I've got their risk departments rattled, with analysts working late into the evening trying to work out how it picks its horses.
They've been studying the pattern of my bets for months now, hunting for the thread that ties them all together.
Each winner I back ends up as a line in a report on somebody's desk at head office.
They're flagging accounts, trimming stakes down to pennies, and trying everything in their power to stem the losses.
And still the winners keep landing, week after week.
From 10 winning bets out of 17.
Stacked on top of 42 winning bets from 62 in June...
And 45 winners from 66 back in March alone.
October has just gone into the books as my best month yet, with £7,120 in profit landing between the first Monday and the final Friday.
Since the first of January, the algorithm has banked £58,675 for me, and it hasn't posted one losing month along the way.
In a moment I'll show you how it does that, and why the bookmakers are so desperate to find out.
This isn't some complicated system that would take you hours to follow every morning...
Nor is it a bloated piece of betting software that needs a brand new MacBook and a degree in computing to run...
It's a refined, structured process that I run each morning in a matter of minutes.
It's less like tipping and more like holding a stopwatch over a race that hasn't been run yet, then backing only the runners the clock says are priced too big.
When the stopwatch and the bookmaker disagree by enough, I bet, and when they don't, I leave the race alone.
That one idea is the backbone of everything you're about to read.
I didn't come up with it overnight, and it didn't come from a hunch or a word in my ear from a stable lad.
It came from 2 years of building, testing and binning ideas until only the ones that made money on races they'd never seen were left standing.
Before I walk you through it, I want you to see what you're really up against each time you place a bet, because it explains why the algorithm matters so much.
Every price you see on a betting site comes from a pricing team backed by serious mathematics, not from a man with a chalkboard and a wet thumb.
The big firms retired the old-fashioned odds compiler years ago and replaced him with traders who sit in front of pricing models fed with the same form data you and I can see.
Those models price up thousands of markets at once, then nudge each price the moment money lands or drifts.
The traders watching over them don't fall in love with a horse, and they don't chase a loss on a wet Wednesday at Fontwell.
So when a lone punter sits down with a form guide and a cup of coffee, he isn't really taking on a bookie at all.
He's taking on a room full of statisticians and the model they've spent years tuning.
And you won't beat a model by squinting harder at the form, you beat it by building a better model of your own.
My aim was never to out-muscle a firm with a trading floor and a stack of servers, because that's a fight nobody at home can win.
What I needed was a way to catch the moments when their numbers slip, and those moments turn up on most days of the racing week.
Point a model at that one narrow job, and it can outperform a much bigger operation that's trying to do everything at once.
Think of it like a fishmonger and a supermarket.
The supermarket sells a bit of everything at a price that works on average, while the fishmonger knows one thing inside out and spots a bargain the moment it comes off the boat.
My algorithm is the fishmonger, and the bookmakers' pricing desk is the supermarket.
A supermarket can't afford to check each fish by hand, and a bookmaker can't afford to study each runner in each race the way my model does.
The algorithm starts work at 5am by pulling in the full race card data for every British meeting that day.
It takes in each declared runner, the course it's running at, and the going description issued by the clerk of the course.
It also collects the early prices from 6 of the biggest bookmakers, so it knows what the market thinks before it forms a view of its own.
For each runner it keeps the best of those prices, because that's the price a member who shops around can take.
That data is cross-referenced straight away against my own private database, built up over years of meticulous record-keeping.
Race history, trainer and jockey win records at specific tracks, how the stalls position plays at each course and trip, timing readings from each stage of a race, market movement, going preferences, headgear changes and handicap mark trajectories, going back several seasons.
The database now holds more than 410,000 individual runs, each one tagged with over 30 separate pieces of information.
Behind the scenes, it's a set of Python scripts talking to a database on a small server in my study, nothing glamorous, but fast and reliable.
I've written every line of that code myself, and I've rewritten most of it at least twice as I learned what worked and what didn't.
The first run of the day takes the declared runners and joins each one to its full racing history, horse by horse.
Then it works out more than 30 measurements for each runner, such as days since its last run, how its weight compares to its last winning mark, and how its trainer's horses have run over the last fortnight.
Those measurements go into the scoring model, which hands back a number for each horse in each race.
To give you a sense of the scale, a dedicated human form student might work through a dozen races in an evening before his eyes give up on him.
The model reads each runner at every British meeting on the card, weighs more than 30 separate factors on each one, and has the full card scored in under 4 minutes.
It isn't cleverer than a good judge of a horse on any one race.
Its strength is that it gives the 4.20 at Southwell the same attention as the big race at Newmarket, and it never lets yesterday's result colour today's thinking.
That last point matters more than most punters realise.
After a bad afternoon, a person bets differently the next morning, whether they mean to or not.
The model starts each day with a clean slate, judging the runners in front of it and nothing else.
The algorithm then runs each runner through a sequence of logical filters, scoring each horse across a range of weighted criteria.
