How we make our leans

For every game in our leagues that has a betting line, we work out our own fair line: the spread and total we think are right. Then we compare it with the market's line, the middle of what the sportsbooks are offering.

• If the gap is big enough, we publish a lean on the side or the total we think the market has wrong.
• If it isn't, the card says "No lean: the market looks fair at {line}" and still shows our fair line.

Every lean is graded in public, wins and losses alike, on a record that never resets.

Ratings (who wins, and by how much)

Every team has a rating. The gap between two teams' ratings, after a few adjustments, is the margin we expect.

• Home advantage. The home team gets a bonus, set separately for each league. At a neutral site (bowl games, conference tournaments, games abroad) nobody gets it.
• Rest. We tested rest in every league and kept it only where it clearly improved our numbers.
  • A team on the second night of a back-to-back gives up about 3 points in the NBA, about 2 points in college basketball, and about 0.4 goals in the NHL.
  • Short weeks and byes in football made too little difference to keep.
  • In baseball, who pitches matters far more than rest, and our ratings don't see pitchers.
• After every game, the winner takes rating points from the loser.
  • A bigger win moves more points, but each extra point of margin counts for less than the one before.
  • A strong favorite that wins as expected moves fewer points, so blowouts against weak teams don't inflate a rating.
• Between seasons, ratings move part of the way back to average, because rosters change. How far is set by each league's own history: in college basketball, last season's rating carries over in full.
• Early in a season, every team keeps last season's rating until its new results move it.

This is a well-known method called a margin-of-victory Elo rating. We use it because it is simple and it is open: given the same results, anyone can rebuild our ratings exactly.

Totals (how many points)

Each team also has:

• an offense number: how many more points it scores than an average team;
• a defense number: how many more points it allows than an average team.

Expected points. A team's expected points are the league average for a home or away team, plus its offense, plus the opponent's defense. Our fair total is the two teams' expected points added together.

After each game, the difference between what a team scored and what we expected is shared between its offense and the opponent's defense. The league averages update too, so the model follows scoring trends from season to season. A baseball game called before 9 innings doesn't change these numbers, because its score was cut short.

Puck lines and run lines

In hockey and baseball the side bet is one team at −1.5 and the other at +1.5. The line never moves, only the price. It is usually the favorite that lays the 1.5, though in close baseball games books often give it to the road team.

• From past games, we know how often a team expected to win by a given margin actually wins by 2 or more.
• That turns the market's price into the margin the market expects. We compare that margin with ours, in goals or runs.

We never lean on the moneyline. In the NHL and MLB we lean on the puck line, the run line or the total. Every lean can then also be a streak pick.

Which teams we rate

• We rate teams in the six leagues we cover: the NFL, NBA, NHL and MLB, FBS college football and Division I men's college basketball.
• When an FBS or Division I team plays someone from a lower division, we don't have enough games to rate the opponent. Those games get "No lean".

The consensus. For each game we take the line from each sportsbook in our feed, and use the median, the middle value, so one odd book can't move it.

Removing the bookmaker's margin. Books charge a margin, the "vig". So we first adjust each book's line for its prices:

• An over priced at −130 means that book really expects a little more than its posted total.
• We move the line by that amount before taking the median.

The result is the median no-vig consensus: our best reading of what the market as a whole expects, before the books' margin.

When we read it. We read the market at 10:00 AM ET on game day. If a game's first line appears later, we read it when it appears.

We lean when our fair line and the market's line differ by at least:

One lean per game. If both the side and the total pass, we publish the one with the bigger gap relative to that league's threshold. If the two are level, we publish the side.

Confidence is set only by the size of the gap:

• Low: the gap is at least the threshold.
• Medium: at least 1.5 times the threshold.
• High: at least 2 times the threshold.

After 200 graded leans in a league, we check on the record whether High really wins more often than Medium and Low. We publish the result and adjust the tiers if it doesn't.

A published lean never changes.

• It doesn't switch sides, and it doesn't turn into "No lean".
• If the market later moves past our fair line, the card says "Line moved since we published", and the lean is still graded as published.

Our editor can withdraw a lean before the game starts, for example when big injury news breaks. The editor can't create one.

• A withdrawn lean stays on the record, marked "Withdrawn by editor" with the reason. It isn't graded.
• We show how often we withdraw.

• The line. Each lean is graded at the line we published: the middle book line. When two lines sit in the middle, we take the one that is worse for our lean.
• The price, for one unit risked:
  • NFL, NBA and college: at −110. A win pays +0.91 units; a loss costs 1 unit.
  • NHL and MLB: at the published consensus price for that side. The price is most of what a puck line or run line is, so −110 would misstate the result.
• Push and void. A push is a push. A game that is postponed or canceled is void. In baseball, a game called before 9 innings is also void, as the books treat it.
• Closing line value (CLV). We also compare our published number with the closing consensus, the market's last line before the game.
  • Beating the close regularly is the best early sign of a method with an edge. Wins and losses take hundreds of games to mean much.
  • For spreads and totals we show CLV in points. For puck and run lines we show it in percentage points of the chance to cover.

Our leans use only results, the schedule, venues, rest and the market. We don't use:

• injuries, lineups or rest days for stars;
• starting pitchers or goalies;
• weather;
• betting splits;
• news.

The market does see these things. So when our line is far from the market's, the market often knows something we don't. Our confidence tiers are checked against the record for exactly that reason.

Before going live we tested the method on every game from the 2023 seasons to October 2026, using only games played before each one. More than 30,000 leans were tested across six leagues.

The short answer: it doesn't beat the market.

• Graded at the closing line, our leans won between 49.0% and 52.2% in every league. At −110 you need 52.4%, and hockey and baseball prices need more.
• Our fair lines were less accurate than the closing market in every league.
• Larger gaps did not win more often than smaller ones.

Against the earlier opening lines it did better.

• NBA and NHL leans finished close to break-even.
• Leans on NBA totals, college football totals and MLB run lines beat the closing line about 60% of the time.

That's a hint of value early in the market, not an edge we claim.

The full test, with every number and its sample size, is published in our back-test report.

Back-tests never count toward our record. The record starts with the first lean we publish. Our method's parameters were tuned on 2019–2022 results only, never on betting lines. They are published with the model version, elo-2026.1.

• A lean is our view of where the market is off, not a promise.
• At −110 you need to win 52.4% of bets just to break even. Most bettors, and most methods, don't manage it over a long run.
• If you bet, set a budget you can afford to lose. If gambling stops being fun, call 1-800-GAMBLER.