You already know the odds feel stacked against you, even on the mornings you've done everything right.
Not because you pick the wrong horses.
Not because you lack knowledge, or because you haven't put the hours in.
It's because every race card you've ever studied is written in a kind of code, and nobody ever handed you the full key to it.
In truth it's a compressed story about ground, trip, class, fitness and intent, squeezed into six characters and left for you to unpack.
Multiply that by every runner at every meeting and you're looking at thousands of these little stories waiting to be read every single morning.
Most of us read them the way a holidaymaker reads a menu abroad, picking out the few words we recognise and guessing at the rest.
The major bookmakers stopped reading cards that way years ago, and every race they price now has software behind it that reads every line at once.
Their pricing tools work through millions of data points and reshape the odds as money arrives in their markets.
They also keep a close eye on individual accounts, and the moment a customer starts winning regularly at decent prices, somebody takes notice.
That's who you're up against every morning, sat with your phone and a mug of tea, trying to read a foreign language by candlelight.
I was losing more than I won, not through bad luck, but because I was bringing a pen and a hunch to a contest the other side had long since automated.
Then I stopped trying to out-guess the market with a form guide and a gut feeling.
I spent 18 months building software that could read the whole card fluently, every runner and every line of it, before the first race of the day.
I called it Race Logic Pro, since unpicking that code every morning is the whole of its job.
£64,410
12 Months
222
Winners
367
Bets
My name is Stuart Taylor, and I'm going to show you precisely what that software produces.
In August 2026, the software I built banked £5,960.40.
It's what settled into my betting accounts across a single month, from win-only bets placed at flat stakes.
It wasn't a freak month either, as the months either side of it tell the same story.
July 2026 brought in £7,080, the best month the software has ever produced, and September has added another £6,460.
Put those three summer months together and you get £19,500 from a routine that takes about 5 minutes of my morning.
Before I explain how any of it works, let me hand you over to a few of the people who saw it first.
When I finally opened the software up beyond my own accounts, I gave a small group 30 days of the morning selections and asked each of them to keep a written record of what actually happened.
"I joined purely to catch you out, and I kept a spreadsheet so tight my wife asked if I'd taken up accountancy. A month later the only thing it had caught was £2,140 of profit, so I've stopped hunting for the trick and started enjoying my mornings."
Kevin Ashdown, Stoke-on-Trent
"My husband used to hide his betting slips in the glovebox, so when he suggested we follow the selections together I very nearly said no. We finished the month £1,880 up between us and, for the first time in our marriage, the horses are something we talk about over tea instead of avoiding."
Julie Pemberton, Hexham
"The email lands while I'm stood on the platform waiting for the 7:52, and the bets are on before the train pulls in. That's the whole routine, and it put £2,015 in my account over the trial without costing me a minute of my evenings."
Gary Lambourne, Ipswich
"How many systems have I bought since 1996? More than I'd admit to anyone, but this is the first one where the notebook I keep shows a figure in black ink at the end of the month, £2,230 to be precise."
Derek Oldfield, Kendal
So what do those four people actually share?
Not one of them had been turning a steady profit from betting before they joined.
Each of them arrived with a profile you'll probably recognise in yourself.
They had a real love of the horses and a nagging sense that something at the heart of their approach was missing.
Every one of them could read a race card well enough to have an opinion, and not one of them could read it well enough to be right about the price often enough.
That missing piece was never more knowledge or another afternoon of research.
It was a systematic way of reading far more of the card than any person sat at a kitchen table could ever hope to take in.
With somewhere between £1,880 and £2,230 in their pockets from a single month of following the selections, they'd finally got their hands on it.
But first, it's worth understanding why this has been so hard to pull off on your own.
Most punters assume the answer lies in backing the right tipster or simply pouring more hours into the form.
None of those things address the real problem.
The real problem is structural, and it's been sitting underneath every bet you've ever struck.
Their prices are set by teams of analysts and refreshed constantly, with one aim in mind: pulling a long-term margin out of millions of bets a year.
Every horse in every race gets a price designed to keep the house in profit, not a price that accurately reflects that runner's true chance of winning.
Now and then that price is simply wrong, like a misprint in a newspaper, and those misprints are the only thing a punter can profit from over the long run.
Spotting them reliably, before the market corrects them, is something no person working by hand can do at the speed and the scale it demands.
There's far too much data for one person to work through before breakfast.
The window before the price corrects itself is too short.
And the cold detachment needed to back a horse purely on the numbers, rather than on a liking for the look of it, isn't something most of us are wired for.
It makes profitable betting impossible with the wrong approach.
A properly built model, feeding on the same data the bookmakers use and hunting for misprints across every card before the first race goes off, alters the picture completely.
That's the difference between finishing the month down and finishing it with a healthy withdrawal from your account.
That difference-maker is Race Logic Pro.
Let me show you what one of those misprints looks like when it plays out in front of you, because you've almost certainly seen it happen and put it down to luck.
How many times did you back a horse at 5/1 that the market had drifted out from 3/1 first thing, and it went in?
The drift told you the bigger money didn't agree with your assessment.
But the horse still won, and if horses like that keep winning more often than their drifting price suggests, the price was the mistake all along.
That's a misprint playing out in plain sight, and paying you by accident.
Now picture software that hunts those misprints down every single morning, across every meeting and every race on the card, and hands you the best two or three of them before 8am.
