Loading the slateGames, win odds and scorer projections appear here once the data loads.
Who wins and who scores, for every NFL, NBA and MLB game on the board.
NFL scores. Built from play-by-play efficiency, not the betting line. Each offense and defense is rated on three signals, each regressed toward average by games played and converted to points: EPA per play (50%; expected points added, the best single measure of moving or stopping the ball), success rate (20%; the share of plays that keep a drive on schedule, the steadiest week-to-week stat) and touchdowns scored or allowed per game (30%). A team's projected points = league average (22.8) + its offense + the opponent's defense, ± 0.75 for home field. Extreme projections are pulled 20% toward average, which keeps totals realistic.
NBA scores. Offensive and defensive rating (points per 100 possessions) and pace from this season, blended with last season's strength (its win total turned into a net rating) worth 10 games. That blend moves toward this season's numbers every night. Projected points = expected possessions × (league efficiency + team offense + opponent defense) / 100, ± 1.25 for home court.
Simulation. Every game is played 20,000 times. NFL games are simulated drive by drive: about 11 possessions a side, each ending in a touchdown, field goal or nothing, with odds tuned to each team's projected points and how TD-heavy its offense and the opposing defense are. Extra points, two-point tries, return touchdowns, safeties and overtime are all played out, which is why NFL finals land on real scores like 24–20 and 27–17. NBA games vary possessions and shooting nights around each team's projected efficiency, with overtime played out.
Predicted final. The typical score in the simulations the favorite wins, snapped to the most common real scoreline nearby. "Avg" is the plain average across all 20,000 games, upsets included, so the predicted winning margin is usually bigger than the average one. "Usually" is each team's middle-80% scoring range. The five boxes show how often the favorite wins by each margin, and how often the underdog wins.
Win probability is the share of simulations each team wins. The betting feed's odds are shown only as a comparison, never as an input; a gap of 10+ points is flagged as a Disagreement. The market is usually the sharper of the two, so treat a disagreement as a reason to check injury and lineup news, not as a better bet. A Strongest pick is a game where the model and the market pick the same team and both give it 70% or more.
Record. Every prediction is saved before tip-off or kickoff (the latest one before the game starts is the one that counts) and graded against the final score. Player picks are graded against the box score when one is available; a player missing from an incomplete box score is shown as "not in box score" rather than as having sat out.
NBA scorers. Each player's season points per game, stabilized with two phantom games at league-average scoring for his minutes, then scaled to tonight's projected team total. That way a fast pace or weak opposing defense lifts everyone. Threshold odds assume a game-to-game spread of 28% of the projection plus 3.5 points.
Injuries. Each day's injury report is applied before the games are simulated. Out and IR count as missing; doubtful players miss 80–85% of the time; questionable players are counted as sitting 45% of the time in the NBA and 25% in the NFL (most NFL questionables play). In the NBA, a missing player lowers his team's offensive rating by his scoring above what a replacement gets in his minutes (up to 7 points per 100 possessions for a star), counted only for the share of games he's actually played, since earlier absences are already in the team's numbers. His points are handed to the teammates who play. In the NFL, a missing scorer's touchdowns go 60% to teammates and 40% off the team's total. Quarterback injuries aren't modeled beyond what the team's recent EPA already shows.
MLB. Uses the record and runs-scored model until the season starts.
Data. NFL schedule, scores, lines and play-by-play from nflverse; NBA schedule and box scores from ESPN via hoopR; odds and injury reports from ESPN. Refreshed every two hours.
NFL anytime TD. Built from season usage, not the spread alone. First, the team's expected offensive touchdowns for this game: projected points × 0.105, so a strong defense on the other side lowers everyone's odds. Each player's slice of those touchdowns comes from two things, weighted 60/40:
The two add to an expected TD count λ, shown as the chance of at least one: 1 − e−λ. Usage covers each league leader list's top 25 (red-zone carries, red-zone targets by position, rushing and receiving TDs). Players outside those lists don't appear, and injuries aren't applied yet.
MLB HR and RBI. Per-game home run and RBI rates, blended with league-typical rates (0.12 HR, 0.5 RBI per game) over three phantom games, scaled by projected team runs.
Player numbers sharpen as more games are logged each day. These are estimates for fun and research, not betting advice.