What moves a game before it starts: pitchers, goalies, quarterbacks and rest
Published October 19, 2026
Team strength explains most of a game's odds. But some news arriving in the last days or hours before a game moves them a lot more than others. This is what the evidence says, sport by sport, and what iPick does with it.
NFL: the quarterback above everything
No single player moves a game like an NFL quarterback. Bookmakers have described moving a line by about 5 to 8 points when a star starter is replaced by a backup, less when the backup is good (ESPN). The market prices this well: over a decade, backups facing a starter early in the season won only 39% of their games but covered the spread 52% of the time (Action Network).
Rest matters much less than it used to: the bye-week advantage, about 2.2 points per game before 2011, has been about 0.3 points since practice time was cut. Home-field advantage has also shrunk, to roughly 1.5 to 2 points.
iPick's NFL model reads who starts at quarterback; it lowered the model's error clearly in testing (twelve seasons, 3,266 games).
MLB: the starting pitcher
In baseball the starting pitcher is the biggest game-specific factor. In FiveThirtyEight's former model, an ace was worth about as much as moving a team up a full tier. iPick's MLB model takes in the announced starter's record, and in testing over ten seasons and 22,375 games it improved the model in every season.
Parks and weather mostly change how many runs are scored, not who wins: each extra degree Celsius adds about 2% to home runs (Dartmouth). Playoff baseball is even closer to a coin flip than the regular season, because the teams left are similar.
NHL: the goalie, less than you think
A goalie change is the hockey news everybody watches, yet a sportsbook's head of hockey put it at “at most” 3 to 5 percentage points of probability in most cases (ESPN). Goalie performance is also very noisy from season to season.
iPick tested it: rating goalies by goals saved above expected added almost nothing to its model, even using the goalie who actually started — so it is not used. What did help was the schedule: the second game of a back-to-back, games in the previous six days and distance travelled, which are now part of the NHL model.
NBA: who plays, and when
In basketball, a star sitting out moves a game more than anything else, and the league's injury report is the key document: teams must file it by 5 p.m. local time the day before a game. iPick's NBA model reads who is out.
The second night of a back-to-back costs a team about 1.5 to 2.2 points of net rating (SportsBettingDime), and Denver has the strongest home court in the league, 8.6 points of net rating better at home than on the road since 1999 (Sportico). iPick tested back-to-backs, travel and altitude on top of its rest-days measure over eleven seasons: they added nothing it did not already know, so they are not used.
The common thread
The information that matters most — who starts, who is out — arrives late, and the market absorbs it quickly. That is why iPick recalculates its pick every few minutes until the game starts, and why the pick shown when a game begins is the one it is judged on. What each league's model reads, and what it does not see yet, is on What each model knows.