The filters weigh things like track suitability from previous visits, whether the class of race suits the horse's current form cycle, whether the going on the day matches its historical preference, the trip, the trainer's form over the preceding 14 days, the shape the race is likely to take, and whether the current market price represents value against what the model calculates as the true probability of winning.
This is where the real work happens.
Let me show you some of what it's looking at, because this is the part that leaves an ordinary tipster standing.
It maps the likely pace of each race first, working out who wants to lead, who needs to be held up, and how that shape will help or hurt each runner before a stall has opened.
It reads the timing of each stage of a race rather than only the finishing positions, so a horse that stayed on strongly behind a slow early pace gets the credit the bare form line hides from the casual eye.
It knows how the draw plays at each track down to the specific trip and the state of the ground, because a low number at Chester on fast ground is worth much more than the very same draw on a soft day.
It tracks going preference for each horse, and it treats good to soft and heavy as the different animals they really are.
It watches for a horse dropping in class at the point in its form cycle when a drop tends to spark a revival.
It flags first-time headgear, a stable switch, or a jockey booking that doesn't fit the yard's usual pattern, the small clues the market is always slow to price.
And it reads the money as well as the odds, so it can tell a price that's drifting because the horse can't win from a price that's drifting only because nobody has noticed it yet.
It even notes how far a trainer has sent a horse, because a yard that boxes up one runner and drives it 400 miles to Perth rarely does it for the scenery.
Each of those factors nudges a horse's score up or down by an amount the model learned from years of results, and those nudges add up to a price.
The model learned them by working through 7 seasons of results and measuring how much each factor shifted a horse's chances once all the others were accounted for.
Some factors that sound important, like a horse's last finishing position, turned out to be worth very little once the rest of the picture was in.
Others that most punters ignore, like how many days a horse has had off since its last run, carried real weight.
A few factors only matter in certain types of race, so the model treats them differently depending on the contest.
The draw, for example, can decide a 5 furlong sprint at Beverley and barely matters in a 2 mile handicap at Haydock.
Rather than force one set of weights onto every race, the model keeps separate weights for sprints, middle distances, staying races and all-weather racing.
This is the technical heart of it, so let me explain the next step in plain English.
Once every runner has a score, the model turns those scores into a set of win chances for the race, and those chances always add up to 100%.
A horse might come out at 24%, another at 18%, and so on down to the outsiders at 1% or 2%.
Those percentages become a fair price for each horse, so 24% works out at roughly 3/1, and 18% at a shade over 9/2.
Then the model looks at what the bookmakers are offering on the same horses.
Bookmakers' prices always add up to more than 100% across a race, often 115% or more on a big handicap, and that extra slice is their built-in margin.
The model strips that margin out of every price so it can compare like with like.
The bookmakers build their prices based on public money flow and their own margin models.
They employ very clever people, and their pricing models are serious tools that reprice thousands of markets a second.
But their models are designed to generate profit across millions of bets, not to identify each mispriced runner on a given afternoon.
Their models only have to be right on average across the whole book, and the margin they build in means that even when they're wrong on a race here and there, the maths still pays the firm handsomely at the end of the week.
That margin, and that right-on-average design, is the weak spot in their armour.
It means the firm can happily leave one horse a couple of points too big, because that runner barely moves their overall numbers.
My model has no interest in their overall numbers.
It only ever hunts for the individual runners their pricing has left too big, and it ignores the rest of the card.
That difference between their price and the true chance is where my algorithm earns its keep.
It finds horses whose real chance of winning is meaningfully higher than the bookmakers' odds suggest.
It's a regular, repeatable and significant advantage that shows up in my bank balance month after month.
There's another reason the timing of the email matters so much.
The early prices that go up the evening before racing are set on thin information, before the bulk of the money has arrived to tell the traders where they've gone wrong.
That's the window my model works in.
By the time the big money lands later in the day, many of those generous prices have been cut, which is why the selections reach you first thing.
You're not really gambling against the bookmaker any more when you work this way.
You're backing the prices they've got wrong and letting the arithmetic pay you over hundreds of bets.
There's one more check I run that most people building models never bother with.
I group the horses by the price the model gave them and count how often each group went on to win.
Across the last 2,000 qualifying runners, the horses it made 3/1 won about as often as a true 3/1 shot should, and the same was true right down the price range.
When a model's numbers line up with reality like that, you can trust the prices it builds, and that's what lets it bet with confidence.
If the horses it made 3/1 had been winning half the time, or hardly ever, I'd know the model was off and I'd stop betting until I'd found out why.
I run that check at the end of each month, and it's one of the first things I look at when I close the books.
Finally, it produces the day's selections, usually two, sometimes three or four, each carrying a high score based on everything the model has weighed.
Each selection arrives with the horse's name, course and race time.
There's no jargon, and no complicated ratings to decode.
You read the short email each morning, place the bets with any standard UK bookmaker, and get on with your day.
My own end starts a little earlier, with the alarm going at ten to five and the first run under way soon after.
A second run at 7am catches any late changes, and the email is on its way to members before 8am.
