The things worth being suspicious about, answered straight.
No. Every pick ever published is on the track record, winners and losers, with the odds taken and the final score. Results are settled by software from the final score — nobody types them in, and nothing is edited afterwards.
Picks are also posted to a public Telegram channel before kick-off, which timestamps them. You do not have to take our word for when they went up.
It is the wrong number to judge this on. At average odds of @2.89, a pick needs to land about 34.6% of the time just to break even. Our win rate is 37.4%.
Someone backing favourites at 1.40 needs to be right 71% of the time. Comparing the two percentages without the prices behind them is meaningless.
Because it is the outcome most people avoid. Bettors back a team; the draw is what you are left with when you have no side to cheer for. That reluctance shows up in the prices, and it shows up most in leagues nobody writes about.
Doing one market also makes the record honest: there is nowhere to hide a bad run behind a different strategy having a good month.
We are not trying to predict draws better than the market. A pick goes up when the price we can take sits above the market consensus for that same draw, and the model finds no reason for the gap. Each pick shows that premium — the edge is in the price, not the prediction. The full version is on the methodology page.
Then the edge may not be there. The odds published are the ones actually taken at the time of posting, and prices move. If your bookmaker is materially shorter, that pick is not the same bet.
Because that is where prices are loosest. In the Premier League thousands of people are correcting the odds every minute. In a second division on another continent, far fewer are — and the price stays wrong for longer.
A unit is our base stake. The whole record is published in units so it scales to whatever you bet: if a unit is £10 to you, multiply everything by ten. Publishing in euros would only tell you about the size of our bankroll, which is no use to anyone else.
Not every pick gets the same. Two units when the price is in the 2.6–3.2 band and above market consensus — the combination that has paid best over eight years of data. Half a unit when the price only matches the market. One unit for the rest. The stake says how much we like it, which is more honest than pretending they are all the same.
For us a unit is currently €130.00, because our unit is 2% of the bankroll and it moves as the bankroll moves.
Because it answers a different question. The unit ROI tells you what the picks did. The € curve replays those same picks with our real staking ladder — 2% of the bankroll, recalculated weekly from €10,000.00 — so each pick ends up weighted by the size of the bank on the day it ran. That makes the two returns differ a little: one measures the selections, the other measures the selections plus the staking. Both are computed from the same settled results, and neither is ever adjusted by hand.
No. Losing runs are part of it — betting draws at these prices means being wrong more often than being right, and stretches of consecutive losses are normal, not a sign something broke. Judge it over months, not days, and never stake money you need.
It costs nothing. There is no paid tier, no VIP group, no subscription and nothing to buy. If that ever changes, the free record stays free and public.
No. This is a public record of what one model does with one market. It is not advice, and nothing here accounts for your situation. Betting loses money for most people who do it.