume there's somebody sat up at dawn picking horses, so let me tell you what really happens each morning.
At around half past five, the system pulls in the day's declarations and going reports, along with early prices from across the market.
It checks each of those feeds against a second source, and if two sources disagree about something as basic as a jockey or a weight, it stops and sends me an alert.
By about six, the model has priced every runner in every race, which on a busy Saturday can mean more than 400 horses.
It then strips the bookmaker margin out of each market, lines its own chances up against the best available odds, and marks any horse where the difference clears its threshold.
Those candidates go through one last filter, which throws out anything on the probation list and anything where the market is too thin to get a fair bet on.
Around seven I get a short summary on my phone, I read through it, and I place my own bets.
The whole run takes the machine about 40 minutes, and it has never once needed me to make a judgement call.
My only job is to check nothing looks broken, which after 25 years of watching servers is the easiest part of my day.
Here's the thing most punters never grasp.
Every price you've ever taken was set by software before any human trader so much as glanced at it.
That software has already done more sums on the race than you could manage in a lifetime.
Because Ray rebuilt the core of their engine, we aren't guessing at how the bookies think.
We already know the blueprint, because Ray spent years testing the original
Every morning, before the first race is anywhere near off, the system pulls in the full card for every meeting in the country.
Then it goes to work on every runner in every race, not just the ones a tipster happens to fancy.
It weighs up more evidence at once than any punter could juggle in a whole morning.
How each horse actually finishes its races, the late pace it shows rather than where it crossed the line.
How the going suits its breeding, right down to the sires that throw runners who handle soft ground.
Course and distance form, and the shape of the track itself, from tight bends to stiff uphill finishes.
Where a horse breaks from the stalls, which only matters in big fields at certain courses on certain ground.
A trainer's form over rolling windows, plus the less obvious angles, like a first run back after a long break or headgear going on for the first time.
Whether a horse is dropping in class, and whether the race it's dropping from was stronger than the bare result suggests.
Whether a jockey actually goes well round that particular track, not just whether their name is famous.
And the weight a horse carries against the direction its handicap mark is travelling
In total it reads more than 140 separate inputs for each runner, many of them built by the model itself rather than chosen by us.
A good human tipster might weigh up 20 or 30 of those factors for one or two races, and it would take them all morning.
The software does it for every runner at every meeting, in seconds, with the same attention paid to the last race of the day as the first.
But raw speed isn't where the advantage lives
It works out a true win chance for every horse, then holds that figure up against the bookmaker's price.
First it has to strip out the bookmaker's built-in margin, because if you add up the chances implied by every price in a race, they come to well over 100%.
In a typical handicap the total might come to 118% or more, and that extra slice is the firm's profit before a single horse has run.
The model removes that slice, so it's comparing like with like, then lays its own chances alongside.
When the software reckons a horse has a 25% chance of winning and the bookies are pricing it as if it's nearer 12%, that difference is the bet.
It won't flag anything where the difference is marginal, because a slim advantage can vanish the moment the price moves a point.
It only has to win often enough, over a long run of those mispriced runners, to leave you well in front.
That's the exact logic the bookies use to build in their margin.
We've just pointed it back in the other direction.
The software doesn't treat every factor the same in every race, because it learned that the rules change depending on the contest in front of it.
At a tight, turning track in a big sprint field, where a horse breaks from can make or wreck its chance, so stall position gets heavy weight.
Over a longer trip at a wide, galloping course, the stalls barely matter, so the model all but ignores them.
In a race full of unexposed young horses, the form of the yard sending them out tells you a lot.
In a big handicap stuffed with exposed old campaigners, it tells you almost nothing.
The software knows all of this, not because anyone typed in the rules, but because it taught itself from years of past results.
That's the difference between a rigid set of if-this-then-that instructions and a model that understands what it's looking at.
The bookies' engine is built to price thousands of races a day and turn a profit across millions of bets.
It isn't designed to get every horse at a wet Tuesday meeting at a small track spot on.
