Why we publish our losses — and what honest statistics look like
How "statistics" get made in tipping
There are only a few ways to dress up a record, and all of them are simple.
Delete the losses. The tip goes out in a private channel, the bet loses, the message disappears. By the end of the month the feed is all winners.
Start counting from a convenient date. Pick a good stretch and present it as the overall result. A bad month simply never enters the sample.
Show the hit rate without the prices. "85% winners" at odds of 1.10 is a loss, but the headline does not say so. More on this in the break-even rate.
Lean on a tiny sample. "6 from 6, 100%" is not statistics, it is a coincidence.
None of these requires an outright lie. It is enough to choose what to show.
What we do
Every signal is written to the database before kick-off and stays there regardless of the result. Nothing is deleted or edited after the fact — losses sit alongside wins on the results page.
On the statistics page the average price sits next to the hit rate, and the break-even rate next to both, so the number cannot be read out of context. Markets with fewer than ten bets are moved into a separate block with a plain warning that the percentage there is noise.
The month-by-month record is the uncomfortable part
A single all-time figure averages things out and hides the direction. Ours:
- March 2026 — 20 bets, ROI +26.5%
- April 2026 — 160 bets, ROI +14.4%
- May 2026 — 87 bets, ROI +9.7%
- August 2026 — 4 bets, ROI −8.7%
The trend is downward. We could show only the overall +13.5% and never technically lie. But someone subscribing today is not buying March — they are buying what happens now.
The explanation is straightforward: March had a small sample (20 bets), and a high percentage there is largely luck. As the number of bets grows the result converges toward a more realistic level. August is the start of the season with only 4 bets, and nothing can be concluded from those either.
Why we bother
The practical reason: we do not know what next month brings. The model can produce five losses in a row — that happens to any system running at 69%. If the reputation rested on an edited feed, the first such run would destroy it. When the record is open from the start, a losing run is part of the chart rather than a catastrophe.
And one more, purely arithmetic. Betting is not a way to get rich quickly. ROI of +11–13% over distance is a good result, but it means a $600 bankroll returns roughly $70 across a hundred bets. Anyone promising more is either showing a short lucky stretch or not showing the losses.
All figures are on the statistics page. They recalculate automatically after every match, and we cannot touch them by hand.