How to Tell If a Tipster Is Actually Good: Sample Size, Yield and Luck

You've seen the screenshot. A football tipster, 100 tips at evens, 55 winners, and a profit line marching up to the top right of the chart. Ten points up, a 10% yield, and a caption telling you the paid group opens on Friday.

It looks like skill. It might even be skill. But hand 100 people a coin each, ask them to call 100 flips at evens, and the maths says around 18 of them would match that record or beat it. Nearly one in five, and not a shred of talent between them.

That cuts both ways, and it's why this is worth learning properly. Most hot records are luck, yet genuine edges do exist, and the numbers will lead you to them. You just need to know which figures to ask for, how many bets it takes before they mean anything, and which tricks to see straight through.

Get this right and you'll never pay for a lucky month again.

The UK press has tracked tipsters the same way for years: assume a level £1 stake on every tip and follow the running profit or loss. It's a sound convention because it strips out staking tricks and leaves only the selections. From there, three numbers do the heavy lifting.

Profit is total returns minus total stakes, in units. It's the headline everyone shouts about, and on its own it tells you very little, because it depends on how many bets were placed. Twenty points from 50 tips and 20 points from 2,000 tips are completely different stories.

Yield (or ROI per bet) is profit divided by total staked. With flat one-unit stakes that's simply profit divided by the number of bets, so +18 units from 400 bets is a 4.5% yield. This is the number you compare tipsters on.

Strike rate is winners divided by bets, and it's the most abused stat in tipping. A 60% strike rate sounds brilliant until you learn the average price was 1.50, where you need 66.7% just to break even. A 20% strike rate sounds dreadful until you see the average price was 5.00, where 20% is exactly break-even.

Average odds (decimal)	Fractional	Break-even strike rate
1.50	1/2	66.7%
1.91	10/11	52.4%
2.00	evens	50.0%
3.00	2/1	33.3%
5.00	4/1	20.0%
10.00	9/1	10.0%

Break-even strike rate = 1 ÷ average decimal odds.

Every tipster starts with a mountain to climb. First they have to beat the margin the bookmaker builds into every price, then find a bit of extra edge on top. That edge is small, and over a short run it gets buried under whatever football decides to do on a given Saturday.

For a single one-unit bet at decimal odds O, where p is the true chance of winning, the maths looks like this:

true yield        y = p × O − 1
spread of one bet s = O × √( p × (1 − p) )

after n bets:
standard error of measured yield = s ÷ √n

Think of the standard error as the wobble luck puts on a record. Roughly 95% of the time, a tipster's measured yield lands within two standard errors either side of their true yield. Plug in a tipster with zero edge and watch what happens:

Average odds 2.00  →  SE after 100 bets = 10%   after 1,000 = 3.2%   after 10,000 = 1.0%
Average odds 5.00  →  SE after 100 bets = 20%   after 1,000 = 6.3%   after 10,000 = 2.0%
Average odds 10.00 →  SE after 100 bets = 30%   after 1,000 = 9.5%   after 10,000 = 3.0%

Go back to that first line. After 100 bets at evens, a tipster with no edge at all can routinely show anything from about −20% to +20%. At average odds of 5.00 the range stretches to roughly −40% to +40%. A +15% yield over 100 tips at those prices tells you next to nothing about whether someone can pick a winner.

The binomial maths agrees. Our 55-from-100 tipster at evens? A pure coin-flipper does that or better 18.4% of the time. Push it to 60 winners from 100 (+20% yield) and luck still gets there 2.8% of the time. Put a 95% range around that 55% strike rate and you get 45.2% to 64.4%, which comfortably includes a losing tipster betting at 1.91.

So how big does the sample need to be? Two things decide it: the size of the edge and the length of the prices. Small edges hide in the noise for longer, and bigger prices make the noise louder.

bets needed  n = ( z × s ÷ y )²

z = 1.96  →  the record sits two standard errors above zero
z = 2.80  →  an 80% chance a genuine edge looks convincing

Here's what that gives for some realistic tipster profiles. The figures come from a normal approximation, which gets rougher at long odds, so treat them as a guide to scale rather than exact counts.

Average odds	True yield	Bets to reach 2 SE	Bets for an 80% chance
1.50	5%	~730	~1,480
2.00	5%	~1,530	~3,130
2.00	10%	~380	~780
3.00	5%	~3,150	~6,430
5.00	5%	~6,370	~13,020
10.00	5%	~14,440	~29,500

The line to remember: a genuinely good +5% tipster at evens needs around 1,500 settled bets just to poke their head two standard errors above zero, and even then a real edge only clears that bar about half the time. At average odds of 5.00 you're looking at around 6,400 bets.

Now put that in football terms. A Premier League season has 380 matches. A tipster who bet on every single one at evens would need four seasons to reach 1,500 bets, and nobody bets every match. For a selective football tipster, proof takes many seasons. One hot October doesn't come close.

Martin Gardner was writing about ESP experiments, not betting, but he nailed the problem. Run enough experiments and someone will always score improbably well. In his hundred-experimenters story, one researcher ends up with a star subject and has no idea the other ninety-nine projects found nothing.

Swap "experimenters" for "tipsters on X" and you've got modern tipping. This is survivorship bias: you only ever see the ones who made it through. The tipsters who went 40 points down closed their accounts, deleted the tweets or quietly rebranded. Whoever's left standing looks like a talented elite.

We simulated 1,000 pure coin-flip tipsters, each placing 100 level-stake bets at evens with zero edge between them. About one in five finished 10 units up or better. Around 3% made 20 units or more. The best of the lot finished roughly 38 units up, a +38% yield built on thin air.

