Previews, reviews and explainers on how the models read the games.
Sunday night: a 12-12 split on the football card, and hockey graded nothing: Sunday finished 12-12 for about even across 24 graded NFL leans. The spreads went 5-2 for +2.5u, the totals went 6-5 for +0.5u, and the moneylines went 1-5 for -3.0u. Hockey played five games and graded no leans, so its season record is unchanged.
How to read a game page: every number on a matchup, explained: Every game page on this site carries the same six parts: the model's numbers next to the market's lines, the three chances, the label, the breakdown of where the number comes from, and the deeper tables underneath. A walk through tonight's Chargers-Ravens game and two games from the weekend shows what each part says.
How to read a playoff percentage: what 99%, 48%, and 4% actually mean: The site's NHL power pages carry a playoff percentage for every team. A percentage is a count of simulated seasons, not a forecast. How the ladder works, why points and rating disagree, and how fast the numbers move.
How football power ratings are built — and how to read a rating change: A power rating is one number per team, and the margin on a matchup page is the difference between two of them. How the model's five parts add up, why the quarterback line moves the most, and what a half-point rating change means, with real games from this week's NFL feed.
Which players play better with which balls?: Four Slams, four balls. Fresh balls favor the server, worn balls favor the returner, and the serve-plus-one numbers show which top-20 players gain most from a lively ball.
How balls and surfaces affect each other: A court has no speed without a ball on it. Court Pace Index readings from 2015 to 2026 show how the ball, the felt, the air, and the surface combine into one playing speed.
How a total is built: from a projected score to a line: A total starts as a projected score, not as a guess about the line. How the model builds its number from a league base and the two teams, how that number becomes a chance, and how the chance becomes a label, with real games from this week's feeds.
What the price really means: implied probability and the vig: Every price on a card is a probability in disguise. How to turn American odds into the chance the market is charging for, why the two sides of a game add to more than 100%, and what that costs over a season on this site's graded card.
Tuesday night: hockey went 7-2, and the college card went 2-2: Tuesday finished 9-4 for +4.0u on 13 tiered leans. Hockey went 7-2 for +4.0u with all three totals under, college football went 2-2, about even on units, on a three-game MAC card, and football had no game. The hockey season now sits at 214-214 for +0.5u.
Why hockey totals sit at 5.5 and 6.5 — and what moves one: Every total on tonight's NHL card is 5.5 or 6.5. That is not a quirk of one night: a typical game lands near 6 goals, and the line has to sit near the middle of the scoring distribution. How the number is built, why six goals is less likely than five or seven, and what has to change for the line to move.
What counts as a system: a rule, a test, and a bar it has to clear: A system is a rule written down before the games, applied the same way every time, and graded on a fixed test. Three rules from this site's own journal show why the rule matters more than the model behind it.
Sunday night: the totals went 6-3, and the moneylines went 0-4: Sunday finished 10-14 for -5.10u. Football totals went 6-3 for +2.38u while its moneylines went 0-4, and hockey lost all three of its graded leans in one game in Calgary. College football had no game on Sunday.
Why some games get no pick at all: Every game gets a model number on all three markets, but only the gaps above a floor carry a label. Sunday's NFL card had 21 labeled picks and 18 passes. What the floor is, why it exists, and what a pass means for the record.
Saturday night: college spreads went 17-23, and hockey went 5-4: Saturday finished 35-40 for -7.37u. College football spreads went 17-23 for -7.55u while its totals went 13-13, and hockey went 5-4 for +1.36u on a nine-lean card. Football had no game on Saturday.
Why this site publishes the tests that do not pass: Every idea on this site is tested against a bar written down before the test runs. When a test comes up short, it stays on the record anyway. Here is why, with the NBA backtest, a losing hockey tier, and nine nights that stayed on the books.
Week 9 line moves: Baltimore swung 14 points on a quarterback change: Fifteen Week 9 games, open to now: the spread moved 2 points or more in six games and the total moved 3 points or more in nine. Two quarterback changes drove the two biggest moves, and five of the six biggest spread moves crossed the model's number.
Friday night: the college spreads went 3-0, and hockey went 1-3: Friday finished 5-6 for -1.45u. College football spreads went 3-0 for +2.73u while its totals went 1-3, and hockey went 1-3 for -2.09u on a four-lean card. Football had no game on Friday.
Thursday night: the slight tier went 1-10, and the card finished 5-15: Hockey went 4-10 for -6.39u as the slight tier won once in eleven tries, Jacksonville beat Baltimore 33-10 with the Over 39.5 winning in the strong tier, and college football went 0-3 as both Thursday totals went over.
Season check: the puck line is carrying hockey, the spread is carrying football: Through Thursday, hockey stands at 197-199 for -1.30u and football at 77-82-1 for -0.98u. One market carries each card: the puck line at +12.81u in hockey, the spread at +6.98u in football. The week's strength split has the slight tier at 9-16 for -6.85u in hockey.
Goaltending in hockey: how one name in net moves a moneyline and a total: A goalie plays every minute and faces every shot, so the name in net is the biggest single-player input on a hockey card. How the model reads a starter, a backup, and a late change, with real examples from this week's feed.
What a unit is, and why every result here is counted in units: A unit is the fixed amount risked on each graded pick. How price turns a win into +0.42u or +3.80u, why a split record can finish at -3.36u, and how to read a season record in units.
What a trend is, and why a trend is context rather than a pick: A trend row is a record of how similar spots have played out before. What each row says, why only the loudest rows appear, and how to read one beside a pick without letting it change the pick.
How pick strength is decided: what slight, moderate, and strong mean on a card: Strength is the size of the gap between the model's number and the market's number. How the three labels are set in each market, why the price does not set strength, and why a strong label still loses, with locked examples from hockey, football, and college football.
How line movement works: what a moving line is telling you, and what it is not: A line is a price, and prices move when news arrives and money arrives. Three real moves from this week show how a spread, a total, and a moneyline each move on their own, why the locked line is the only line that counts for grading, and why a move is information, not a result.
How to read an NHL power rating: one number, three parts, and a whole season behind it: A power rating compresses attack, defence, goaltending, and finishing into one number. How the rating is built, why it disagrees with the standings, how it becomes a game prediction, and what the season simulation does with it, using the ratings as of October 30.
How to read a spread, a total, a moneyline, and a puck line: One game, four different questions. What a spread, a total, a moneyline, and a puck line each ask, how the half point works, and how to read the model's number against the market's number, with locked lines from two real games.
What variance is, and why a 60% model still loses 4 of 10: A 60% edge means losing 4 of every 10 picks on average. The math of 10-bet and 100-bet stretches, the juice bar, and what a real losing week looks like on this site's own record.