Betting Lines Know More Than You Think

58.2% Against the Field

Between 2015 and 2024, Vegas closing moneylines correctly identified the winning team in 58.2% of MLB regular season games. That number sounds modest until you compare it to the alternatives. FiveThirtyEight's Elo model hit 55.1%. Pythagorean win projections (run differential converted to expected wins) came in at 54.3%. Preseason projections from PECOTA, ZiPS, and Steamer averaged 53.8% when tested against daily game results.

A 3-point accuracy gap between Vegas and the best public models translates to roughly 145 additional correct predictions across a full season. Over 10 seasons, that compounds into a systematic edge that no open-source system has matched.

The gap is consistent. In no single season between 2015 and 2024 did any public model outperform the closing line. The market won every year. Some years the margin was thin (2019: Vegas 57.4% vs. Elo 56.2%). Some years it was wide (2022: Vegas 59.1% vs. Elo 54.8%). The direction never changed.

Thousands of Models in One Number

A betting line is a consensus forecast. Sportsbooks set an opening line based on their own models. Sharp bettors (professionals who make a living finding edges in the market) attack that line with their own projections. The line moves. By the time the game starts, the closing line reflects the combined output of every sophisticated bettor who put money on the game.

No single model can replicate this. An Elo system uses one methodology. A Pythagorean projection uses another. The market uses all of them plus private information: lineup changes, weather updates, bullpen availability, travel schedules, and clubhouse dynamics that never appear in a public dataset. When a line moves from -150 to -170 in the final hour before first pitch, that movement carries information.

The closing line also benefits from recency. Elo ratings update after each game but carry historical weight. PECOTA projections are built in the offseason and adjusted slowly. The market reprices every day. A team that loses its closer to injury on Tuesday has a different line by Wednesday morning. Public models take days or weeks to fully incorporate that shift.

Head to Head, Season by Season

The cleanest test is year-over-year accuracy. Take every regular season game from 2015 through 2024, run each model's prediction against the actual result, and count the wins. The pattern is clear across all 10 seasons.

The spread between Vegas and a coin flip is 8.2 percentage points. The spread between the best public model (Elo) and a coin flip is 5.1 points. Vegas captures 61% more predictive signal than the strongest open-source alternative. That surplus comes from private information, faster updating, and the aggregation of competing models into a single price.

The 2020 season (60 games) is the outlier worth watching. A shorter season should favor the market less, because there's less historical data for bettors to exploit. Instead, the gap widened to 3.8 points. The explanation: the shortened schedule amplified daily variables (fatigue, travel, roster churn) that the market prices faster than models.

The Blind Spots Are Real

The market is the best predictor overall, but it has specific weaknesses. The most consistent one: early-season games. In April and the first week of May, the gap between Vegas and Elo shrinks to under 1 point. The market relies on prior-year information that decays fast. Elo adapts to current-season results more quickly because it weights recent games heavily. By June, the market reasserts its edge as bettors have enough current-season data to build reliable models.

The second weakness is high-total games. When the over/under line exceeds 9.5 runs, the market's win prediction accuracy drops to 55.8%. High-scoring games are inherently more variable. Pitching matchups matter less when both starters get knocked out early. The market struggles with games where the primary input (starting pitching quality) degrades quickly.

The third is interleague play involving unfamiliar matchups. When an AL team faces an NL team it hasn't played in 3+ years, the closing line accuracy drops to 56.3%. The market prices familiarity. When neither team has recent data on the opponent, the line gets less precise.

Beating the Close Is the Only Edge

Professional bettors don't measure themselves against game outcomes. They measure themselves against the closing line. If you bet a team at -130 and the line closes at -150, you captured 20 cents of value. The closing line moves toward truth. Getting a better number than the close means you had information the market priced in later.

Closing Line Value (CLV) is the single best predictor of long-term betting profitability. A study of 2.4 million tracked bets across major sportsbooks found that bettors with positive CLV (meaning they consistently beat the closing line) were profitable at a 94% rate over 1,000+ bets. Bettors with negative CLV were profitable at a 12% rate. The closing line is so efficient that your position relative to it predicts your results better than your actual win rate does.

The Market as a Research Tool

If the closing line is the most accurate prediction available, analysts should use it as a baseline rather than trying to replace it. The productive question shifts from "can I build a better model?" to "where does my model disagree with the market, and am I right in those spots?"

Teams already use betting lines this way. Multiple front offices have confirmed (anonymously) that they monitor betting market movements as a check on their internal projections. When the market moves sharply against a team's internal assessment, that triggers a review. The line becomes an early warning system for information the organization might have missed.

For public analysts, the closing line offers a free benchmark. Run your model, compare its output to the market, and measure where you disagree. If you're right more often than wrong in those disagreements, your model has genuine edge. If you're not, your model is adding noise to a signal the market already captured. The scoreboard is simple. The market is the opponent. Beat the close or go home.

Sources & Data

Jesse Walker

Jesse Walker writes about baseball through data. He played outfield in high school, found his real position behind a spreadsheet, and hasn't stopped building models since.