How Our Predictions Work

For each game, we produce a single number: the estimated probability that the home team wins. We use three inputs, applied in order of reliability:

1. Market moneyline odds — when ESPN carries a consensus line, we convert it to implied probability and use that as the base.
2. Season record differential — if no odds are available, we estimate win probability from the gap between each team's win percentage.
3. Home-court/field advantage — we apply a small upward adjustment (~3 pp) for the home side. This is not applied at neutral sites or for UFC bouts.

We display a confidence tier (High / Medium / Low) that reflects the model's certainty, not its accuracy:

• High — live market odds available; win prob ≥ 65%.
• Medium — odds available or clear record differential; win prob 55–64%.
• Low — no odds; small record gap; near 50/50.

When odds are available, we show whether our model pick agrees or disagrees with the betting market. "Agrees with market" means the model and the market favor the same side. "Disagrees with market" is displayed as a note — not as a betting signal.

• It does not use machine learning or historical training data.
• It does not account for player-level injuries or line-up changes.
• It does not incorporate weather, referee tendencies, or travel fatigue.
• It does not predict margins, totals, or point spreads.

Team records, scores, and odds are sourced from ESPN's public scoreboard API. Data revalidates every 60 seconds during live games. ESPN is not a sponsor or affiliate of MatchupLens.

Disclaimer: MatchupLens predictions are for informational and entertainment purposes only. They are not betting advice and do not guarantee any outcome. Past model performance is not indicative of future results. If you choose to wager, please do so responsibly and within your means. Sports betting involves risk of financial loss. Responsible gambling resources →