That isn't a fantasy or a projection.
It's what the month-by-month story further down this page lays out for you in full.
Monthly Profit At £50 Flat Stakes
September 2025 To September 2026
Here's something most punters never fully take on board about how the bookmakers actually operate.
They don't just price races from the same form data you can read for free every morning.
They watch how money moves across every major platform as it happens, right down to the individual account, and they pick out the customers who keep coming out in front.
If you start winning at fair odds with any regularity, they don't sit back and wait to see how long it lasts.
Your account gets flagged for a look, and your maximum stake gets trimmed, sometimes within a week of a good run getting going.
It's a business protecting itself, just as it was designed to.
Their pricing is far from perfect, though, and that's the important part.
It's built to generate margin across an enormous volume of bets, not to price every runner at every meeting with pinpoint accuracy.
At smaller midweek meetings, where public interest is thin, prices lurch around with whatever money arrives first rather than settling from deep independent analysis.
That leaves mistakes in the pricing, small and brief ones, the kind that vanish the moment serious money starts to move.
But they're repeatable mistakes, and the software is built specifically to catch them before they disappear.
That's the territory Race Logic Pro works in, and it's the whole reason it works at all.
It also explains why the software would rather work a wet Wednesday at a small track than a sunny Saturday at a big one, which is the opposite of how most of us like to bet.
I want to paint you a proper picture of what those 25 years of losing actually looked like, as I suspect parts of it will sound familiar.
It wasn't one single catastrophic failure.
There was a tipster service I followed for four months that started brilliantly and then quietly came apart at the seams.
There was a staking system that looked mathematically sound on paper and produced a steady bleed in practice.
There was a spell when I decided I finally understood the market well enough to back only my own selections, and I lost £1,400 over 6 weeks before I stopped.
And there was a Saturday in the autumn of 2019 when I chased a losing morning across seven meetings and turned a bad day into a properly stupid one, sat in the car outside a garden centre while my wife did the shopping.
I remember the rain on the windscreen and the little spinning wheel on my phone as the last bet of the day went down at Wolverhampton, and I remember deciding not to tell her how much.
That's the part that stings when I look back, not the money itself, but the small secrets it made me keep.
Each experience taught me something about what doesn't work.
The racing writers whose columns I read every morning had their good spells, but they were as prone to blind spots as I was.
The systems I bought either worked for a fortnight and then died, or never worked at all.
And the tipsters I trusted, every last one of them, shared one fatal flaw.
They were all working from the same public information as everyone else, with no data behind their selections and no reliable way to find the horses the market had priced wrong.
What I didn't understand then, but understand completely now, is that the problem was never the horses.
The problem was the method.
I'd been reading the card like a fan, when it needed to be read like a code.
My name is Stuart Taylor, and I'm a 51-year-old family man from just outside London.
I've had two obsessions running in parallel my entire adult life.
One is my family, and the other is betting on the UK horses.
The second one started young, on a cold Saturday at Kempton when I was about nine, holding my grandad's betting slip while he lifted me up to see over the rail.
He backed one at 12/1 that afternoon that came from nowhere off the home turn, and I've been chasing that feeling, on and off, for the better part of 40 years since.
These days I take my own daughter, Ellie, to Kempton once a year, and she's far more interested in the ice cream van than anything happening on the other side of the rail.
The horses got me into bother more than once, never the kind that wrecked anything that mattered, but the slow, compounding kind that builds over years of knowing you're putting in the work and still coming up short.
I studied form the way other people revise for exams, tracking trainers, marking jockeys who were finding their stride, watching the going closely, and keeping a record of every bet I struck going back to 2007.
That record ran to more than 11,000 rows by the time I stopped adding to it, and it lived in a spreadsheet I'd named, with more hope than sense, "Profit".
Those records told me the same uncomfortable story year after year.
Slowly and steadily, more money was leaving my accounts than was finding its way back in.
But it was enough to tell me something was fundamentally wrong with my approach.
I tried tipster services and bought systems off the internet, following famous names and completely obscure ones alike.
I had good patches, weeks where everything clicked and I felt like I'd cracked it.
Then the run would end and the losses would quietly undo most of what I'd built.
I kept thinking I was one more piece of information, or one smarter angle, away from finally getting on top of it.
That feeling kept me going for a long time.
It also kept me losing, right up until the day I stopped thinking like a punter and started thinking like the engineer I'd been all along.
By profession, I'd spent 18 years working as a software systems architect.
My job was designing the data-processing infrastructure that large organisations rely on to handle huge volumes of complex information in real time.
These were systems that swallowed millions of records a day and turned the patterns hidden inside them into a reliable answer in the time it takes to blink.
I understood how to take a messy, complicated data problem and engineer something reliable out of it.
Logistics firms, a regional rail operator and a couple of names you'd know all paid me to do essentially one thing: find the signal buried in the noise and build something dependable on top of it.
One project I'm still proud of predicted which parcels in a national delivery network were about to go missing, hours before anybody noticed they were running late.
It didn't know where the parcels were; it just knew what "about to go wrong" looked like in the data, having seen it happen hundreds of thousands of times.
Keep that idea in mind, as it turned out to be the seed of everything that followed.
Work was work, and the horses were the hobby I couldn't quite bring to heel.
That lasted until one Tuesday afternoon in late 2023, when I was sat at my desk staring at a race card I'd already spent two hours on and something finally clicked.