It leans on public money and big-name jockeys to shape the smaller markets, and that's where the wrong prices creep in.
Those openings are small, and they close fast as the money comes in...
Our software is built to find them before anyone else does, and to get on while the price is still there.
Every result, winner or loser, goes into the evening check I described earlier.
The model looks at what it got wrong, adjusts which factors it leaned on too hard, and comes back a fraction better tuned the next morning.
The bookies' engine improves every month, and so does ours.
Picture two horses in different races on the same autumn afternoon, both priced at 7/1 first thing.
The first has been running well on fast ground, its trainer is in decent form, and a well-known jockey has been booked, so it's attracting a stream of small bets.
The model looks at it and finds nothing the price doesn't already account for, so it leaves it alone.
The second has trailed home in its last four runs, which is why nobody wants to know.
But the model clocks that those defeats all came on quick ground the horse can't cope with, that today's rain-softened surface is what its breeding is crying out for, and that the yard has hit form in the last couple of weeks.
It works the horse's real chance out at nearer 3/1 and flags it as value.
Same price on the screen, two completely different bets underneath it, and only one of them is worth your money.
When a horse like that is backed in to 4/1 by the off, you've had it at 7/1 first thing while the rest of the market was still rubbing its eyes.
No crystal ball and no inside word on the day, just a machine spotting a wrong price quicker than anyone else can.
A man with a newspaper and a hunch is bringing a torch to a floodlit stadium.
He's working off the same public form everyone can see, while the firm he's betting into runs software that reprices the race every few seconds.
For the first time, you'd be sitting on the right side of that fight.
I spent 25 years in IT learning one hard lesson.
People are brilliant and people are hopeless, often within the same hour.
We spot patterns that were never there, and we let one rotten morning sour every decision that follows.
It doesn't get bored by the seventh race of the day.
It doesn't chase a loss, and it doesn't fall in love with a horse because it once won at a big price.
It runs the same careful process on every runner at every meeting, morning after morning.
It has no favourite trainer and no urge to win its money back before lunch.
For years I tried to be that machine with a pen and a racecard, and I failed, because no person ever could.
Now I don't have to be, and that's the most freeing thing that's ever happened to my betting.
Let me put your mind at rest about something first.
If you've never turned a regular profit backing horses, it isn't because you're daft, and it isn't because you haven't tried hard enough.
It's because the deck was stacked before you ever opened the betting app.
You've been pricing races in your head against software that prices them with an entire department behind it.
You bring a gut feeling and a tip from someone at work.
They bring a model refined across hundreds of millions of bets that updates the second money starts to move.
That isn't a fair fight, and it was never meant to be one.
Back three winners on the spin and what happens?
You get bold and start staking more in races you'd normally leave well alone.
The firms have studied that exact reflex for decades, because a hot streak is usually what tips a punter into giving it all back, and then some.
You back horses you like instead of prices that are wrong.
There's a world of difference between a horse you think will win and a horse whose odds are too big for its real chance.
Almost nobody asks that second question, and it's the one the bookies hope you'll never ask.
Find a bit of an advantage and your stake gets cut, or your account gets limited to coppers.
Spreading your bets across a few firms and keeping stakes sensible buys you time, and time is what lets a winning approach pay out.
Ask most punters if they're up or down over the last year and they truly don't know.
They remember the big winner and forget the slow drip of losers.
You can't fix a system you've never once written down.
I learned that the hard way at the bathroom mirror, and it's the first rule I'd apply to any server in my care.
None of that is your fault
But all of it can be turned around the moment you stop guessing and hand the arithmetic to a machine built for it.
I've paid for more tipster services than I'd care to add up, and the vast majority failed me in the same handful of ways.
Once you've seen how the firms really operate, you understand why.
What they have is a knack for marketing the winners and going very quiet after a bad week.
There's no model underneath it, nothing that holds up when you line up three months of results side by side, just opinion dressed up as expertise.
Their records have a funny habit of starting the week after their last bad run.
