The truth about the markets

Statometrics is not a tipping service. It is an attempt to tell the truth about betting markets: how they work, where the value really is, and where it is not.

We work only on the Betfair Exchange, and only on the markets with real money in them: Match Odds, Correct Score, Over/Under Goals, and First and Second Half Goals. Everything we show is measured on real prices from real matches.

If you are looking for someone to send you three bets a day and tell you they can't lose, this is not the place. If you want to understand why a bet is worth making, or why it isn't, read on.

There are ways to make money from betting. There are ten times as many ways to lose it, and most of them feel like winning while you are doing them.

The hardest ones to spot are the ideas that look brilliant on paper:

• ·A strategy that worked last season. Look back over enough games and you will always find a pattern that made money. Most of those patterns are luck wearing a disguise, and they stop the moment you put real money on them.
• ·A good run. Ten winners in a row feels like proof. Often it is just a coin landing heads ten times. That's about a 1 in 1,000 chance, so across thousands of bettors it happens to dozens of people every week.
• ·A high strike rate. Winning 80% of your bets means nothing if the 20% you lose cost more than the 80% you win. Backing short-priced favourites, or laying big-priced outsiders, wins most of the time, right up until it doesn't.
• ·Tweaking until it fits. Add enough filters to any strategy and you can make its history look perfect. You have not found an edge; you have described the past.

There is a new wave of "AI betting". Sites now promise AI picks that win 95% of the time. That tells you nothing. Anyone can win 95% of their bets by backing 1.03 shots, and still lose money doing it. What matters is whether the prices are worth taking. When you see a claim like that, the question is not how the AI does it. The question is what they are not showing you.

The bigger problem is that the claim misunderstands what AI is. AI cannot win you bets, and it will not. It has no special view of how a match will finish and it cannot see the future any better than you can. Selling AI as a machine that picks winners is not a bold claim about a clever tool. It is a complete misunderstanding of what the tool is and what it does.

What AI really is, is leverage. It takes the knowledge and skill you already have and multiplies it. Someone who understands the markets can use it to learn faster, test ten ideas in the time one used to take, and check their thinking from angles they would never have tried alone. Used well, it can make a skilled person ten times more effective in their own area.

But leverage works both ways, because it multiplies whatever you put in. Give it a clear question, clean data and an honest way of measuring the answer, and it will get you to the truth quickly. Give it a vague hunch, bad data, or a question built to confirm what you already believe, and it will hand you a confident, polished, wrong answer just as quickly. AI and its output are only ever as good as what goes into it.

So the right question is never "does it use AI?" It is "what went in, and how was the answer checked?" That is the question we ask of everything on this site, our own work included.

We embrace AI fully. It is a big part of how Statometrics was built and how it will keep improving. But we use it for what it is genuinely good at, not as a marketing badge. What AI does for us:

• ·It helped build the whole platform: the database of matches and prices, the tools, and the pages you use.
• ·It lets us test far more ideas, far faster, than one person ever could. An idea that would once have taken a week in a spreadsheet takes minutes.
• ·It acts as a second opinion. Machine learning runs in the background, checking our main models. When it disagrees, we look harder.
• ·It is relentless about checking. It will happily run the same test a thousand ways to see if a result holds up or falls apart.

What AI does not do:

• ·It does not pick bets out of a black box that nobody can explain.
• ·It does not make a weak idea strong. If the numbers say an edge isn't there, no amount of AI changes that.

The honest use of AI is to find the truth faster, including the truths you'd rather not hear.

The betting world is full of noise: screenshots of winning slips, tipsters who only post their winners, and systems that are never shown losing. We want Statometrics to be the opposite, a place where the numbers are straight.

That means three things:

• ·We show what works, and what doesn't. Telling you an idea is dead is just as valuable as telling you one is alive. It saves you money.
• ·We show our members everything. When we show results to members, we show all of them, every game, winners and losers. Never a highlight reel, never the best ten.
• ·We explain the market. How prices form, why they move, what a price is really telling you. If you understand the market you can judge any idea for yourself, including ours.

We have moved completely to statistical modelling. Not gut feel, not "this team is due a win", not a pundit's opinion. Every idea is tested against the record before it goes anywhere near a stake.

The most important thing the statistics do is not find winners. It is tell us what is statistically viable and what is not. Most ideas fail that test, and that is the point. A filter that throws out nine ideas to keep one real one is worth far more than a system that keeps all ten.

What that looks like in practice:

• ·A large, real history. Over 107,000 matches with goal timings. Around 55,000 of those, and growing, have the Betfair Exchange prices that were actually traded on them: more than 142 million price points in all.
• ·Only what you could have known. Every figure for a match is worked out from games played before it. No hindsight creeps in.
• ·Models you can read. Our core is built on well-understood methods: team ratings and goal-scoring models. When they make a call, we can say why.
• ·Agreement across independent systems. We trust a call more when separate methods, built different ways, point the same direction.
• ·Beating the closing price, not just the result. The fairest test of a bet is whether you got a better price than the market settled on by kick-off. Results swing with luck; beating the closing price over hundreds of bets does not happen by accident. It is the measure we trust most.
• ·Betfair Exchange only. You bet against other people, not a bookmaker, so winning doesn't get your account closed. All our results include Betfair commission.