The week before, I'd backed a horse with complete confidence at 3/1 in the morning.
By the off, it had drifted out to 9/2.
Then it won easily, pulling clear inside the final furlong.
The market had been telling me something, and I'd ignored it completely.
Money drifting away from a horse before a race usually means the bigger players don't share your confidence in it.
But the raw data I'd never properly examined suggested that horse was actually a far better bet at 9/2 than it had been at 3/1 first thing that morning.
I should have been looking at it as a data engineer.
The maddening part was that I'd been that data engineer the whole time, 40 hours a week, solving somebody else's problems.
That thought took root and wouldn't shift.
That evening I opened the "Profit" spreadsheet, all 11,000-odd rows of it, and for the first time I looked at it the way I'd look at a client's data rather than my own diary.
It took me about 20 minutes to find the pattern I'd been blind to for 16 years.
I'd backed horses from three yards I was fond of more than 1,400 times, and those bets alone accounted for well over half of everything I'd lost.
I hadn't been reading the card at all; I'd been reading my own favourite bits of it and skipping the rest.
Software doesn't have favourite bits, and that was the moment I decided to build some.
I started pulling historical race data that same week.
I wanted raw data from several feeds covering thousands of races across multiple seasons: every declared runner, every result, every going description and every price movement from the overnight show through to the off.
The first thing that stopped me was the state of the data itself.
Field names didn't match across sources, trainer names were spelled three different ways depending on the feed, going descriptions meant slightly different things at different tracks, and course names wouldn't line up cleanly between providers.
One feed recorded a horse's age on the day of the race, another as of the 1st of January, and for a fortnight I couldn't work out why half my three-year-olds were apparently four.
Before a single predictive model could run, that entire pile had to be cleaned, standardised and stitched together properly.
That step alone swallowed weeks of evenings after the kids were in bed.
I spent those evenings cross-referencing records, patching holes and building the data pipeline that everything downstream would depend on.
I'd spent enough of my professional life watching predictive systems fail because they were built on a bad foundation.
Feed a model corrupted or unreliable data and it learns the wrong lessons, producing plausible-looking outputs that fall apart the moment real money is on the line.
That isn't a shortcut I was willing to take here.
There's a phrase we used to throw around at work: rubbish in, rubbish out.
In my old job that meant a wrong number in a quarterly report, which was embarrassing but survivable.
Here it would mean my own money walking out of the door dressed up as a confident selection, so I treated the cleaning stage like the foundations of a house rather than a chore to hurry through.
By the time it was finished, the cleaned archive held a little over 190,000 individual runs, every one of them described in the same language and stitched to the same course and going records.
For the first time in my betting life, I had a card I could read from end to end without guessing at any word of it.
I'd love to tell you the first thing I built worked perfectly.
It didn't, and the way it failed taught me more than any early success could have.
In early 2024, impatient to see something working, I wrote a quick rules-based filter over a long weekend.
It picked horses that ticked a list of boxes I'd believed in for years: recent form in the first three, a trainer in decent form, a previous course win, and ground the horse had won on before.
Tested against the 2022 season, it looked fantastic, and I sat up past midnight grinning at the screen.
The rules hadn't found anything real; they'd simply described one season of racing very well and the next season very badly.
My wife found me the following morning at the kitchen table with cold toast and a face like thunder, and asked whether the horses had won or lost overnight.
I told her it was worse than that, since this time a computer had done the losing, and computers aren't supposed to.
That weekend taught me the lesson the rest of this letter is built on: a model has to earn its opinions on races it's never seen, or its opinions are worthless.
Once the foundation was solid and the rules-based filter was in the bin, the real work began, and it looked nothing like handicapping.
I didn't sit down and write another set of rules that said "back the horse if this and this and this are true."
That's how most betting systems are built, and it's exactly why most of them fall apart the moment they meet a race that doesn't fit the mould they were carved from.
Instead, I trained a machine learning model on the history, which is a different animal altogether.
In plain terms, I fed it tens of thousands of past races, each one broken down into hundreds of measurable details, and let it work out for itself which of those details actually moved the needle on the result.
Nobody told it that a certain kind of horse tends to run well in a certain kind of race.
It found those relationships on its own, by being shown what happened tens of thousands of times and being scored, over and over, on how close its estimate came to reality.
It was a probability: a cold percentage chance that each runner would win, built from the data rather than from a feeling.
And a probability is something you can test, which a hunch never is.
Think of it like learning a language by living in the country rather than from a grammar book.
A grammar book gives you rules, and the rules break down the moment somebody speaks with a strong accent.
Living there gives you a feel for what usually follows what, and that feel survives accents, slang and whatever else real life throws at you.
My rules-based filter had been the grammar book, and the model was the years spent living in the country.
There's a concept in this world called calibration, and it became the yardstick I lived by while I was building the thing.
When a good forecaster says there's a 70% chance of rain, it should actually rain on roughly 70 of every 100 days they say that, no more and no less.
I held my model to precisely the same standard.
When it said a horse had a 30% chance, I went back and checked that horses it rated at 30% won close to 30 times in every hundred across the history.
If they were winning far less often, the model was overconfident and useless to me, however clever it looked.
Getting that calibration right across every band of probability took me the best part of the winter of 2024.
Horses the model rated at 5% or 6% were winning closer to 3% of the time, which sounds like a rounding error until you realise that's where the long-priced, tempting bets live.