You know the routine
A big-priced winner gets framed and hung on the wall, while the 30 losers either side of it never make it into the picture.
That's advertising, not a record, and it costs you every time you fall for it.
A tipster on a good run feels bold and decisive.
The same tipster after a rough fortnight starts tinkering with his own rules and ducking bets he'd have snapped up a month earlier.
Our model has no moods to wobble, and it runs the same process every morning whatever happened the day before.
If someone won't put their own money down on a selection before it reaches you, that tells you everything about how much faith they have in it.
I'm on every bet myself, before a single member sees it.
When a tip loses, they forget it and move on.
Nothing gets learned from it at all
Our model does the opposite, checking every result against what it predicted so the next morning's picks are a touch better judged.
A human tipster can't run that check on himself, however many years he's been at it.
That's why so many services vanish within a year and pop up again under a shiny new name with a fresh set of early results.
A machine doesn't need to do any of that, because it doesn't run out of road.
I'm not going to bury you in tables and charts, because you've seen a hundred of those and half of them are made up.
So let me just give you the shape of it, straight from my own betting records.
14 months of live betting, at level stakes, with nothing deleted on the bad days.
A total of £84,375.20 taken out and banked, every penny of it withdrawn and tax-free
Not one losing month across the whole run, not even December when half the country was off work and the cards had thinned right out.
September and October have both comfortably beaten the same months last year, and October still has a week left to run.
Every month of 2026 so far has cleared £4,000, and eight of the ten have cleared £5,900, with October still going.
That isn't a projection or a best-case guess dressed up to look impressive.
It's what happens when you let a real advantage run for long enough to show its true colours.
Back in the spring, I let a handful of people I trust follow the bets for a month and asked them to tell me the truth about it, warts and all.
Here are four of the replies that came back.
★★★★★
"Do you know what it feels like to open your betting account and find more in it than you put there? I didn't until the third week, and by the end of the month mine was sitting £1,940 higher than where I started."
Stuart Fenwick, Hexham
★★★★★
"31 selections across 18 racing days, and a balance £2,215 higher than when I began. I was a quantity surveyor for 34 years, so I checked every one of those figures twice before I let myself believe them."
Raymond Holt, Stamford
★★★★★
"I put my bets on at the bus stop on the way to work and don't give them another thought until I'm home. That little routine has added £1,680 to my account this month, and it's never once made me late."
Joanne Pryce, Hereford
★★★★★
"My first month's winnings went straight on four new tyres and a full service for the car, with £600 still left over afterwards. It's the first time any betting system has paid for something I can actually walk outside and touch."
Lee Garside, Wakefield
Now, you might be wondering why I'm sharing this at all.
If the software works this well, why let anyone else near it?
It's a fair question, and it deserves a full answer.
Ray can't bet under his own name, or through anyone he's connected to.
He's watched the results build for well over a year, and he wants more people winning off the back of this.
As for my side of things, I've taken as much out of my own accounts as I can manage without drawing attention.
Three of my accounts have already been restricted to stakes so small they're barely worth logging into.
Remember what I told you about that flag going up the moment you start winning?
Every account I lose means a morning's work from the model that earns nothing, and that's a waste neither Ray nor I can bear.
Members betting through their own accounts means the selections keep earning, even as mine get squeezed.
On top of that, the software doesn't care whether it's working out bets for one account or fifty.
It does the exact same job every morning either way
The membership fees also cover the racing data feeds and the server costs, which aren't cheap, so the model pays its own way.
So letting a small group of the right people in costs me nothing, and it keeps the whole operation running well into the future.
It's in both our interests to grow this slowly, and with the right people.
In September I ran a wider trial with a fresh group of people.
Friends of friends and a couple of online contacts, with no insiders among them.
I asked them to follow my daily bets for 30 days and report back at the end, good or bad.
I just wanted to know what happened.
Here's a selection of what came back.
★★★★★
"Will, I owe you a pint or three, because the £2,385 I made this month is more than my last three years of betting put together. Please don't give my place to anyone who isn't going to stick with it."
