Backtesting that holds up
A backtest is only worth what it predicts. WinPicks grades every system the way a professional would: on data it has never seen.
The number that separates the keepers
Sweep 9,317 combinations against a database and roughly 424 will clear a two-sigma threshold by chance. That is not a flaw in the data, it is arithmetic — and once you know the figure, you know exactly how high a real result has to clear.
WinPicks puts that bar on the screen next to your results, so a genuine find announces itself instead of hiding in the crowd.
Four numbers you get back every time
Where the bar sits
Every scan reports how wide it looked, so you can weigh what you found against what chance alone would produce. That is what turns a list of results into a verdict you can act on.
How it held up out of sample
Fitted on one block of seasons, judged on another the search never saw. Clear both and you are holding something that has already proved it travels — which is the only kind worth taking into a season.
What the ride actually feels like
A system is only worth what you can hold through. The deepest drawdown and the longest losing run sit right beside the record, so you can size the bet for the season you are really going to have.
What it did at the prices that traded
Graded against the archived prices wherever we hold them, rather than a flat −110 stand-in — so the return on the screen is the return you would have taken. When the two figures differ, that gap is often the most useful number in front of you.
Why the edge is in your ideas
We ran four separate studies against the raw box-score fields — scores, rest, streaks, home field — and the market has had every one of them for forty years. A closing number already contains that much.
Which is precisely why WinPicks is built the way it is. The value was never in mining the same four columns harder than the next man. It is in the situations, filters and weightings you bring, tested properly against forty seasons, because those are the ones nobody else is running.
We tell you where the dead ends are. A tool willing to mark its own is a tool you can believe when it says something survived.