Fixing it meant teaching the model a little humility about outsiders, and the day I did, a whole category of attractive-looking losers vanished from its shortlist.
The other trap I spent months dodging is the one that quietly kills nearly every betting system ever sold.
It's called overfitting, and it's the difference between a model that has learned and a model that has simply memorised.
If you let a model study the same history for long enough, it will eventually "explain" every result perfectly, right down to the fluke wins and the freak days.
It looks spectacular on the races it was trained on, and it's worthless on the race being run tomorrow, because it has memorised the past instead of learning the patterns that carry into the future.
That's what my rules-based filter had done, only with far fewer moving parts.
I trained it on one block of seasons, then set it loose on a completely separate block it had never encountered, and judged it only on those unseen races.
Then I did it again, and again, rolling the window forward through the years, so that every judgement was made on races that were, as far as the model was concerned, still in the future.
It's a slow and humbling way to build something, because it kills your favourite ideas one after another.
It's also the only way I know to end up with a model you can hand real money to without lying to yourself.
Once the model could be trusted, I started asking it which factors actually predicted outcomes.
I wanted the ones that, tested coldly across thousands of historical races, kept lining up with results once you isolated them from everything else.
Some findings were straightforward, some were mildly surprising, and a handful of them flattened assumptions I'd carried around for 20 years.
Factors the racing public treats as gospel turned out to carry almost no independent predictive weight once they were separated out properly.
A big-name jockey in the middle of a hot streak.
A stable that had sent out a cluster of winners at a particular track recently.
A horse that looked overdue a win after a run of near-misses.
A well-backed favourite from a yard with a big reputation and a bigger following.
They don't move results nearly as often as people think they do.
When a market prices a horse based on factors the data says don't reliably predict outcomes, that price is wrong.
When a price is wrong and you already know the true probability, you're holding a real advantage.
The entire software is built around spotting that, over and over, before anyone else does.
Just as useful were the quiet factors nobody in the pub ever talks about.
How far a horse had travelled to the course that morning turned out to matter in some kinds of race and hardly at all in others.
So did the number of days since its last run, but in a way that varied from one yard to the next, because some trainers win first time back from a break while others plainly use that first run to get a horse fit.
A good punter can hold a handful of those trainer habits in their head.
The model holds every one of them, for every licensed yard it has data on, and updates them every week.
Here's what sets this apart from anything a human tipster can offer you.
The very best tipster alive weighs up somewhere between 20 and 30 factors when forming a view on a race.
That's their ceiling, not because they're cutting corners, but because it's the upper limit of what the human brain can manage reliably before fatigue and personal bias start bleeding into the process.
The model assesses hundreds of weighted variables for every declared runner across every meeting on the day's full card.
It doesn't just ask whether a horse has won at this course before, but on what going, over what trip, in what grade of race, with which rider, coming off what kind of recent run, carrying what weight, and how its price has shifted from the overnight show to the morning market.
It also weights those variables differently depending on the exact shape of the race in front of it.
Where a horse starts from in the stalls matters enormously in a big-field sprint handicap round a tight, turning track where one side of the course has long held an advantage.
In a staying race over two miles on a wide, galloping track, that same starting position barely registers.
Trainer form is a strong signal in maiden races and novice events where there's limited individual performance history to analyse.
In a competitive open handicap with 20-plus runners, it counts for far less, because that information is already baked into the price.
Ground preference matters hugely for a horse with a real leaning towards one particular surface, and hardly at all for one that handles most conditions equally well.
How a price has travelled, from its opening show through to the morning market, gets its own weighting that shifts depending on the race type and the track.
Kempton and Wolverhampton are both artificial surfaces, but they aren't the same surface, and horses that thrive on one can look very ordinary on the other.
The model spotted that split on its own and now treats each all-weather track almost as a separate dialect, with its own grammar of pace, position and trip.
The model learned these distinctions from thousands of historical races.
I didn't program them in as fixed rules.
That's the critical difference between a rule-based filter and a trained machine learning model.
Fixed rules crack at the fringes of the data they were built for, or the moment conditions shift in a way the person who wrote them never saw coming.
A trained model adapts, because it's absorbed enough context to understand which signals actually matter in which situations.
People often picture something grand when I talk about the software, so let me describe what's actually sitting in my house.
Under the stairs, next to the Hoover and a box of Christmas decorations, there's a small server I built from second-hand parts for a little under £600.
My wife calls it "the lodger", partly because of the noise its fan makes and partly because it's the only member of the household that never has a lie-in.
At 5am every morning, the lodger wakes up and pulls the day's declarations, the overnight prices and the latest going reports from the data feeds.
It then runs the cleaning stage, the same one I spent all those evenings building, so that every name, course and going description speaks the same language before the model sees any of it.
Next it builds a profile of every declared runner, several hundred numbers long, and hands each profile to the model.
The model returns a probability for every horse in every race, and those probabilities are checked against each other so that every race adds up to 100%, which sounds obvious but catches more errors than you'd imagine.
Only then does it compare its own figures with the prices on offer, looking for the handful of runners where the market has made a clear mistake in your favour.
At around 7:30am there's a second pass, which re-checks non-runners, jockey changes and any overnight change to the going, and drops anything that no longer stands up.
Whatever survives that second pass, usually two or three horses, goes into the email.
