How to Evaluate a Sports Picks Service

Handicapper is the oldest of the three. It described the person who set the handicap — who worked out how much better one side was and priced the difference. It now generally means anyone who analyses games and publishes a verdict on them.

Capper is the same word, shortened, and it is what the industry mostly calls itself. Tout is the same job again with a harder edge: someone selling picks, used approvingly by sellers and sceptically by everyone else.

None of the three is a credential. There is no exam, no register and nothing to revoke — anyone can use any of them from their first day, and many do. So the words tell you what somebody does, and nothing whatsoever about whether they are good at it.

Which is why “best sports picks” has no answer you can look up. There is no ranked list, because there is no body producing one and nothing being measured. What there is instead is a test you can apply yourself, and the rest of this page is that test.

The difficulty is not that people lie, though some do. It is that betting results are so noisy that an honest person can sincerely believe they have an edge they do not have — and produce a real, unfalsified record that appears to confirm it.

A strong professional edge is around three percentage points of win rate. The swing between one run of a hundred bets and the next is many times that. So over any period a customer would consider “a while”, the noise is louder than the signal, and a good record and a lucky one look identical.

Which means the right question is never “is this record impressive”. It is “could this record have happened by chance” — and that has a numerical answer.

Every record has a confidence interval — the range of true win rates consistent with what was observed. Until that interval clears the break-even rate of the prices used, the record is compatible with having no edge at all.

A 30-20 record — the kind that fills promotional material — is consistent with a true win rate of 46%, which loses money steadily. The service advertising it is not necessarily dishonest. The number simply does not carry the weight being put on it.

Run any record you are shown through the win rate confidence calculator. It takes ten seconds and it answers the only question that matters.

Here is the part almost nobody applies, and it is the most useful thing on this page.

The bigger the claimed edge, the fewer bets it takes to prove. A small edge is hard to demonstrate because it hides in the noise. A large one is easy, because it does not.

So a service claiming 65% needs only 238 graded bets to establish it beyond reasonable doubt. That is a few months of ordinary volume.

Which produces a clean test. If a service has been operating for years, claims a very high win rate, and still cannot show a complete graded record long enough to demonstrate it, the claim is false. Not unproven — false. They have had the sample. Publishing it would settle the question permanently, and they have not, and the only reason not to publish a record that would vindicate you is that it would not.

Take a thousand people with no skill whatsoever and have each bet ten coin flips.

P(9-1 or better in 10)  =  1.07%   →  about 11 of the 1,000
P(a perfect 10-0)       =  0.098%  →  about 1 of the 1,000

Eleven of those thousand people now have a screenshot of a 9-1 run. One has a perfect record. None of them did anything, and none of them know that — from the inside, an excellent run of pure chance feels exactly like being good at something.

Now let those eleven advertise, and the ones who go on to lose quietly stop. The industry does not need liars to produce a steady supply of spectacular records. It only needs enough people trying and a mechanism that shows you the survivors.

This is why an individual winning slip is worth nothing as evidence, no matter how large the payout or how genuine the image. The question is never whether a bet won. It is what happened to all the bets.

• How many graded bets, in total? If the answer is not a specific number, there is no record. If it is under a few hundred, it cannot demonstrate anything either way.
• Were the losses published before the results were known? This is the load-bearing question. A record assembled afterwards is a selection, and selection is precisely the mechanism that manufactures impressive records from nothing.
• At what prices? A 55% record is a real edge at −110 and a losing operation at −140, which needs 58.33%. A record without prices cannot be evaluated at all — check against the break-even rate of the odds actually used.
• Is the ROI on turnover or on bankroll? The same season is 5% on turnover and 50% on bankroll. Only the first is what ROI normally means, and the second is the one that appears in marketing. The ROI calculator uses turnover.
• Are the graded prices the ones subscribers could actually get? A record graded at the best number available anywhere, or at a price that had already moved by the time the pick was sent, is not a record of what a customer would have achieved.
• Does the record start at the beginning? A track record that begins on the date the winning started is not a track record.
• Can you verify it independently? Timestamped posts, a third-party tracker, or an append-only public log beat any screenshot.

Every question above, with what a good answer looks like beside the answer that should stop you. It works on us as well as on anyone else, which is the only way to know a checklist is honest.

• “Lock”, “guaranteed”, “can’t lose”. Every bet can lose. Anyone using these words is telling you something about themselves, not about the game.
• A win rate with no bet count. Not a claim. A decoration.
• Bet slips as the primary evidence. See above — eleven of every thousand no-skill bettors have a great one.
• Guarantees of profit. Nobody can guarantee an outcome they do not control. A refund policy is a commercial term; it is not evidence of an edge.
• Urgency and scarcity. A real edge does not expire in the next four minutes, and a genuine one is generally harmed by more people betting it, not helped.
• Never a losing week. This is not a boast, it is an admission. A winning bettor at −110 hits a five-loss run about 74% of the time within a hundred bets. Somebody who has never shown one is not showing you everything.

It would be dishonest to publish this and exempt ourselves, so: apply every question above to ProPickz.

Our record is public, it includes the losses, picks are graded at the price they were posted at, and nothing is deleted once it goes up. Where a page is running on sample rather than live data, it says so on the page. Our editorial policy states what we will not publish, and fabricated testimonials and invented rating markup are on that list.

What we cannot claim is a sample large enough to have proven an edge statistically, because — as the table above shows — almost nobody in this industry has one, and saying otherwise would fail our own test. The honest position is that the record is public and still accumulating, and you should judge it with the same confidence interval you would apply to anyone else’s.

That is the whole argument for a public ledger. It does not ask you to believe anything; it lets the sample grow in the open where you can check it.