Ask it the way you would say it
“Home dogs after two straight losses.” “Teams off a bye, split by line size.” “Does it hold up out of sample?”
Type it, and WinPicks AI writes the query and calls the same C engine as every other screen — the one that has been grading these databases since 1986. The answer comes back as the app's own grid, sample size and out-of-sample column attached. The model writes the sentences; the engine supplies every figure in them, and even the prose is checked against the engine's output before you see it.
It runs the real engine
WinPicks AI is not a model guessing about football. It cannot see a game, a price or a record. What it does is write the query, call the same C engine every other screen calls, and hand back the app's own grid — games, record, percentage, units, drawdown, longest losing run.
The model writes the reading. The app draws every figure.
It cannot invent a number
Every statistic in every answer is checked against what the engine actually returned — rates, records, unit figures, decimals. Anything that cannot be traced back is flagged on screen rather than quietly printed, and the tables beside it are always the real ones.
It always tells you how strong the answer is
Ask about a filter with nineteen games behind it and it says so, in plain words, so you know exactly how much weight to put on it. Ask for a scan and it reports how wide it looked before it found what it found.
What it is for
Exploring
Ideas you would never bother writing a query for, because typing the question takes four seconds.
Checking
“Does this hold out of sample?” is one button under every result.
Learning the language
Every answer shows the query it compiled, so the clause language teaches itself.