Most members tell me they have their bets placed within 5 minutes of opening it.
The email is plain text, so it opens in a second on any phone, even on a slow connection on a train.
There's no app to download, no password to remember and no dashboard to log into.
Now, that might sound a little technical depending on your level of expertise...
But here's the good news...
Anyone can make this work, whatever their background or know-how.
If you can open an email and navigate to a bookmaker website, you have everything you need.
You don't need to understand a word of what I've just explained about win chances and margins, because the model handles all of it before the email is sent.
To make up to a further £5,500 before the end of the month...
No matter your betting past, how much you've won or lost...
Regardless of your level of expertise...
Whether you're retired or holding down a busy full-time job...
You can follow my algorithm's selections and earn as much as £2,000 in pure profit this week.
Now, before we go any further...
I want to tell you about 4 of my members who've been following my bets over recent weeks...
£1,340 up since the middle of August, and the only thing I've changed is where my £10 goes each morning. I've been betting on horses since Red Rum was winning Nationals, and I've never been able to write that sentence before.
Russell Brookes, Stockport
What do you do when the selections arrive at 7.45 and your first meeting starts at 8? You place them on the office stairs in under a minute, which is how I've banked £860 in 5 weeks without missing a single morning.
Denise Fairclough, Preston
If you're the type who has to see something working before you trust it, start with £5 stakes like I did and let a month convince you. My little £5 bets made me £410 in October, and I've just moved up to £15 a time.
Clive Mottram, Grantham
I was standing in the queue for the bar at Redcar in late September when my phone buzzed with a 9/2 winner at Wolverhampton that the algorithm had picked that morning. That was my sixth winner of the week, so the round was on me, and I didn't have to think twice about it.
Joanne Thirlwell, Manchester
Those 4 members joined over the summer as part of my second trial intake, and their results have been excellent across the board.
A few of them had a history of picking the odd winner, making a modest amount from their bets now and then.
But most had lost regularly and never found anything that put the odds on their side.
As things stand, you have a chance to become one of my members...
Take the Bet Enigma Pro challenge, and see how much you can win this week...
Before I tell you my story in full, I want to share something the betting industry would much prefer you never thought about too hard.
Most people who bet on horses lose money.
Not because they're stupid.
Not because they haven't tried hard enough.
But because the entire system is structured to beat them before they even open their betting app.
The bookmakers employ entire teams of mathematicians, statisticians, and data scientists whose full-time job is to make sure the odds are always tilted in their favour.
They've been doing it for decades, and they're extraordinarily good at it.
They have pricing models and computing power that most punters couldn't dream of competing with.
Their models aren't perfect.
There are repeatable pricing errors buried in the data that the market gets wrong week after week.
Not because the bookmakers are slow or careless.
It's because their models are built to generate profit across millions of bets, not to find each mispriced runner on a Monday afternoon at Ffos Las.
Those mispriced runners are where my profits come from.
And here's why ordinary punters and most tipsters never find them.
People back horses they like the name of, jockeys they admire, or trainers they've had a good run with recently.
None of that has anything to do with value.
The bookmakers price the market on hard numbers.
You need hard numbers of your own to fight back.
The moment you back a horse because you're fond of it, you've handed the firm the advantage before the stalls even open.
A horse wins twice on the bounce and suddenly the whole country piles on at 4/6 when the true chance is closer to evens.
The bookmakers know this, and they price accordingly, luring casual money onto poor-value favourites and cashing in day after day.
My model weighs those two wins against everything else the horse has done, so it never gets swept up in the excitement.
A punter hits a losing run, doubles their stake to chase losses, and wipes out two months of decent work in a single bad afternoon.
Without a fixed staking plan, even a good selection method falls apart over time.
Every selection I send is backed at the same stake, win or lose, so one bad afternoon never decides a month.
There are hundreds of tipping services out there sending out selections based on gut feel, newspaper form guides, and recycled market information.
Ask most of them for a full record of every bet they've sent this year, and watch how quickly the conversation changes.
They don't have a model that survives a full season of results, and they can't tell you why a selection qualified beyond a hunch.
Five tipsters in your inbox means five conflicting strategies fighting each other for your money.
You can't build a lasting advantage when you're being pulled 5 ways at once.
What you need is one clear, data-led method that you follow each morning without fail.
And I'll tell you what I learned about the tipster trade, because I spent years funding half of it myself.
A tipster studying the form by hand is doing slowly the job the firms handed over to their pricing models the best part of a decade ago.
He's working from the same public form everyone else can see, so by the time he spots an angle the market has usually already swallowed it whole.
He can't test whether his hunches actually made money across a thousand past races, because he's never kept the records to find out.
That isn't a dig at his knowledge of horses, which is often excellent.
It's just that knowledge of horses and a model that prices them are two very different tools, and only one of them scales.
I need to be clear with you on a few things...
I can't promise to make you rich overnight.
I won't promise 100% winners, and you will lose some races, but you'll barely notice them when the winners come in so often...
