How our picks are made.

No gut calls, no “locks,” no cherry-picked screenshots. Every Da Vinci Bets pick comes out of the same pipeline: a multi-factor model, a calibrated confidence score, a quality gate — and the best available price across approved books.

1. Factor models, not hunches

Team-sport picks today come from a factor-scoring model: it rates both sides on 15-20 factors per game — team strength ratings, current form, injuries and rest, matchup history, home/road splits, and market context — with set weights. The gap between the two scores sets the side and a starting confidence, which is then calibrated and bounded (per-market floors and ceilings, plus penalties in weak confidence bands and where the model disagrees with the market) before a pick is served. When the gap is too small, or the price leaves no qualified edge, the game gets no pick. Tennis match winners come from our own machine-learning model, trained on years of ATP and WTA results. We currently publish moneyline markets across NBA, NFL, NHL, and MLB, plus ATP/WTA tennis match winners — the paused markets below are excluded. We don’t offer player props.

2. Calibrated confidence

A raw model probability isn’t a confidence score. Before anything is published, model output passes through a calibration layer — isotonic calibration fit to how past predictions actually resolved, plus per-sport confidence policies (floors and ceilings) that our team maintains from graded results. The goal is simple: when we show a confidence number, it should track reality, not flatter the model. Only the strongest picks earn the HIGH CONVICTION tag, which requires at least 62% calibrated confidence.

3. Quality gates — and paused markets

Picks below our minimum confidence and edge floors are never shown. And when a whole market runs below our accuracy bar over a graded sample, we pause it entirely rather than keep publishing losers — the page shows an honest empty state instead of a pick we wouldn’t bet ourselves.

Currently paused

  • NBA spreads
  • NBA totals
  • NFL spreads
  • NFL totals
  • NHL spreads
  • NHL totals
  • MLB spreads
  • MLB totals

Paused markets return only when they clear the bar again on graded results. This list is generated from the same configuration that gates the live product.

4. Every pick comes with model analysis

Automated settlement compares predictions with official final scores for internal model evaluation. Public pages provide current picks, confidence scores, and matchup analysis.

What we do not do

  • No guarantees

    Sports outcomes are inherently uncertain. A 62% confidence pick loses often — that's what 62% means. Past performance never guarantees future results.

  • No “locks”

    We publish probabilities, not certainties. Anyone selling a “guaranteed lock” is selling something else.

  • Not a sportsbook

    Da Vinci Bets is an analytics product. We don't accept, place, or facilitate wagers of any kind.

  • Ongoing evaluation

    Internal evaluation helps us monitor model behavior and review quality gates.

21+ (or the legal gambling age in your jurisdiction). If you choose to bet, bet responsibly — and if gambling stops being fun, call 1-800-GAMBLER or visit our responsible gaming resources.

About Da Vinci Bets

Da Vinci Bets is operated by Da Vinci Bets LLC, based in Orlando, Florida. We build and maintain the prediction models, calibration systems, and settlement pipeline described on this page. Model evaluation informs internal quality controls.

Questions about our methodology, data, or model evaluation: support@davincibets.io