On some mornings nothing survives at all, and on those mornings the email says so plainly.
The whole run takes about 11 minutes from start to finish, and I'm usually still asleep for most of it.
The software also gets better on its own, every day, without me touching the underlying logic.
Every evening, the full results from that day go back in.
Winning selections confirm the model's probability estimates were in the right range.
Losing selections get pulled apart to work out which variables were over-weighted for that specific race context.
None of these adjustments are dramatic on any given day, and that's rather the point.
It's a nudge here and a fractional re-weighting there, the sort of small correction that would be invisible over a week and is unmistakable over a year.
Applied every day since the first live morning, that compounding improvement is the reason the 2026 numbers are running ahead of the ones that came before them.
There's one safeguard I added after a nervous week in the autumn of 2025.
A model that learns every day can also learn the wrong thing from a freak day, so the lodger now keeps the previous version of the model on standby.
Every Sunday night, the new version and the old one are both tested on the past fortnight's races, and the new one only takes over if it's earned the job.
Most weeks it does, and some weeks it doesn't, and that quiet Sunday contest has stopped more than one bad habit before it could take hold.
The core logic behind the selections is this.
I stopped trying to find horses I thought would win.
I started finding horses whose true chance of winning sits meaningfully above what the published odds are quietly implying.
Let me show you the maths, as it's simpler than it sounds and it fits into a single paragraph.
Odds are just a probability in disguise.
A price of 3/1 is the bookmaker telling you a horse has roughly a 25% chance of winning, because at 3/1 a horse needs to win one race in every four for you to break even.
So if my model has done the work and rates that horse's true chance at 25%, and the market is offering me 9/2, I'm being paid a 25% chance at an 18% price.
That gap isn't a feeling or a fancy; it's a misprint, and backing enough misprints is how you end a month in front instead of behind.
Model
3/1
Market
9/2
A horse with a 25% chance of winning should be priced around 3/1.
If the market has drifted it out to 9/2 or 5/1, that runner is mispriced, and that's precisely the kind of horse the software is built to find.
The horse doesn't have to win that specific race for the bet to be the right decision.
It just needs to win often enough, at prices that reflect the gap between the model's probability and the bookmaker's price, across a large enough sample to produce a steady profit.
That's the same logic the bookmakers use against you, taking a margin by being right about probabilities slightly more often than not, across a colossal number of bets.
The software turns that logic around, hunting for the runners where the bookmaker's own assumptions have slipped.
But they win often enough, and at prices big enough, to build a real profit month after month.
It also means the software never chases a loss, since it has no idea what a loss feels like; it only knows whether the next price is right or wrong.
Just as important as what the software looks for is the list of things it refuses to touch.
It won't back a horse in a race where too much information is missing, such as a field full of unraced two-year-olds with nothing but pedigree to go on.
It won't go near each-way terms, forecasts, accumulators or anything with more than one leg to it.
It won't send more than a handful of selections on any one day, however busy the card, as a longer list would only dilute the best of them.
And it won't touch a race where its own confidence in the numbers is low, even when the price looks tempting.
That last rule cost me a few winners in the early days, and I grumbled about each one of them.
It also saved me from far more losers than it cost me winners, which is the only sum that matters in the end.
The first live morning was the 5th of June 2025.
Before that, the model had spent 3 months running in shadow mode, logging every selection it would have made without a penny of real money involved.
Shadow mode very nearly ended in disaster, and I've never told anyone outside my family about this part.
In the April, one of the data feeds changed the way it listed non-runners without any warning, and for 9 days the model happily rated horses that weren't even going to line up.
Nobody lost a penny, as it was all on paper, but it frightened the life out of me.
That's where the 7:30am second pass came from, and it has checked every declaration against a second source every morning since.
Those shadow results gave me confidence, but a paper trade and a real stake are two very different animals, and I felt the difference in my stomach.
I placed real bets on the model's first two live selections before breakfast, then went to work and tried very hard not to think about it.
Day one profit: £290.
I didn't celebrate, not even a little bit.
Two winners on day one proves absolutely nothing, and I'd have told anyone else the same.
What matters is what a system does over hundreds of bets across many months.
But the feeling that afternoon was different to anything I'd experienced from betting before.
It was more like watching something you'd engineered carefully finally work the way it was designed to.
By the end of that first June, 13 of the model's 21 selections had won, and £4,210 of profit had landed.
That was more than I'd made from betting in any full year before I built it.
Then came the month that made it real to my family rather than just to me.
August 2025 stepped things up hard, with 26 winners from 40 selections and £5,940 in the black.
That was the month I pulled £5,000 out of my betting accounts, cleared the last of a credit card I'd been dragging behind me for years, and took my wife out for the dinner we'd been putting off for far too long.
She raised a glass, looked at me across the table and said the words I'd been waiting the best part of 25 years to hear about the horses: "So it's actually working, then."
It's still the nicest thing anyone has ever said to me about betting, and she said it in the tone she normally saves for the day the washing machine gets fixed.
The momentum carried on through the winter and spring that followed.
September 2025 turned out to be the quietest month of the whole run at £3,620, and even that covered a family week in Pembrokeshire with plenty left over.
October brought in £5,840, and November went better again at £6,380, the best month the software had produced at that point.
There were quieter spells too, weeks where the selections didn't land at the same rate.