The algorithm isn't 100% accurate, as you'd expect with a sport as unpredictable as horse racing...
But what I can promise is that you'll make money over time, reliably and month after month.
Most likely in the first week, almost definitely in the first month...
And you'll see results at a level that no tipster or system you've tried before has ever come close to.
And across the 18 months I've logged its results in full, the total profit stands at £95,475.
All from the click of a button, without any gut-feel guesswork on my part...
No hard graft, and no insider knowledge needed...
With no reason to ever follow another tipster again...
Or wade through the chatter they push out on the tip sites and across social media each morning...
I can promise to turn you from a losing punter into a winning one.
And I can promise that each selection you receive will have passed the same tests that produced the figures on this page.
With a chance to make up to £2,000 this week from my selections...
Let me spell out what this is not...
This isn't the usual service that sends you overpriced favourites and calls it tipping...
I won't forget to mail your tips every morning, the way most so-called tipsters manage to on the very days when there's serious money to be made...
And I'll be there via email to help you every step of the way.
I know you haven't reached this page by accident...
You're searching for a real, competitive advantage when selecting your bets...
You love the sport of horse racing, and you know there's money to be made if you can only find the right method...
Until I discovered this algorithm, I was in the same boat as you...
My name is Geoff Stephenson.
And for years, I was the worst kind of punter.
Trusting mediocre tipsters...
Listening to so-called experts who talked a good game and delivered nothing...
Building half-arsed staking systems that worked for a fortnight and then collapsed without warning...
Chasing losses on the last race at Wolverhampton on a Tuesday night when I should have been asleep...
I'm from Berkshire, I've followed the horses since my early twenties, and for the best part of a decade I was the definition of a losing punter.
It started with a works outing to Newbury and a £5 win on a grey I'd backed because I liked the way he walked round the paddock.
I collected £35 at the window, and for years afterwards I chased the feeling of that afternoon without ever getting close to it again.
The trips to Newbury became regular, and before long the betting app on my phone was where I spent most of my evenings.
I'd read the form for hours before each meeting and still find myself backing the horse the man next to me in the queue had mentioned.
Every Friday I'd top up my betting account from my wages, telling myself this week would be different.
I'd win enough to keep me interested, then give it all back and more within a few days.
Every week I'd put more funds on my credit card to bet with...
I followed hopeless tipsters and backed bad-value favourites, making every mistake in the book.
I can't bear to think about the total amount I handed over to the major bookmakers across those years.
My family were sick of me talking and thinking about betting around the clock...
My wife used to pull a face whenever I disappeared upstairs with my laptop and a coffee after dinner.
She knew what I was doing, and she knew it was costing us.
Our daughter was saving for her first car at the time, and I remember feeling sick when I worked out I'd lost more that spring than she'd saved all year.
I'd had a shocking week at work, money was tight, and I'd convinced myself that a big Saturday accumulator was the way to fix everything in one go.
Five horses, a fortnight's spare money riding on it, and four of them duly won.
The fifth got beaten a nose in a photo, and I sat out on the back step in the dark for the best part of an hour without saying a word.
That was the night I finally understood that hoping harder was never going to be a plan.
If I was ever going to beat this thing, I'd have to treat it as a problem to be solved rather than a feeling to be chased.
That accumulator would have paid £4,300, and I'd backed it with money we needed for the gas bill.
I went back inside, deleted the betting app from my phone, and kept it deleted for 3 whole months.
It was the longest break I'd taken in years, and it gave me the space to think about betting as a puzzle instead of a habit.
Those 3 months were hard, and there were Saturday afternoons when I sat in front of the racing on television with my hands in my pockets.
But for the first time I watched races as a puzzle to be studied, noticing which horses ran better than their price suggested and wondering whether there was a way to spot them in advance.
I knew that with this much dedication and experience, something would eventually click...
By trade I'd spent the best part of 20 years building and testing data systems for businesses, the sort of work where you're judged on whether the numbers can be trusted.
After 20 years of that work, one habit had stuck with me above all others, which was never to believe a number until I'd tried my hardest to prove it wrong.
Looking back now, it seems obvious.
I'd been trusting tipsters' claims and my own hunches without ever testing them, which is the one thing I'd never have done at work.
At work, if somebody told me a new process would save the company money, my first question was always to see the figures from the last 3 years.
Yet I'd been handing over £30 a month to tipsters on the strength of a few lines of sales copy and a photo of a betting slip.
During those 3 months away from the betting app, I started gathering all the racing data I could get my hands on, and I kept going long after the break had ended.
Every result and race detail went into a private database that I built up over more than 2 years.
That database is the foundation of everything I do now.
After that, I was ready to start hunting for patterns and build something worth trusting...
I bought 7 seasons of historical results from a commercial data supplier, and I paid for a daily feed of race cards and early prices on top.
The daily feed arrives overnight as a set of files, and the first job each morning is checking those files are complete before the model is allowed near them.