Run it again at odds of 5.00 and it gets wilder. Around 45% of the coin-flippers were in profit after 100 bets, and the luckiest posted a yield of around +75%. That looks superhuman. It's really just the jammiest draw from a big crowd.

Fund managers know this trap well. Fund tables that quietly drop closed funds overstate returns, and Elton, Gruber and Blake put the effect at 0.9% a year in US funds back in 1996. A tipster who deletes losing weeks is pulling exactly the same stunt. Scan 100 zero-skill tipsters for anyone whose record passes a "two standard errors" test and you should expect two or three to pass by luck alone.

This is the bit that should worry anyone following a tipster who loves an outsider. Take a tipster with a true yield of −5%, roughly what the bookmaker's margin costs you when you have no edge at all. What are the chances they're showing a profit anyway?

Average odds	Chance of profit after 100 bets	After 1,000 bets
2.00	30.8%	5.7%
5.00	39.9%	21.0%
10.00	43.2%	29.5%

At evens, the losing tipster gets found out fairly quickly. At average odds of 10.00, nearly three in ten are still in profit after 1,000 bets. For most people that's years of tips, so a long-shot specialist can run a healthy-looking record for a very long time with no edge whatsoever.

That doesn't make big-price tipsters bad. Their records just need far more bets before you take them seriously, because one 33/1 winner can carry a mediocre record for months. Our guide to form, variance and luck in betting shows how wild those swings get, and which sports are the most unpredictable shows how much of any football record is simply noise.

It also explains why last month's star is so often this month's also-ran. The bigger the luck component in a hot run, the harder it falls back. It's the same regression to the mean behind the Sports Illustrated cover jinx and the sophomore slump.

Most dodgy tipster operations give themselves away if you know where to look. Reputable services keep an accurate record precisely so you can judge their form. The rest have every incentive to massage the figures or invent them outright, and back in 2008 the Office of Fair Trading put UK losses to tipster scams at between £1 million and £4 million a year.

Derren Brown showed how the oldest trick works in his Channel 4 programme The System. Give different horses to different people and someone always backs the winner. That person only sees their own lucky streak, assumes there's real insight behind it and pays for more. Any channel that only ever posts winners is running a version of the same con.

Walk away when you see:

• No settled, timestamped record. Screenshots of winning slips are not a record.
• Tips posted after the event, edited posts, or "we were on this" messages after the final whistle.
• Prices you couldn't get. Odds from one obscure book, stale lines, or prices that vanished in seconds.
• Profit with no context. "+120 points this month" with no number of bets, average odds or stake sizes.
• Varying stakes in "points", which can make a chasing strategy look brilliant until one bad run wipes it out.
• Strike rate without average odds, or a hot streak marketed off a few dozen tips.
• Guarantees, "risk-free", insider or fixed-match claims, and pressure to join a paid channel today.

If you pick up tips on social media, read our guide to using Twitter, Reddit and Discord tips responsibly alongside this one.

A credible record is gloriously boring. It's a full ledger, not a highlights reel. Every tip goes in before kick-off with a timestamp, the market, the price, the stake and where that price was available, and nothing gets touched afterwards.

It shows everything: total bets, average odds, units staked, yield, strike rate against the break-even rate for those prices, and the worst drawdown along the way. A good tipster will happily tell you their longest losing run. Anyone with a real edge at decent prices has had one.

Then there's volume. We want 1,000+ settled bets before calling anything evidence, and more for long-priced tipsters, with the standard error at that sample in mind. Even then, a big sample tells you a record is unlikely to be pure luck. It can't promise the edge survives next season.

Price quality is the quickest tell of all. If a tipster's selections keep shortening after they're posted, so the price taken beats the closing price, that's a far cleaner signal than profit. Our explainer on closing line value shows how to measure it. Closing prices are the market's best estimate of the true chance, and beating them week after week is very hard to do by accident.

Watch for backtests dressed up as results, too. A model fitted to last season's data and shown as "what we would have made" is not a record. Only live, pre-event tips count. Our guide to building a betting model covers why honest backtesting is so hard.

That's exactly why we love a tipping competition. The tips go in before the event, the full historical record is public, and nobody gets to delete the bad weeks.

None of this maths is meant to turn you into a cynic. It's there so you can spot genuine quality and stop paying for luck. Run any tipster through these checks before a penny goes their way.

Our tipster-judging angles:

• Yield first, profit second, strike rate last. Ask for yield and average odds before you look at anything else; a strike rate on its own is noise.
• Under a few hundred tips, it's a trailer, not a film. Enjoy the run, but at evens a zero-edge tipster swings ±20% over 100 bets. Don't pay for it.
• Scale your patience to the prices. A 5% edge at evens needs ~1,500 bets to show; at average odds of 5.00 it needs ~6,400. Long-shot records need years.
• Price beats profit. A tipster who consistently beats the closing price is more convincing than one sitting on a hot month's profit.
• The leaderboard top ten always contains lucky people. That's simply what happens when hundreds of people tip at once. Look for the ones who stay near the top over thousands of bets.
• No ledger, no money. If you can't see every tip, timestamped, with the price, assume the missing ones lost.
• Copy the stakes sensibly. Even a genuine edge comes with long losing runs, so stake flat or a small percentage of your bank; our flat vs percentage staking guide shows how.

Fancy turning your own record into a living? Our honest look at whether you can make a living from betting puts you through every one of these tests. Whoever you follow, pick a staking plan you can live with through a bad month, because even the best records have them.

• Closing line value (CLV) explained
• Form vs variance vs luck in sports betting
• Social media betting tips: Twitter, Reddit and Discord
• Can you make a living from betting?
• Flat betting vs percentage betting

18+. Please gamble responsibly. For free, confidential support visit BeGambleAware.org.