But the model soaked those patches up the way a sound process should, running to the same criteria whatever the recent results, and feeding every losing selection back in as fresh training data.
December was a good example, with frozen tracks and abandoned meetings thinning out the cards, and it still finished at £4,120.
Not a single month across that stretch finished at a loss.
And then the year I'd been quietly waiting for arrived, because a machine learning model is only as good as the amount of real history it's been able to chew through, and by 2026 it had chewed through plenty.
January 2026 opened with £6,720, a new record at the time, and most of it came from the all-weather tracks while the turf was waterlogged.
That was the month the separate all-weather dialects I mentioned earlier really paid for themselves, since winter cards are packed with races the public barely looks at.
February followed with £5,280, from a shorter month with fewer selections, and still comfortably ahead of anything I'd ever managed on my own.
March was the quietest month of 2026 at £3,850, and I can tell you precisely why.
It's festival month, when the big jumps meetings pull enormous amounts of public money into a small number of races, and the prices in those races get picked over by everyone.
The model sent fewer selections than usual and patiently waited for the ordinary midweek cards to come back round.
I'll take a £3,850 month in which the software knew when to keep quiet over a flashy week that hands it all back the week after.
May was the less spectacular of the two, and here's the part that matters: the profit still landed, because the model wasn't chasing winners, it was backing the right prices.
Spring is also when the model has to relearn the turf, as horses come back from winter breaks carrying unknowns, and that's when the trainer-by-trainer comeback habits earn their keep.
In April 2026 I dropped to 4 days a week at work, which is the first time in my adult life the horses have given me time back rather than taken it away.
Then July 2026 delivered the best single month since the software went live, with 23 winners from 33 selections and £7,080 in profit.
By that point, the model had digested another full year of live results on top of everything that came before, and the jump in selection quality showed up plainly in the numbers.
August kept the run going at £5,960, and that's the £5,960.40 I told you about at the very top of this page.
September 2026 kept it going with another £6,460, making it the fourth month of 2026 to clear £6,000.
That's six months in a row, April through to September, every one of them in profit, with the two quieter ones still clearing £4,300 apiece.
That isn't a hot streak, and I've had enough hot streaks over 25 years to know the difference.
Here's the comparison I find most telling of all.
June, July and August 2025, the first three live months, brought in around £13,850 between them.
It was the same summer racing and the same 5 minutes of my morning, yet it produced roughly £5,650 more in profit, purely because the software had another year of experience behind it.
Add September's £6,460 and the first 9 months of 2026 come to £50,910 between them.
By the middle of 2026, I'd stopped being surprised by the monthly figures.
Not because the numbers had become ordinary, but because the process behind them had become so reliable that a profitable month felt like a confirmation rather than a result.
By then the model had well over 350 live selections behind it.
Each one had fed back into the training data.
Each losing selection had adjusted a weighting somewhere in the model, marginally, but cumulatively in a way that was showing up clearly in the outputs.
From September to December 2025, the software averaged £4,990 a month.
From January to September 2026, it averaged just over £5,650 a month, and that's with festival March dragging the figure down.
At £50 a bet, that's close to £670 extra every month for doing the same 5 minutes of work.
That's the compounding effect of a model that keeps learning.
Here's the full 12-month picture in plain figures.
From September 2025 to August 2026, 222 of the software's 367 selections won, and the total profit came to £64,410 at £50 flat stakes.
The best month was July 2026 at £7,080, and the quietest was September 2025 at £3,620.
Even the two quietest months, September 2025 and March 2026, still came in above £3,600.
The 2026 months are the strongest the software has produced since it first went live.
That's what happens when a machine learning model absorbs a full year of live results and keeps refining its approach from them.
Winners 222
Losers 145
Let me put those figures into practical terms for a moment.
At £50 flat stakes per selection, which is what every figure on this page is based on, the softest month in the last 12 (September 2025 at £3,620) still represents a profit of more than 72 points.
The best month, July 2026 at £7,080, represents a profit of more than 141 points.
The average across the 12 months comes to just over £5,367 per month at those stakes.
The same percentage returns apply at any stake level.
At £10 per bet the numbers scale down proportionally, so that softest month would have come to about £724 and the best to around £1,416.
At £100 per bet they scale up, and the same two months come out at £7,240 and £14,160.
The point is that the advantage the software produces is steady and measurable, and it never relies on backing long shots or stacking up an accumulator that has to land perfectly.
These are straightforward win-only bets on UK horse races, placed one at a time, with flat stakes.
It's the kind of betting that doesn't need a complicated staking plan or a large bank to get started.
If you're starting small, a bank of 30 or 40 stakes is plenty, and you can raise your stake as the bank grows rather than all at once.
There are dozens of tipping services running right now, and most of them won't be here in 18 months.
Not always because of any bad intent, but because the model they're built on has a ceiling, and most reach it quickly.
The first problem is the information they're working from.
Most tipsters lean on the same public form data every punter can pull up for free, along with the same newspaper analysis and stable whispers available to anyone with a phone.
That information is already reflected in the price before the tipster has finished reading it.
The real pricing mistakes live in the overlap between less obvious data points that nobody's looking at together.
A person working by hand simply can't juggle enough variables at once to keep finding those combinations.
A tipster in the middle of a winning run feels invincible, backing selections they might have hesitated over a month earlier.
The same tipster after a rough fortnight starts second-guessing their own method, bolting on conditions that conveniently explain the recent losses and quietly skipping selections they should be backing.