It cost more than I'd ever spent on a tipster, which tells you how much I'd been spending on tipsters.
The first thing I learned is that the data itself is a swamp.
Going descriptions that mean different things at different tracks, non-runners that need stripping out, timing readings that go missing for whole meetings, and horses that change names when they're sold abroad.
I spent months cleaning and standardising all of it before one useful pattern could surface, and that unglamorous grind is the part no ordinary tipster ever does.
I'd already been writing Python and database code at work for years, so the building itself didn't frighten me.
What took the time was deciding how to handle each awkward case, such as a race declared void or a meeting switched to a different track at the last minute.
By the end of that clean-up I had one tidy record for each run, and I could ask the database any question about the past 7 seasons and get an answer in seconds.
Next I built a basic rating for each horse, based on how fast it had run against the standard time for each course and distance.
Then I noticed the problem, which was that it found the same winners everybody else could see, and they were all trading at prices too short to make any money.
A rating that agrees with the market can be right all day long and still leave you skint.
That was the first big lesson, and it shaped everything I built afterwards.
Picking winners was never the goal, because the bookmakers are very good at pricing obvious winners.
The goal was finding horses with a better chance than their price suggested, whether they looked like winners to the eye or not.
Looking for combinations of variables that, historically, produced a meaningful profit over a large sample of races.
It's not as easy as it sounds, believe me...
Even finding a system that worked in back-testing was painful...
Only for the method to fall apart within weeks of running it live...
It took me far too long to work out why that kept happening.
I was fooling myself, over and over, with a trap that catches almost everyone who tries this.
If you torture past data for long enough, you can always find a rule that would have won a fortune last year, in the same way you can always find shapes in the clouds if you stare hard enough.
But a rule that's been bent to fit yesterday's results perfectly has latched onto coincidences, and it falls to pieces the moment it meets a race it has never seen.
I started building each idea on one stretch of racing history and then testing it on a separate stretch it had never touched, so the model had to prove itself on races it couldn't possibly have been tailored to.
If an angle only worked on the seasons I'd built it from, it went in the bin, no matter how pretty the profit line looked.
What survived that process was a small, stubborn set of patterns that held up on going they'd never seen and at tracks they'd never been tuned for.
Out of more than 200 ideas I tested in that period, only 11 made it through.
The rest went in the bin, including a few I'd grown very fond of.
One of my favourites was a rule about horses dropping back in trip after a run over further, which looked brilliant on the seasons I'd built it from and lost money on each season it hadn't seen.
It hurt to let it go, but that's the test working as it should.
Every month I'd build and test multiple systems to see what would hold up...
I kept detailed records of everything so I could track performance accurately...
Some lost heavily, a few nearly broke even...
Occasionally one would turn a profit over the first 30 days, only to flatline shortly after...
I'd win for a week or two and then watch the advantage evaporate.
Some nights I'd be at my desk until gone midnight, rerunning the same test with one setting changed, hoping the numbers would come out differently.
I came close to packing it all in completely on more than one occasion.
My wife suggested, very patiently and very reasonably, that perhaps I should redirect all that energy into something that actually paid.
She had a point, and I knew it.
But I couldn't let it go.
Because I could see, when I looked at the data from a purely analytical perspective, that the patterns were there.
I just hadn't found the right combination of variables to bring it to the surface month after month.
That was until one quiet evening over the winter heading into 2025...
I wasn't even trying to force a breakthrough that night.
I was flicking through some old race card data almost out of habit, a glass of wine beside me, the house quiet.
A set of conditions that kept cropping up in the winning columns with a frequency that shouldn't have been coincidence.
Horses returning within 14 to 21 days of a previous run, running at a track where they had a strong historical profile, in a class that suited their current form cycle, where the market had them at a longer price than the data suggested they deserved.
I pulled up every qualifying example across the previous three seasons and ran the numbers.
The winners among them had gone off at prices much bigger than their chances deserved.
Profit on level stakes at starting price, positive across all three seasons tested.
I stayed very still for a moment.
Then I checked it again.
Then I tested it on the 4 seasons before those, which I hadn't touched, and it held up there as well.
I spent the next few days coding that core pattern into a full scoring framework, then layering the surviving angles on top of it, going preference, how the draw plays at specific tracks, trainer and jockey form, likely race shape, and market position at the moment of declaration.
By the first full week of January 2025 I had a working model.
The last piece I bolted on was the part that actually turns all of this into money.
For every runner the model likes, it works out its own estimate of the horse's chance and turns that estimate into a fair price.
Then it compares that fair price to whatever the bookmakers are offering, and it only ever pulls the trigger when their price sits far enough above fair to clear their built-in margin and still leave a profit.
In practice, that means the bookmaker's price has to be at least 20% bigger than my fair price before a bet goes out.
A horse I make a fair 4/1 shot needs to be available at 5/1 or bigger, and if it's only 9/2, it stays in the database.
That bar sounds high, and it means some days only one horse makes it through, but it's the reason the profits have held up month after month.
