Tipster record checker — skill, or luck?

Enter the record somebody is selling: wins, losses, pushes and the price they usually bet at. This works out what it would be worth, and how often a tipster with no edge at all would post a record that good by luck.

Win rate

58.0%

Break-even at these odds

52.4%

Units (1 unit a bet)

+10.73u

ROI, flat stakes

+10.7%

Chance by luck alone

15.3%

If this tipster had no edge at all — a true rate of exactly 52.4%, the price's break-even — a record at least this good would still turn up 15.3% of the time. Of a thousand tipsters with no edge, about 153 would post it.

Could well be luck. Better than most no-edge records, but nowhere near rare enough to say.

A few hundred decided picks is a real sample, but still a short one — streaks of this length come and go.

All of this assumes the record is complete: every pick, posted before kick-off, nothing deleted and no second account quietly dropped. A record that cannot show that is a claim, not a record — this is why we seal every pick before kick-off. How a sealed pick is checked · How our pick is chosen

How often luck alone produces a record, at −110

Each cell is the record at that hit rate, and the chance that a tipster with no edge at all, betting into −110, posts one at least that good.

How the check works

At American odds the break-even rate is 1 divided by the decimal price: −110 is 1.909, so a tipster has to be right 52.38% of the time just to stand still. Pushes are left out, because they return the stake.

The luck figure is a one-sided binomial test. It assumes the tipster's true rate is exactly the break-even rate — no edge at all — and adds up the probability of every record at least as good as the one claimed. A small number means the record is hard to reach by luck; it does not measure skill, and it means nothing if losing picks were left out.

Sample size matters more than the percentage. At −110, 11 right out of 20 turns up by luck about half the time, and 55 out of 100 about a third of the time. Even 550 out of 1,000 turns up by luck about one time in twenty; 600 out of 1,000 almost never does.

Ask for the picks, not the summary: every one, with a timestamp from before its game, somewhere the tipster cannot edit afterwards. A record put together after the results are known can be any record at all — which is why we publish a hash of each pick before kick-off.