They call it refining the method.
It's doubt getting into the decision-making.
My software doesn't have bad weeks in that sense.
It doesn't wake up on a grey Thursday morning and decide it doesn't fancy the card.
The same process runs at the same standard every single morning without exception, regardless of what happened the day before.
A standard tipster sends a selection out, it loses, and they move straight on to the next one.
Nothing about that lost bet feeds back into how the next selection gets made.
Every losing selection in my system gets analysed before the following morning.
The model identifies which variables it overweighted for that race context and adjusts the weighting accordingly before it runs again the next day.
Over time, the losing selections actually make the whole system stronger.
No tipster working manually can replicate that, however experienced they are.
The moment a successful tipping service grows its subscriber base significantly, the collective weight of money following the same selections starts to move the market price before everyone can get their bets placed.
The very mistake that made the selection worth taking gets corrected before most of the members can use it.
That's a structural ceiling every successful tipster eventually runs into.
Keeping the group deliberately small is the only way to protect that advantage for everyone inside it.
That's why there are only 75 places in Race Logic Pro, and it'll stay that way.
People often ask me whether the bookmakers will eventually work out what the software is doing and close the gap.
It's a fair question.
The bookmakers' own pricing tools do improve over time, and they'll keep improving.
My model improves too, and it improves specifically in response to the same market conditions the bookmakers are creating.
Every day the results go back in, including the prices that were available, how they moved, and how the selections performed against those prices.
The model is learning from live market conditions in real time, not from static historical data.
That's why the gap between the software's probability estimates and the bookmakers' published prices hasn't narrowed since the system went live.
If anything, the 2026 figures suggest it's widened slightly as the model has matured.
There's also a structural reason this advantage is more durable than most people assume.
The bookmakers price thousands of events a day across multiple sports.
Their pricing resources are spread across an enormous range of markets.
That's an extremely narrow specialism next to everything the bookmakers' systems are asked to cover.
It means the model can go deeper into racing-specific data patterns than any general-purpose pricing algorithm the bookmakers operate.
The bookmakers are building tools designed to manage exposure across millions of customers and thousands of events.
Race Logic Pro is built to find two or three specific horses per day where their margin assumptions have slipped.
Those are completely different objectives, and that's why the two systems aren't really in direct competition with each other in the way people imagine.
The bookmakers want volume above all else.
I want precision, and precision is a far easier thing to win at when you only have one pool to fish in.
There's one more thing worth understanding before I explain how to join.
This isn't a system that works in a bull market and falls apart the moment conditions change.
The last 12 months cover a wide spread of racing conditions, from summer Flat through to the depths of the winter jumps, on ground running from firm to heavy.
The model produced a profit in every one of those months, not because the selections only came from the most favourable race types, but because the software adapts its variable weighting to the specific conditions of each race.
It doesn't have a strong month in summer and a weak one in winter.
It processes the available data for whatever races are on the card that morning and finds the clearest pricing mistakes within them.
That steadiness across every kind of condition is what makes the 12-month record actually mean something, rather than being a run of results from a narrow window of good weather.
In February 2026, I let the first handful of people outside my family see the selections.
It was 12 people to begin with, mostly old colleagues and friends from the pub quiz team, and one neighbour who'd spotted the lodger's lights blinking through the window at 5am.
I wanted to know whether ordinary people, with jobs and children and patchy phone signal, could follow the selections as easily as I could.
The answer was mostly yes, with one important exception.
Some of them were getting to the bets an hour or two after the email, by which time a price or two had shortened and the mistake had been corrected.
That's why every selection now carries the lowest price worth taking, so nobody ends up backing a horse after the reason for backing it has gone.
The group grew slowly through the spring and summer, by invitation only, and its feedback shaped almost everything about the email you'd receive.
One of them, a retired lorry driver, asked me to make the horse's name bigger, as he reads the email on his phone without his glasses.
Here's how this works for you on a practical, day-to-day basis.
Every morning, before the first race of the day goes off, the model works through the full declared card.
Every runner at every meeting gets assessed against hundreds of weighted variables.
The selections where the model's calculated win probability sits meaningfully above the available market price are flagged and packaged into a short email.
You open it, place the bets at your usual bookmaker, and get on with your day.
The whole thing takes around 5 minutes on your side.
You can do it from your phone before you leave the house.
Then you check the results later when you've got a minute.
There are no form guides to plough through and no statistics to decode, because the model has done the heavy lifting hours before you're even awake.
Each email is laid out the same way every day, with the course, the race time, the horse and the lowest price worth taking.
That last figure matters, because if the price has shortened below it by the time you get there, the mistake has been corrected and you just leave that one alone.
It won't require you to have...
Technical ability of any kind (all you do is copy the selections and place the bets).
Significant free time (5 minutes in the morning is the full commitment).
Prior racing knowledge or experience (the model has assessed everything there is to assess).
A large betting bank (you can start at £2 or £5 a bet while you get comfortable).
There's no learning curve, and nothing to buy beyond a phone and a betting account you almost certainly already have.
Here's what four more members told me after following the bets through the summer of 2026.
"I'm a postman, so I'm out of the door at half five, and I place the bets from the van at my first tea stop. By the end of July I'd cleared £2,380, and the lads at the depot have stopped taking the mickey and started asking me for the link."