Lower it to 10% and the model sends more bets, but the extra ones are the marginal calls that eat away at the bank.
No value in the price, no bet, however lovely the horse might look on paper.
That one rule separates backing winners from making a profit, and almost every punter alive ignores it.
The very first week I ran the finished model live, back in January 2025, it returned £750, and the week after that it returned £550...
Then the results really took off.
A remarkable early run pushed my betting bank up towards £5,000 inside a matter of weeks...
I knew this was something big, but I kept my feet on the ground...
I kept feeding new results in, refining the filters and running the model each morning with more care than I'd ever given anything betting-related in my life.
By the end of that first month I'd banked over £4,500 from the live model.
I remember transferring the first serious withdrawal into my current account and just sitting there looking at it on my phone.
The winners kept coming through the spring, and I was making withdrawals on a regular basis.
I kept the betting bank separate from our household money from day one, and I only ever withdrew profit, never the stake money.
That habit kept the bank growing and gave me a clear record of what the model had made.
My wife noticed the difference before I said a word, because for the first time in years I was closing the laptop at night in a good mood.
The model kept improving through 2025 as more results fed back into it, with each filter tightened and each unusual case accounted for.
That May I also started logging every bet, every price and every return in full, so I'd have a record nobody could argue with.
The first real test came in September 2025, when profit dipped to £3,325, and I wanted to know why before it happened again.
When I pulled the month apart, most of the losers were in races for 2-year-olds, where many runners have only had a run or two and there's very little form to judge them on.
So I changed the rules, and now a 2-year-old has to have run at least 3 times before the model will consider it at all.
October 2025 came back with £6,880, and that rule has stayed in place ever since.
2025 was the year the algorithm proved itself, but 2026 is the year it grew up, and that's the part of the story I want you to hear in full.
I used my second income to upgrade our family home, with a new kitchen and the garden room my wife had wanted for 10 years...
Taking the family somewhere hot for a fortnight, the sort of holiday we'd only ever looked at in brochures before...
And clearing the last of a debt that had been hanging over the two of us for far longer than I'd care to admit...
I also paid for our daughter's second car outright, which felt like settling a score I'd been carrying around for years...
Things my wife and I had talked about for years but had always pushed back to "one day".
By the time 2026 came around, the model was a very different beast to the one I'd switched on that first January.
I'd added the pace-mapping layer over the winter, and it cut out a whole category of losers, because it stopped me backing hold-up horses in races that were always going to be run at a crawl.
Building it took most of November and December, and it meant teaching the model to read each horse's running style from years of in-running comments.
Every run in the database carries a short note describing where the horse raced, so I wrote code to turn those notes into a simple label for each horse, such as led, prominent, midfield or held up.
Count up the likely leaders in a race and you get a good idea of how fast the early gallop will be, and that tells you which runners the race will suit.
January 2026 opened the year with £6,240 in profit, and it felt like the model had finally come of age.
Most of that month's racing was on the all-weather at Lingfield, Kempton, Newcastle, Southwell and Wolverhampton, where the pace layer earned its keep straight away.
February was quieter at £5,470, one of those grinding months where nothing spectacular happens and the bank climbs anyway.
Then March 2026 gave me my best month at the time, £6,995 from 45 winners in 66 bets, and it was the run that convinced me this was ready to put in front of other people.
The turf season got under way at Doncaster towards the end of that month, and the model handled the switch from all-weather to grass much better than I'd feared.
April cooled off to £4,725, and I'd ask you to take note of that one.
The ground was changing almost daily that month, from good to firm one week to soft the next, and the model was pricing some horses on going they'd never run on.
Finishing £4,725 ahead in those conditions tells you more about a method's foundations than any flashy month could.
It also showed me where the model was weakest, so in May I added a breeding layer to judge a horse facing ground or a trip it had never met.
Instead of guessing, it now looks at how the horse's sire and dam's other offspring have handled those conditions, along with clues from the horse's own action and earlier runs.
May 2026 bounced straight back to £5,880, and June went better again at £6,610 from 42 winners in 62 bets.
June was the busiest month of the year for qualifying bets, and the extra races gave the model more chances to find runners the firms had priced too big.
July came in at £4,540, the softest month of the year so far, and even that was more than I used to lose in a typical month as a punter.
That summer I made the improvement I'm proudest of, which was a second run of the model at 7am.
Overnight rain, a late non-runner or a change to the going can alter how a race will be run, so the 7am run catches those changes before the email goes out.
If a selection no longer clears the bar, it's dropped, and if a new runner qualifies because the ground has turned in its favour, it's added.
August 2026 brought in £5,935, with much of it coming from small-field handicaps at the northern tracks.
September added £5,160, during a spell when the form book was getting muddled as many horses passed their peak for the season.
In September I also started printing a minimum price beside each selection in the email, so members know when a price has been cut too far to be worth taking.
That one line has stopped members backing horses at prices the model would never accept.
Then came October, and it gave me £7,120, the best month the algorithm has ever produced.