Alison Wardle, Stockport
★★★★★
"I managed a high street betting shop for 11 years, so I know which customers the firms are pleased to see walk through the door. A month on these selections made me £2,140 and turned me into the sort of customer my old area manager used to warn us about."
Russell Ingham, Preston
★★★★★
"Saturdays used to mean pacing the living room and shouting at the telly until my throat hurt. Now I get the bets on early, go fishing with my brother, and came home this month £1,975 better off without watching a single race."
Glen Hartwell, Kendal
★★★★★
"I once backed a horse because it shared a name with my ex-wife, so believe me when I say I'm no judge of form. Not having to judge anything is the reason I'm £2,290 up, and the reason I've stopped explaining my bets to anyone."
Barry Gill, Scarborough
Results like those are what happens when a well-built model keeps finding prices that are wrong.
I've decided to open up Bond Street Bets to a wider group.
No more than that.
Ray was firm about the number, and I agreed with him.
He wants to know who's on the list, and so do I.
Plus, I've only so much time in the day to handle correspondence and support.
I answer every email myself, and 100 members is about the limit of what one man with a day job can reply to well.
Keeping it at 100 means every member gets a real reply from me, usually the same day.
But they are going quickly
I don't expect all 100 to last the day.
Every morning, before 8am UK time, an email lands in your inbox.
Inside, you'll find the exact selections the software has flagged for that day's racing.
I'll have placed my own bets moments before sending it.
Which means you get on at the same odds I do
You open the email, head online or down to your local shop, and get your bets on.
Then you close the email and get on with your day as normal.
No need to watch the races unless you fancy it.
There's no form to read and no decision to second-guess.
If you can open an email and place a bet, you can do this
All the hard thinking has been done for you hours before you wake up.
In the evening, you check your account to see the results.
You don't need to know a thing about racing, form, or betting markets.
You don't need any technical ability whatsoever.
It works whether you're retired, studying, working shifts or raising a young family.
It's open to anyone over 18 who can legally bet on UK racing.
There's no learning curve and no catch
You open the email and place the bets, day after day.
I'm not going to insult you with promises this can't keep.
'Bond Street Bets' won't make you wealthy
It won't win every bet, and there'll be the odd losing day along the way, because no system on this earth wins them all.
What it will do is put a pricing model on your side that reads races the way the bookies do, then backs their mistakes morning after morning.
Over a month, and over a year, that's an advantage most punters never get within a mile of.
The 14-month record I've shared is what that advantage looks like once you let it run.
Let's talk about the price.
I'll lay it out plainly.
14 months of results totalling £84,375.20 isn't something you can stick a modest price on.
A professional tipster with half this track record would charge £100 a month without blinking.
But I've no interest in running a high-priced subscription business.
My bets are worked out and placed every morning regardless of how many people receive them.
It costs me next to nothing to send them out to members.
And after years of losing my own money chasing bad advice...
There's real satisfaction in watching other people finally get ahead.
To become a lifetime member of Bond Street Bets right now...
The one-time joining fee is £20
Not £20 a month.
Not £20 to start and a subscription kicking in later.
You pay once, and you receive my daily bets for life.
There are no hidden charges and no renewal fees.
That's less than most people lose on a single bad Saturday afternoon.
And to make this as easy a decision as it should be...
Bond Street Bets is fully backed by my 30-day money-back guarantee.
Join today and follow my selections for the next 30 days, tracking every result as you go.
If at any point inside those 30 days you're not completely happy with what you see...
Drop me a quick email and I'll refund every penny of your joining fee.
Your joining fee comes straight back to you, the same day.
Every member of my trial group has stayed on.
Not one of them has asked for a refund.
But the guarantee is there, and it's iron-clad.
Because that's how confident I am in what you're about to receive.
There are 100 memberships in total.
87 have been taken so far, as I write this.
That leaves 13 spots
I don't know how long they'll last.
If you're reading this and the page is still up, there's still a chance.
To get lifetime access to Bond Street Bets, click the button below.
