You should not, on our say-so. Nobody should take a betting site's word for its own figures, including this one.
The specific worry people raise is cherry-picking: that a site publishing research about betting will show the runs that flattered it and quietly lose the rest. It is the right thing to be suspicious of, because it is exactly what the incentives push towards.
So rather than assert that we do not do it, here is how the numbers are made and what you could check.
Every bet is logged before the race. Our model's selections are published each morning and settled afterwards. The tipster league's picks are timestamped before the off. A selection cannot be added later, and one that loses cannot be removed, because the daily files are public and dated.
The sample is everything, not a window. Our track record covers 10,478 races and 10,297 bets, with no start date chosen to flatter it. The betting systems board covers 30,082 races. Choosing a good month is easy; choosing a good three years is not available to us.
The losers are published as prominently as anything else. The headline on our systems board is that 0 of 24 approaches make money, and the best still loses -5p in the pound. Our own model returns −14.4%. Those are the numbers we lead with.
A control runs alongside. The tipster league includes a baseline that backs the favourite and nothing else, and it is currently beating most of the AI models. We could have left it out.
None of that is proof. It is a set of arrangements that would be awkward to fake, which is the most any publisher can honestly offer.
The better test of a site is what it does when it gets something wrong.
We built a page showing a fair price next to a live best price and did not notice, for a while, that the two were computed from markets nine hours apart. It made the comparison meaningless. A reader pointed at it and we rewrote how the columns work.
We ran a racecards page that served a two-day-old card under the current date, for several hours, because a cache had pinned an unusually small day's data. It said racing had finished while five meetings were still to run. We wrote the fix, published what had happened, and built an hourly check that now grades those pages against the racing feed.
We have also had to retire figures. When a betting-system recompute produced a positive return that turned out to rest on information only available after the race, we quarantined it rather than publish it.
We would rather list those than claim a clean record. A site with no corrections is not a careful site, it is a site that is not looking.
If nothing works, how do some people make a living at this? That question does have an answer, and it is not about picking more winners.
None of these pages is a tip, and none describes a way to win. They describe what betting costs, which is a different and more reliable subject. If your betting has stopped being fun, BeGambleAware has free, confidential help, and the National Gambling Helpline is on 0808 8020 133.
The samples are everything we hold rather than a chosen window: 10,478 races for the model, 30,082 for the systems board. Selections are logged before races run and settled in public.
Yes, and they are the headline. Our model returns −14.4% and 0 of 24 tested approaches make money.
The site carries affiliate links. Nothing on the research pages is ranked or reported by who pays us, and the figures here are measurements of the market rather than reviews of firms.
The daily picks and results are published with dates, and the methodology pages set out how each figure is computed. The underlying convention for margins is stated on every page that uses one.
Yes, more than once, including a comparison built on stale prices and a positive system result that depended on information only available after the race.
Every figure here is pulled live from our data and nothing beats the bookmaker's margin. For whether anyone holds a real edge, see our track record. 18+, please bet responsibly.