The ground softened across the country in the first week, and dozens of horses found themselves running on going they'd never faced before.
The bookmakers priced many of them on their summer form, while the model priced them on what their breeding and earlier runs said about soft ground.
That breeding layer from May paid for itself many times over, from the opening Monday of the month to the closing Friday.
Members who'd joined over the summer were sending me screenshots of their October accounts, and more than one asked whether the model had been tweaked for the winter ground.
It had, back in May, and October was the first month the ground gave it the chance to show it.
I've only just closed the October books, and that figure still puts a smile on my face.
Ten months into 2026, the model has already banked £58,675, which is more than many people take home from a full-time job.
Every one of those months began with the same routine, the same two runs of the model and the same short email, and the results came from sticking to it.
It's that across the whole of 2026, there isn't one losing month anywhere to be found.
The same holds for each month I logged in 2025, which makes it 18 in a row.
Not a single month in the red, and not one wobble that ever lasted longer than a bad week.
My family started to really take notice...
I helped a few mates out with their bets too...
The idea of running my own small betting club slowly started to take hold...
Sharing my winners with a small community...
Helping ordinary people claw back a bit of money from the bookmakers for a change...
I took in my first outside members at the start of 2026, to see whether the results would hold up for ordinary punters following the selections remotely, with no background in data or systems.
It worked, and they've all reported back with results that track almost identically to my own.
A second small group joined over the summer, and their figures have matched mine just as closely.
Some of them bet £5 a time and some bet £50, but every one of them is ahead.
I answer their emails myself, usually the same evening, and some of the best ideas for improving the service have come from their questions.
The minimum price in each email, for one, came from a member who asked how far a price could shorten before a bet stopped being worth it.
I'm here right now to tell you...
I can help you make serious money from your betting, starting this week.
£58,675 in tax-free profit across the first 10 months of 2026.
The quieter months are still profitable.
Even the softest month of the year cleared £4,500, which tells you how sturdy this method is.
There's not a losing month anywhere in the run.
That reliability is what separates this from anything else on the market.
You've given this page your time, and that tells me a lot about you...
You're here for a reason, whoever sent you to this page thinks you're perfect for this opportunity...
They, like me, want you to succeed, to grab back a small slice of the pie the bookmakers have taken from all of us...
To fight back against the firms and bring home some real cash from betting for a change...
No more payday loans loaded into your betting balance...
No more stressing over a string of losing horses...
No more hiding the bank statement when it drops onto the doormat...
No more lying awake working out how to win back yesterday's losses...
True peace of mind with your bets...
A chance to win up to £2,000 this week from my algorithm's selections...
I've coached junior cricket for 20 years, and the thing I drum into the youngsters is that you score runs by leaving the bad balls alone. This algorithm bats the same way, letting the poor races go past, and it's earned me £1,120 since July by only playing the shots worth playing.
Barry Lockwood, Doncaster
My one complaint is that it doesn't send more bets, because some mornings I open the email hoping for 4 and get 1. Then I look at my account, sitting £975 higher than it was in August, and the complaining stops pretty quickly.
Sandra Heppell, Darlington
On the Monday I joined I had £200 in my betting account, and by the Friday of my second week it had reached £520. Six weeks on it stands at £1,090, and I haven't topped it up from my wages once.
Lee Hollins, Colchester
I build databases for a bank, so the first thing I did was ask Geoff how he stopped his model fitting itself to old results, and his answer was better than the one I'd have given. I've since made £780 following the selections, but it was that conversation that told me this had been built by someone who knows what he's doing.
Ian Prescott, Shrewsbury
"I've heard it all before and it's never amounted to anything. What makes you any different?"
"I've been where you are, and I know how it feels. I spent years losing money and trusting the wrong people, and it cost me dearly. The difference is that this isn't built on gut feel or recycled form guide tips. It's a data model that's been running live and making a profit since January 2025, and I've just told you what every month of this year produced. It costs me very little to help a small community of punters make similar money to me, so why wouldn't I? My daily needs are already covered by my winnings. I want to do something worthwhile with what I've built."
"I'm so tired of losing, how is this actually going to make me money?"
"I feel you. I've lost money in every way imaginable at betting, and it took me years to find a method that works month after month. The algorithm does something that no gut-feel tipster can do: it takes emotion out of the selection process, finds price errors the market has missed, and backs only horses whose chance of winning is meaningfully higher than the odds reflect. Follow the bets as sent, with flat sensible stakes, and the maths does the rest."
"What if I can't afford to lose more money on the horses right now?"
"Then Bet Enigma Pro is what you need. You can follow along without staking a penny for the first week if you prefer, tracking results before you commit any betting funds. The performance will speak for itself within a few days. And the only upfront cost is your £20 membership, which is fully protected by a 30-day money-back guarantee."
"When will I start to see money coming in?"
"Potentially from day one. The selections go out every morning and results typically come in throughout the afternoon and evening. Most members are in profit within their first week. By the end of your first month, you'll be looking at figures that would have seemed unrealistic before you joined."
