What cherry-picking is
Cherry-picking means showing the numbers that support a story and leaving out the ones that do not. The figures themselves may all be accurate. What misleads is the selection: which period, which players, which results and which measure were chosen to be shown.
In gaming, this appears in advertisements, strategy videos, forum posts, screenshots of big wins and sales pages for “systems”. Because games of chance produce so much random variation, there is nearly always some true-but-unrepresentative result available to quote.
Chance will give someone a great result
The simplest demonstration needs no cheating. Suppose 100 people each flip a fair coin 10 times. The chance that any one person gets 9 or more heads is 11 in 1,024, about 1.1 percent. That is a small chance for one person, but across 100 people the chance that at least one reaches it is roughly two in three. (The calculation is 1 minus 0.989 raised to the power 100, which is about 0.66.)
So if you hear from the person who flipped 9 heads, and not from the other 99, you might conclude they have a talent. The same logic applies to anyone who tries many strategies, many tables or many games and then reports the best one. The winning story is visible; the ordinary ones are quiet.
Common ways numbers get selected
- A convenient time window. “Up 40 percent this month” may follow a bad quarter. A different start or end date can reverse the picture.
- Survivors only. Forums show people who kept playing and posting. Those who lost and left rarely write about it. Reading only the survivors makes success look common.
- No denominator. “Thousands of winners” says little unless you know how many players there were. A big count can be a tiny share.
- Tiny samples. “I won 8 of my last 10” is easy to produce by luck. The role of sample size is the first thing to check, and a proper range around the rate would be very wide.
- Many tests, one report. If twenty betting or play approaches are tried and the best is shown, it will probably look good even if every one is no better than chance. This is sometimes called the multiple comparisons problem.
- Relative versus absolute change. “Doubles your chance” can mean going from 1 in 10,000 to 2 in 10,000. Both numbers matter.
- A favourable measure. A table might show wins per session while hiding total money lost, or quote the best single payout rather than the typical result.
- “Up to” and “as high as”. These phrases quote a ceiling, not an average.
Claims specific to gaming
Be especially cautious with these:
- Return and payout claims. A quoted percentage return is an average over a very long period of play. It does not describe what happens in one session, and it does not say you will not lose. For the real meaning of house edge and its relation to long-run losses, look at how the figure is defined and who calculated it.
- Screenshots and clips. A picture of a big win shows a result, not how many tries it took or what was lost on the way. With slot odds this matters a great deal because rare large payouts are exactly what get posted.
- Hot-streak and pattern claims. “This number has not come up in 40 spins, so it is due” is not statistics. Independent events have no debt. Belief in streaks in either direction is covered by the idea of hot hand belief.
- Esports or sports stats. Win rates built on small match counts and selected opponents can look decisive and mean little. This is a recurring issue with esports statistics.
- Systems that claim to beat the odds. If a method really beat a game with a built-in disadvantage, quoting a few good sessions would not be the evidence offered. Such claims often rely on the selection effects above, and no staking pattern can turn a game with negative expectation into a profitable one.
Questions to ask
- How many cases is this based on, and how many were left out?
- Over what period, and would a different start or end date change the story?
- Who is missing? Do I only see winners?
- How many things were tried before this one was reported?
- Is this a ceiling, an average or a typical case?
- Is the number compared with a sensible baseline, such as what pure chance would produce?
- Who is making the claim, and what do they gain if I believe it?
Question 6 is where a little simulation helps. You can run a Monte Carlo simulation of a coin or a dice game on your own computer to see how often random play produces an eye-catching result. Seeing that chance alone gives “winners” fairly often makes it easier to resist a persuasive story.
Keeping a fair mind
Being sceptical does not mean assuming every claim is false. Some numbers are honest and informative. The aim is to ask for the context that lets you judge them: the full sample, the method and the alternatives that were not shown.
A statistic that sounds like a reason to play, spend or stake more deserves extra care. Informed decisions about games depend on knowing that the odds are set by the rules and not changed by a clever-sounding figure. Outcomes are uncertain, losses are a normal part of games with the odds set against the player, and anyone who feels play is getting out of hand can reach out to a local support service.