Complete the 100% secure checkout in under a minute.
The members area opens the moment your payment goes through.
Once all 100 memberships are claimed, this page comes down, and I can't promise when or if I'll open the doors again.
You're fully protected by a 30-day money-back guarantee, so the only real risk is doing nothing and letting someone else take your place.
Tomorrow morning's email could be the first of hundreds landing in your inbox.
£20 is a small price for a seat on the right side of the counter.
I look forward to welcoming you aboard.
I didn't want to write the usual list of questions and answers, because you've read a hundred of them and they all sound alike.
So I gathered up the questions my trial group fired at me, got Ray on a video call one Sunday evening, and read them out to him one by one.
Where the answer was mine to give rather than his, I've added it underneath.
Ray: "You should think that, and I'd worry about anyone who didn't."
Ray: "There's no magic in it, just a model doing sums faster than any person can, and 14 months of live results showing it does them well."
Will: That's also why the guarantee exists, so you can prove it to yourself on my money rather than take my word for it.
Ray: "Not as far as I know, and I've no intention of telling them."
Ray: "What I've built is mine, written from scratch at home, and it doesn't touch a line of their code or a byte of their data."
Ray: "They update it all the time, and every update I tested in my years there was aimed at the big markets where the serious money goes."
Ray: "The smaller races get the same treatment year after year, because fixing them would cost more in staff time than they'd ever win back."
Will: And if anything did shift, the evening check would pick it up within days, because the model would see its own predictions slipping off course.
Will: Good, because that turns out to be an advantage rather than a problem.
Will: The model does the thinking, and what you know or don't know about horses makes no difference to the result.
Will: If you can type a horse's name into a search box, you're already overqualified.
Ray: "Because the price is what we're judging, not whether the horse is the most likely winner today."
Ray: "A horse with an ugly recent record at a big price can be a far better bet than a favourite everyone loves at a short one, and the model knows the difference."
Will: Usually two or three a day, sometimes a single standout, and now and then none at all.
Will: Days without a bet are some of the most valuable, because a forced bet is a losing bet waiting to happen.
Will: A tipster who serves up ten bets a day is feeding his ego, not your bank.
Will: Start smaller than you think you should, and let the results give you the confidence to build.
Will: My own record runs at level stakes, and a small fixed slice of your bank on each bet lets the model do its work without any single day stinging.
Ray: "Then it means you're winning, and the firm's software has noticed before you have."
Will: Use a few different firms and keep your stakes sensible, and you'll keep your accounts open far longer than the average winning punter does.
Will: Neither can I, and neither can most of the people already doing this.
Will: The bets come with their race times, so you get them on first thing and forget about them until you fancy a look at your balance in the evening.
Will: And if you miss a morning, nothing is carried over, so you just pick it back up the next day.
Ray: "Most of those are a basic stats filter with a clever name bolted on the front."
Will: This is a rebuild of the core logic of a real bookmaker's pricing engine, by the man who spent years testing the original, and it's been running live and in profit for 14 months straight.
Ray: "Part of every fee keeps the server and data feeds running, and the rest pays for the hours I put into improving the model."
Ray: "The real reward is watching something I built in my garage earn its keep for people who need it."
Will: There's no minimum, and many people begin with £100 and build the rest from winnings.
Will: Set aside a dedicated pot you're comfortable with, keep it well away from the money that pays your bills, and let it grow at its own pace.
Ray: "Because he was the only one offering to pay for an hour of my time rather than asking for a favour, and I was tired of watching my work make money for everyone except me."
Ray: "It turned out to be the best hour I ever sold."
Kind regards,
Will Chapman
P.S.
As I write this, just 13 of the 100 spots remain.
October is already the model's strongest start to a jumps season, with a week of racing still to go.
£20 once, and the selections keep coming for as long as you want them.
With the 30-day guarantee in place, you've nothing to lose and everything to play for.
The doors close the moment the last spot goes, and I'd hate for you to come back tomorrow and find them shut.