Right, I've got to tell you one more important thing...
Memberships inside Bet Enigma Pro are severely limited right now...
This is the first full public intake I've run, and I need to keep numbers tightly controlled to protect the quality of the selections for everyone inside.
Only 50 memberships are available, and once they're gone...
You'll have to get your name on the waiting list for the next intake...
That's likely to be well into 2027, and at a significantly increased price.
The time is now, and the next email goes out first thing tomorrow morning...
I fully expect all 50 memberships to be claimed within hours, not days.
Read on to the end of my proposal before you decide anything...
Welcome to the turning point in your betting journey...
What does the morning email look like?
It's short. You'll see the date at the top, then each selection listed with the course, the race time and the horse's name, followed by the minimum price the model wants before the bet is worth placing. If anything changed after the 7am run, I'll mention it in a line above the selections. Most mornings the whole email fits on a phone screen without scrolling.
How can I be sure a selection wasn't sent after the race was run?
Each email is stamped by the mailing system with the time it left, and I have no way of altering that stamp. The selections always land before 8am, hours before the first race of the day, and you can check each one against the result yourself.
Should I take the morning price or wait until the off?
Take the morning price, and use a bookmaker that offers Best Odds Guaranteed on UK racing wherever you can. That way, if the horse drifts and wins at a bigger starting price, you're paid at the bigger price anyway. My selections qualify because the early price is too big, so waiting usually costs you money.
Does the algorithm ever back each-way?
Very rarely. The model is built around win chances, and it does its best work backing a horse to win. If a selection is a big price in a big field, I'll sometimes note that an each-way bet makes sense, and I'll explain why in the email.
Will I pay tax on my winnings?
No. In the UK, betting winnings aren't taxed for punters, so every pound you make from the selections is yours to keep. I'm not a tax adviser, so if your circumstances are unusual, it's worth checking with someone who is.
What happens on days when there's no British racing at all?
There are only a handful of those each year, such as Christmas Day, or days when frost or flooding wipes out the whole card. On those days the email tells you there's nothing to bet, and the selections pick up again with the next day's racing.
Why does it cost £20 and not £200?
Because I'm not trying to make a living from membership fees. The algorithm already pays my bills, and I'd like this to be within reach of anyone who's lost money to the bookies over the years. A £20 fee makes that possible without opening the doors to thousands of people at once.
Can I ask you why a particular horse was selected?
Yes, and members often do. I'll tell you which factors pushed its score up and how far the bookmaker's price sat above my fair price on the morning. I enjoy those emails, because they show me people are taking an interest in how it works.
Does it matter which day of the week I join?
Not at all. Each day's selections stand on their own, with no running sequence or linked bets to catch up on. Join on a Tuesday or a Sunday and your first email arrives the next morning that racing is on.
How do I get my £20 back if I decide it isn't for me?
Send a short email to the support address at the bottom of this page, or contact ClickBank directly, within 30 days of joining. You don't need to give a reason, and the refund goes back to the card you paid with.
Is there anything else I'll be charged for later?
No. Your £20 is a one-off payment for lifetime access to the daily selections. There's no subscription, no renewal date and nothing hidden in the small print. If I raise the price for future members, yours stays as it is.
What's the best way to get started on day one?
Set aside a betting bank you can afford, decide on a flat stake of around 3% to 5% of it, and open an account with a bookmaker that offers Best Odds Guaranteed if you don't already have one. Then wait for the email, place the bets, and leave them alone until the evening.
Follow my daily tips for up to 30 days and see how much you win.
Follow along on paper or place real money on them, it's up to you.
If, in the unlikely event that you're not entirely satisfied with my results, you can get 100% of your membership fee back, right away, with no hassle and no awkward questions.
Right, onto the last thing...
How much does lifetime membership cost?
Not as much as you'd expect, given what you've just read...
I'm giving you a lifetime's supply of daily selections for a one-off fee of £20.
You won't have to pay any monthly subscription fees...
There are no hidden charges anywhere...
A model that's banked £58,675 so far this year, available for less than the cost of a tank of fuel.
I think that's a bargain personally, and if you agree...
All you need to do now is click the button below...
Complete my 100% secure checkout form, and your first selections will be on their way to you...
Remember, you have up to 30 days to see regular profits...
More likely, you'll be on the road to regular weekly profits within your first few days.
Each of my trial members so far is seeing results beyond what they expected when they joined.
And I think you will too, just click the button above and let's get started.
There's a welcome guide waiting for you the moment your order goes through,
Yours sincerely,
Geoff Stephenson
P.S. If you're still on the fence, think about what you've just read.
October's £7,120, the best month it has ever produced.
Not a single losing month in the entire run.
Trial members who are all seeing similar results within a few weeks.
A full 30-day money-back guarantee on your £20 membership fee.
And only 50 memberships available before this closes.
The only way you lose today is by clicking away and letting someone else take your spot.
There are very few of them left.