Every result we show comes with numbers most betting sites never show you: the number of bets, the t-statistic and the p-value. Together they answer the most important question about any strategy: could this result have happened by luck alone?

Here is why that matters. Take two strategies that have both made a 5% profit. One did it over 20 bets. A couple of lucky winners at a big price would explain that, so it proves almost nothing. The other did it over 2,000 bets, which is far harder to fluke. The profit figure is the same. The statistics are what tell them apart.

What each number tells you:

• ·The number of bets. Betting results swing hard, and more so at bigger prices. A real edge needs hundreds of bets, sometimes thousands, before it shows itself clearly. A figure with no count beside it is not worth reading, so we always show the count.
• ·The t-statistic. It measures how far a result sits from breaking even, allowing for how many bets are behind it and how much the results swing. The bigger it is, the harder the result is to explain by chance. As a rough guide: below 2, we treat a result as unproven. Around 2, it is worth a closer look. Above 3, it starts to be convincing.
• ·The p-value. It asks the same question the other way round: if the strategy had no edge at all, how often would luck alone produce a result this good or better? A p-value of 0.05 means about 1 time in 20. A p-value of 0.001 means about 1 time in 1,000. The smaller it is, the less likely the result is a fluke.

There is one more trap these numbers help with. Test enough ideas and some will pass by luck. At a p-value of 0.05, roughly one in every twenty ideas with no edge at all will still look like a winner. So we never stop at one test. We hold seasons back, run the idea on games it has never seen, and check whether the figure holds up or falls apart. This is called out-of-sample testing: the games used to find an idea are the "in-sample" games, and the ones it is tested on afterwards, which played no part in finding it, are "out of sample". A real edge survives that. A lucky one usually does not.

None of this is there to make the page look clever. It is there so you never mistake a lucky run for an edge, and never throw away a real edge because of a bad month.

Numbers on their own are not much use. A percentage means nothing if you don't know what it was measured against, or how many games are behind it. So we show you the statistics with the understanding that goes with them.

• ·Free to everyone: league and team statistics, calculators, a bet log to record your own bets, and a place to upload your Betfair statements and see your real record.
• ·Members: My Desk (the day's matches with our recorded prices), the Strategy Simulator, team ratings and goal models. Elite adds in-play entry, trading out, and what our own engine finds as it finds it.

The tools matter as much as the signals. The Strategy Simulator lets you put a question to years of real matches and prices and see every game it would have bet on, winners and losers. The Model Tree shows what tends to happen next from any score at any minute. You don't have to take our word for anything; you can check it.

This is not a final model, and it never will be. Markets learn. An edge that works today gets found by others, gets priced in, and fades. Anyone who tells you their system is finished is telling you it will stop working and they won't notice.

So Statometrics is built to keep moving:

• ·Every edge is watched, all the time. Its results are tracked live, bet by bet, not just judged once in a backtest.
• ·We know the difference between a losing run and a dead edge. Every real edge has bad spells. The statistics tell us how bad a run can get by chance alone. Inside that range, we hold our nerve. Outside it, we stop, without sentiment.
• ·New ideas go through the same gate. Nothing reaches members just because it looks good. It has to earn its place.
• ·The tools keep improving. More data, better models, sharper measures. What you see this month will be better next month, and we will tell you what changed.

Betting for profit is a tough business. It is not easy money and it is not a side hustle you can do half-asleep. Nothing on this site changes that.

What it takes:

• ·Focus. Knowing exactly which bets you are making and why, and not drifting into ones you aren't.
• ·Mental clarity. Losing runs will come, even with a real edge. The people who make money are the ones who don't chase losses, don't double stakes to win it back, and don't abandon a sound plan after a bad week.
• ·Patience. Real edges are small. They pay over hundreds of bets, not over a weekend.
• ·Discipline with money. Only stake what you can afford to lose, and size your bets so a bad run can't knock you out.
• ·Honesty with yourself. Record every bet. Most people remember their winners and forget their losers. Your bet log won't.

Statometrics is for over-18s only. If betting stops being fun, or starts costing money you need, stop. Free, confidential support is available 24/7 from the National Gambling Helpline on 0808 8020 133 (gamcare.org.uk), and you can block yourself from gambling sites through GAMSTOP at gamstop.co.uk.

• ·We will never claim a win rate we can't prove.
• ·We will show our members every result, not a selection.
• ·We will tell you when an edge has died, and stop using it.
• ·We will explain the numbers, not just hand them over.
• ·We will never use urgency, countdowns or "last chance" tricks to sell you a seat.

It is for you if you want to understand the markets, you are happy to think in hundreds of bets rather than one, you like to check things for yourself, and you would rather hear an uncomfortable truth than a comfortable lie.

It is not for you if you want guaranteed winners, you want to get rich quickly, or you are looking for someone to blame when a bet loses.

Most of what is sold to bettors is built to look good, not to be good. Statometrics is built the other way round. We would rather show you a small, real edge with every result behind it than a big, shiny number with nothing underneath.

That is the why. Not to tip you winners, but to give you the truth about the markets, the tools to test it, and the honesty to tell you when something has stopped working.

Welcome to Statometrics.