Barry Winstanley, Wigan
"As a retired maths teacher, I tested the prices before I trusted the profit, checking every morning's odds against the results for a full month. The numbers held up, and so did the £1,960 sitting in my account at the end of it."
Maureen Dobson, Hartlepool
"The best thing I can say is that I don't chase any more, because there's nothing left to chase once the day's bets are already decided. I finished the summer £2,470 ahead, and I haven't placed a single bet of my own since June."
Alison Thwaite, Scunthorpe
"My brother swore blind it was too good to be true, so I followed the selections at £2 a bet just to prove him wrong. I've since moved up to £20 a bet and bought him a pint out of the profits, which he drank without saying a word."
Phil Easterby, Taunton
Before I let a single person outside my house near the software, I let my brother-in-law Keith have a go at it.
Keith spent 30 years as an auditor, he's never placed a bet in his life, and he regards the whole racing industry the way a vicar regards a nightclub.
Over one very long Sunday lunch, he fired every awkward question he could think of across the table, and I've kept the best of them here as close to word for word as I can remember.
"If this thing is so clever, why on earth are you selling it for the price of a takeaway?"
Because the membership was never meant to be where I make my money, Keith; the bets are.
A small, paid group keeps out the tyre-kickers and covers the running costs of the lodger, and that's all I need it to do.
No, any standard UK betting account will do, and members place the bets at whatever prices they can get that morning.
"What happens when the price you've sent has already shortened by the time somebody gets to it?"
Every selection comes with the lowest price worth taking, and if the market has gone below that, you leave it alone.
Missing a bet costs you nothing, whereas backing a mistake after it's been corrected costs you slowly, and the email is set up to protect you from that very thing.
They can, but I'd rather they didn't, since the profit comes from the whole run of selections rather than from anyone's favourite bits.
That's the exact habit that cost me 16 years of losses, so I'd hate to see it creep back in through the side door.
"What happens on a day when it gets something badly wrong?"
It gets individual races wrong all the time, as a losing bet is nothing more than a race that went the other way.
The difference is that every one of those losses is taken apart that night and fed back in, so the same mistake gets harder to make twice.
"Won't the bookmakers close the account of everyone who follows it?"
Two or three sensible win-only bets a day at flat stakes, spread across ordinary races, looks nothing like the pattern that gets accounts flagged.
Keeping the group to 75 people keeps the whole thing quiet, which protects your account and the prices at the same time.
Not at all, and some of the best-performing members I have couldn't tell a hurdle from a fence.
The email tells you the horse and the price, and that's the full extent of the racing knowledge required.
"What if there's nothing worth backing on a given day?"
Then nothing goes out except a short note saying so, because a day with no selection is a decision the model has made, not a day off.
"And what if it all stops working one day?"
Then I'll be the first person to know, since my own money goes on every selection before anyone else's does.
Keith didn't have an answer to that one, and he's been quietly following the selections himself since the spring.
For the first time since I started running Race Logic Pro privately with a small group, I'm opening 75 places to the public.
That's 75 spots, and not one more than that.
When too many people back the same selections at the same time, the collective weight of money going on those horses starts to shift the market price before everyone can get their bets placed.
The mistake that made the selection worth taking starts to close before it's been properly used.
By keeping membership at 75, the odds available when I send the selections out each morning are still there when you log in to place your bets.
It's the only way to protect the advantage that produced the last 12 months of results.
If you're reading this right now, there's a decent chance one of those spots is still available.
But I'd strongly suggest not relying on it being there if you come back later.
It's not my intention to make a profit selling memberships.
I make my money from the bets, the same as you will.
So the price for lifetime access to Race Logic Pro is a one-time payment of £20.
One single payment of £20 today, and nothing more to pay.
No recurring monthly fees.
No hidden charges tucked away anywhere either.
For context, that's less than half of one £50 stake at the level every figure on this page is based on.
And if you're on the fence about whether this is right for you, here's what backs it up.
Join today and follow my bets for up to 30 days.
If for any reason at all you're not satisfied within that window, send me one message and I'll return every penny of your £20 immediately.
No conditions, and no questions asked.
You can follow the bets at £1 a time if you want.
You can even write them down and track them on paper with no money on the line at all, just to watch the results play out in real time.
Do whatever it takes to be certain this is real before you commit to anything.
You have nothing to lose and everything to gain.
Thank you for taking the time to read through this.
The 75 spots will fill quickly.
If you're here now, take your place before someone else does.
I'll see you inside.
Stuart Taylor
The model runs every morning regardless of what's on the card.
Whether it's a busy Saturday with seven meetings and 50-plus races, or a quiet Tuesday with three meetings and a thin card, it processes everything available and only sends a selection where the pricing mistake is wide enough to act on.
On days where nothing clears the threshold, no selection goes out and I'll say so clearly in the email.
Quality over volume is the principle it runs on, and it's the reason each of the last 12 months has finished in profit.
The lodger will be humming under my stairs again at 5am tomorrow, whether you join or not, and I'd much rather its work landed in your inbox as well as mine.
The daily email goes out before 8am every day, and if you join now, today's selections will already be waiting for you in the members area.
P.P.S. Right now, the software is running better than it has at any point since the day it went live.
The figures on this page show you what the last 12 months have produced, and with a 30-day money-back guarantee covering your whole membership, there's nothing to lose by finding out what the next 12 could do for you.